Home

PAUP 3.1 Manual

1. NYUN DOWN EN RW Y Y Y MN au wo Note that these minimum and maximum lengths are not the same as those generated by the table of branch lengths and linkages which only consider MPRs The retention index is undefined for uninformative autapomorphic characters this is indicated in the matrix by 0 0 The consistency index is not calculated for stepmatrix characters due to the difficulty in estimating the minimum amount of change for those types of characters Character State Reconstructions PAUP will also display reconstructed character states for the internal nodes of a tree according to the optimization criterion currently in effect It will also display the possible reconstructions over all MPRs In this way you can compare the reconstructions required by the current tree and optimization with the states possible over all most parsimonious reconstructions There are several ways to obtain this information assigned character changes are displayed superimposed on the current tree s when CHGPLOT or character changes selection in the Describe Trees dialog box is selected Selected reconstructions will be output with a tree for each character States for the terminal taxa are also displayed NOTE If no characters are specified for either a CHGPLOT or POSSPLOT command the characters are taken to be those plotted in the last invocation of
2. The Warnings amp Errors dialog box The only item requiring further explanation is Warn before resetting the active datafile The active datafile is the file containing the most recently processed DATA block If another data block is encountered when executing a file the previous data matrix and any trees that may have been generated using this data matrix will be cleared from memory By default PAUP warns before resetting the active data file on the chance that the Execute command was issued inadvertently If you wish to suppress this warning uncheck the Warn before resetting the active datafile item NEXUS Format This option allows you to choose different options for NEXUS files Block termination Require ENDBLOCK Allow END as abbreviation All digit taxon and character labels Don t allow Allow but issue warning Allow do not issue warning Cancer The NEXUS Format dialog box In this dialog box you can choose to have end be a valid block terminator in the file as well as endblock You can also choose to allow or disallow all digit taxon and character labels PAUP 3 1 USER S MANUAL 245 Note MacClade uses end to terminate blocks If you created your data matrix with MacClade you must allow end to terminate a block otherwise PAUP will not recognize end as a block terminator Startup Preferences Use this command to change the default settings that
3. The Lake s Invariants dialog box Lake s method uses quartets of taxa at a time You have three options for searching and in each you must select which taxa will be included or excluded First you can explicitly select four taxa to make up one search quartet When you do this you must select the taxa you want and click on the Include button You can also double click on a taxon to include or PAUP 3 1 USER S MANUAL 259 exclude it The second option is to evaluate all possible quartets In that case you must choose which taxa will make up the group from which quartets will be taken The third option is to partition the taxa into four groups which will then make up the quartet In that case you will have to select which taxa will be included in the groups designated Group A through Group D Once a taxon has been assigned to a group it is no longer available to be assigned to subsequent groups You can also specify the output that will appear in addition to the invariants spectral components branch lengths or just summary tables See the section Lake s Invariants for more detail on this type of search and the output it produces Bootstrap The Bootstrap command implements Felsenstein s bootstrap approach to placing confidence estimates on groups contained in the most parsimonious trees Bootstrap Options Number of replicates 100 pType of search i Random number seed gt dar 3 OBranch and bound E Show
4. AAA The Apple 6 menu Along with your usual set of desk accessories the Apple menu contains choices for the About PAUP window which displays among other things the current version number The help system is no longer located in the Apple menu instead it is found in the Windows menu see below THE FILE MENU New The File menu is used to create open convert and save input output files to control certain printing options and operations and to exit the program The options available on the File menu are nm escrbedbelow Page Setup Print File Echo to Printer Print Selection Log Output to Disk Execute manual Export File Import File The File menu Use the New command to open a new untitled editor window e g to create a new PAUP input file See The PAUP Editor for more information on how to use the editor Open Use the Open command to execute or edit an existing PAUP input file On the standard file dialog box that appears choose either Execute or Edit as the initial mode then select the name of the file you want to open and click Execute Edit or just double click on the file name Note that by default PAUP shows all files To restrict the display to files created by 226 Close Save PAUP 3 1 USER S MANUAL PAUP or MacClade click on the Show PAUP MacClade files only box A sample file dialog box is shown below 3 Sample data files v Hard Disk D 5
5. tilde character For example PAUP 3 1 USER S MANUAL 71 symbols 0 9 A F is equivalent to symbols 0123456789ABCDEF Note that the tilde is used rather than the hyphen minus sign in order that present and absent may be used to designate character states SYMBOLS as in the following example format symbols matrix Taxon_one E Taxon_two AD a Taxon_three Taxon_four id By default PAUP does not distinguish between the lower and upper case representations of the same alphabetic characters Thus if you set the SYMBOLS list to abcd A is equivalent to a B to b and so on If for some reason you want to treat the lower and upper case representations as different character states you can use the RESPECTCASE option For example the command format respectcase symbols ABCDabcd defines a SYMBOLS list containing eight distinct character state designations The RESPECTCASE option pertains only to symbols used to represent character states in the data matrix Commands taxon and character names etc can still be entered using any combination of upper and lower case characters Using alphanumeric character names PAUP automatically uses the integers 1 through NCHAR to identify characters for input and output purposes You may also provide alphanumeric character names to supplement the numeric identifiers using the CHARLABEES command For example the command charlab
6. PAUP 3 1 USER S MANUAL 1 ii Several of the newer features of PAUP have been heavily influenced by many hours of lively discussion with David and Wayne Maddison and by their truly outstanding program MacClade David served as a consultant during the early stages of development of the Macintosh version of PAUP and has patiently provided advice assistance and testing throughout the development of PAUP Version 3 In addition to countless ideas for interface and documentation improvements David suggested significant improvements in tree filtering filtering by constraints stepwise taxon addition random addition sequence enhancements and tree file operations most importantly the use of Boolean operations for combining trees in memory with trees in a file Wayne was also heavily involved in the testing and made many additional suggestions for improvements including introducing me to the notion of user defined character types and sharing algorithms for dealing with multistate taxa I have also benefited tremendously from discussions with Joe Felsenstein particularly with respect to the branch and bound algorithm The unselfishness of these individuals who after all have a vested interest in the advancement of their own programs is a constant source of inspiration to me Many other individuals have reported bugs made helpful suggestions and or provided stimulating ideas These include Larry Abele Vic Albert Jim Archie Paul Berti
7. He ristie Searches its ici ve sdebiseseb cade EN ei 100 Branch and Bound Search ceeeeseesseceseeeeeeeeneeenaeenes 104 Exhaustive Search cds 105 Lake s Method of Linear Invariants oooonnconoccconcnnonnn nncnnno 107 Assessing Confidence using Bootstrap Analysis 112 Random Trees eioten er lee ee ese aes 115 Diagnosing LCOS a na e n Ii 117 Cladograms and Phylograms oooooonnnocccnonccinocccnnnnnnnnnos 117 Consistency Indices and Goodness of Fit Statistics 118 Table of Branch Lengths and Linkages o oooooncccnnnnnoncc 118 Change and Apomorphy ListS ooooonnnnccnnnnccinoccninnncnnnnos 119 Character Diagnostics ci isis 121 Character State RECOMStructiOns oooooooccnocononncannninnncnnncnnno 122 Stepmatrix Character Reconstruction Special Co siderations ir rea eas 124 The Pairwise Homoplasy and Patristic Distance NIC 2 A hee Bade de ee 128 Lengths and Fit Measures iis iscsssccsevisesesnsedbesstcsesedsutscnacaasedacndes 129 len pl abi e EES pcan ess cassctae ev eaedeosuaavneoves ets 131 Rooting Trees for Output and Character Diagnosis 131 Saving Trees t Fes 133 Recovering Trees from FileS ooooconnnnccnnonccinoncconnccninnnnnos 134 Compara LTCC Scen do 136 Calculating Consensus Trees ooooooocccconcccnoncconnncconanacinnnnos 137 PUES A sedans ai E E ak asker 138 Condensma Tes en 140 Rooting and Derooting Trees cc eeeeeeeseceesseeeeeneeeennees 140 Printing and Pl
8. If maximum tree length is nonzero only trees of length less than or equal to the specified length are retained MINLENGTH minimum tree length If minimum tree length is nonzero only trees of length greater than or equal to the specified length are retained NUMLE highest tree number to keep Keep only trees numbered 1 through highest tree number to keep NUMGE lowest tree number to keep Keep only trees numbered between lowest tree number to keep and the number of trees currently in memory CONSTRAINTS constraint tree name Only those trees that are compatible with the specified constraint tree are retained SD symmetric difference distance FROM tree number Only those trees within symmetric difference distance units of the specified tree number are retained LESSRESOLV Keep a tree only if more highly resolved compatible trees do not exist 198 PAUP 3 1 USER S MANUAL If NOT is specified trees that do not satisfy the specified filtering criteria are retained If PERMDEL is specified trees not retained by the filter are permanently deleted from memory Otherwise they are only temporarily deleted and can be recovered by issuing the command FILTER OFF GETTREES Use the GETTREES command to load trees into memory from a file containing a NEXUS format TREES block The syntax is GETTREES FILE file name options The available options are FROM starting tree number If nonzero star
9. Note that in some cases the assignment of the would be most parsimonious state is ambiguous However the length required is the same regardless of how the ambiguity is resolved so that for the purposes of calculating the length of a particular tree or comparing the lengths of alternative trees ambiguity does not present a problem In the example below either state 0 or 1 could be assigned to taxon B reconstructions A and B respectively In either case the required length of two steps is greater than that required by the trees above Thus although this character is not informative for placing taxon B it is still useful for discriminating among tree topologies PAUP 3 1 USER S MANUAL 25 D JE D lt 0 0 0 0 Loh E NE E 0 0 0 0 A Nai A 0 1 0 1 Cc O va E al 7B e B 0 0 1 gt 1 NE 0 1 0 0 G G A B Two equally parsimonious reconstructions on the same tree demonstrating the handling of ambiguity in the presence of missing data Although the two reconstructions are different the length of the reconstructions is two steps in each case When reconstructing characters PAUP always assigns a nonmissing state to all internal nodes Occasionally users have complained about this behavior For example some users have argued that for the reconstruction shown below state should have been assigned to the nodes at the base of the D E and C D E clades rather than state 1 This choice would imply
10. Printing is currently on if the item is checked in the menu as shown below and off otherwise Page Setup Print File 8P Echo to Printer HE Print Selection Log Output to Disk 3L Portion of File menu indicating the printer echo is currently activated This method of printing uses low level printer driver calls that are not supported by some non Apple printers If you encounter problems when attempting to print using the Echo To Printer command you will have to use a different method for obtaining printed output For example you can save output to a disk file and then print it later using PAUP or a text editor word processor of your choice See the Log Output to Disk command for how to save PAUP output to a file PAUP 3 1 USER S MANUAL 229 Print Selection This option will print any selected text either from the display buffer or a data file If you want to print the entire display buffer choose Edit Display Buffer from the Edit menu select all text and then choose Print Selection Alternatively you can save the contents of the buffer to a file which can be printed later see Log Output to Disk below Log Output to Disk The Log Output to Disk command allows you to control the saving of PAUP output to disk files The following dialog box will appear Sample data files v Hard Disk O S fasoa test Eject O fnacuelos i Fngiasporms Desktop O fingiasperms treas D Berberidaceae New O Cartaon
11. If you run PAUP from a floppy disk always do this from a copy of the original disk You may also want to copy the Sample Data Files folder Be sure the read the file READ ME FIRST It contains information pertaining to errors in the manual or changes made to the program after the manual was printed This is a standard text file and can be read by any text editor word processor or by PAUP Note to PAUP Mac 3 0 users You can delete the PAUP Defaults file if any in your System Folder PAUP Mac 3 1 stores its settings in a file named PAUP Preferences located in the Preferences Folder under System 7 0 and later or the System Folder under systems earlier than 7 0 222 PAUP 3 1 USER S MANUAL THE PAUP EDITOR The PAUP editor has most of the features expected of a Macintosh text editor including horizontal and vertical scroll bars cut copy paste clear and undo facilities selection of a word by double clicking etc You may also select an entire line by triple clicking You may set the default font size and tab width by choosing Editor from the Options menu Although not required use of a monospaced font like Monaco or Courier is recommended so that columns of the data matrix will line up cleanly The default font PAUPMonaco is a modification of Monaco that overcomes some of the shortcomings of Monaco Specifically it distinguishes between zero and the upper case oh between the lower case el and the upper case ey
12. MAU Phylogencic Analysis amp sing Parsimnraeny Version 3 1 March 1993 Program by David L Swofford User s Manual by David L Swofford and Douglas P Begle Laboratory of Molecular Systematics MRC 534 MSC Smithsonian Institution Washington DC 20560 USA Development of versions prior to 3 1 supported by Center for Biodiversity Illinois Natural History Survey 607 E Peabody Drive Champaign Illinois 61820 USA License Agreement and Disclaimer PAUP is licensed to individual users for the sole purpose of facilitating the scientific research of the licensee This software may be used in more than one location or by more than one person provided that there is no possibility that it will be used by two or more people simultaneously Generally this qualification means either of the following 1 the program may be used by a single researcher on any number of machines in his or her possession or 2 the program may be installed on a single machine and used by one or more individuals who have access to that machine If you cannot abide by the terms of this agreement please return the software to the Illinois Natural History Survey immediately for a refund of the distribution fee THIS SOFTWARE IS PROVIDED AS IS WITHOUT WARRANTY OF ANY KIND DAVIDL SWOFFORD THE SMITHSONIAN INSTITUTION THE ILLINOIS NATURAL HISTORY SURVEY AND THE UNIVERSITY OF ILLINOIS DO NOT WARRANT GUARANTEE OR MAKE ANY REPRESENTATIONS REGARDING T
13. TAXSET Use the TAXSET command to define a taxon set Taxon sets are simply groups of taxa that can be referred to by a single name in other commands The syntax is TAXSET taxon set name taxon list The taxon set name must not be identical to any of the original taxon names m See Simplifying Input with Sets taxon sets in Chapter 2 for examples on the use of TREEDIST Use the TREEDIST command to request output of a matrix of tree to tree distances computed according to the symmetric difference or partition metric Penny and Hendy 1985 The syntax is TREEDIST options The available options are COMPARE tree number If tree number equals 0 a matrix of all pairwise comparisons is calculated If tree number gt 0 the specified tree is compared to all others PAUP 3 1 USER S MANUAL 219 NO ED By default the frequency distribution of tree to tree distances is output Use NOFD to suppress this output or FD to reverse the effect of a previous NOFD specification NO SHOWALL By default a matrix of the tree to tree distances for all comparisons performed is output Use NOSHOWALL to suppress the output or SHOWALL to reverse the effect of a previous NOSHOWALL specification TREEINFO Use the TREEINFO command to obtain information on the status of trees currently in memory No options are available TSTATUS Use the TSTATUS command to obtain information on which taxa if any are deleted and w
14. The Editor dialog box You may change the type font size and number of spaces that are equivalent to one tab If you change the font it is strongly recommended that you use a fixed width font like Monaco or Courier so that columns will line up evenly In addition you may enable or disable the autoindent facility If autoindenting is on when you hit return at the end of a line the cursor is automatically indented to the same column position as the starting point of the previous line You can make your changes apply to all open windows by clicking the box Apply to all currently open windows Warnings Errors Use Warnings amp Errors to specify what action PAUP will take when it encounters unrecognized commands and keywords in a file to specify 244 PAUP 3 1 USER S MANUAL sound options and to enable or suppress the warnings PAUP issues before resetting the active data file Warning and Error Preferences Action when unrecognized commands OPTIONS command keywords or formats are encountered in NEXUS blocks O Issue warning to display window and or output file and continue Stop processing and enter editor O Warn before resetting the active datafile Warn before performing treefile operations that will potentially delete unsaved trees O Warn before quitting with unsaved trees Sound J Beep on error and warning messages J Beep on prompts for input from user J Beep on keyboard errors tance CK
15. The following options apply only if MAJRULE is in effect CUTOFF required majority value Specifies the percentage of the trees on which a group must appear in order to be retained in the majority rule consensus A group must appear on more than this percentage of the trees before it is retained The default is 50 NO JLESO If LESO is specified groups occurring on less than 50 of the trees are retained in the consensus if they are compatible with the groups already on the tree NOLESO reverses the effect of a previous LESO specification NO GRPFREQ By default a table s output that shows all partitions or groups occurring on at least one tree and the frequency of each such group NOGRPFREQ can be used to suppress this output NOJADAMS Specify ADAMS to request computation of an Adams consensus tree NOADAMS reverses the effect of a previous ADAMS specification 190 PAUP 3 1 USER S MANUAL NOJINDICES Specify INDICES to request calculation of a variety of consensus indices The indices computed by PAUP are described in Rohlf 1982 and Swofford 1991 NO SAVE If SAVE is specified a description of all consensus trees computed is output to a file containing a NEXUS format TREES block This option is not persistent you must specify it on every CONTREE command for which you want a tree file to be saved The following options apply only if SAVE is specified FILE file specification Specifies the name for t
16. all of the taxa that possess a particular derived character state could unambiguously be interpreted as descendants of the ancestor in which the state originated and taxa sharing the ancestral condition could be definitively excluded from that group Inevitably however character conflicts or incompatibilities arise For example consider the data shown below Table 1 Hypothetical data for example discussed in text Character Taxon 1 2 3 4 One 1 1 1 1 Two 1 0 1 1 Three 0 1 0 1 Four 0 0 0 0 For now we will assume that 0 represents the ancestral state for each character Thus we observe that taxa One and Two share the derived state for characters 1 and 3 while taxa One and Three share the derived state for character 2 Consequently if One and Two were in reality sister taxa tree A below the possession of state 1 for character 2 in One and Three would be a homoplasy whereas if One and Three were sister taxa the PAUP 3 1 USER S MANUAL 3 derived characters 1 and 3 shared by One and Two would be homoplasies tree B below Tree B Two trees for the data of Table 1 Faced with such a conflict a practitioner of traditional manual methods would usually decide between these two alternative hypotheses of relationship by invoking the parsimony criterion and therefore choose tree A which requires fewer assumptions of homoplasy But this choice is exactly the one that would have been made by the computer method as well since
17. 266 PAUP 3 1 USER S MANUAL Available information includes a list of apomorphies character changes by branch list of changes character change by character table of linkages assigned minimum and maximum branch lengths character diagnostics goodness of fit statistics for each character states for terminal taxa states for internal nodes possible state assignments the f value of Farris for each tree chosen a patristic distance matrix between taxa and finally a pairwise homoplasy index No matter what character information is selected a tree will always be output along with basic tree statistics This is the one of the best places to explore where characters changed on a tree and which nodes are supported by which characters Keep in mind that the output for many of these options will directly be a function of the method of optimization chosen Show Reconstructions This command produces either a a plot of the chosen tree s with the current reconstruction for each selected character mapped onto the tree Character changes or b a plot of the chosen tree s with the states that could be assigned to each internal node in at least one most parsimonious reconstruction Possible states assignments character changes f O possible state assignments ERE J Compress trees vertically Optimization options Cancel 3 The Show Reconstructions dialog box Branches are shaded at lo
18. G S and L R Moore 1982 A statistical evaluation of multiplicative congruential random number generators with modulus 23 1 Journal of the American Statistical Association 77 129 136 Fitch W M 1971 Toward defining the course of evolution minimal change for a specific tree topology Systematic Zoology 20 406 416 Funk V A and D R Brooks 1990 Phylogenetic Systematics as the Basis of Comparative Biology Smithsonian Institution Press Washington D C PAUP 3 1 USER S MANUAL 277 Gould R 1988 Graph Theory Benjamin Cummings Menlo Park California Harary F 1969 Graph Theory Addison Wesley Reading Massachusetts Hendy M D and D Penny 1982 Branch and bound algorithms to determine minimal evolutionary trees Mathematical Biosciences 59 277 290 Hendy M D M A Steel D Penny and I M Henderson 1988 Families of trees and consensus Pages 355 362 in H H Bock ed Classification and Related Methods of Data Analysis Elsevier Amsterdam Hennig W 1966 Phylogenetic Systematics University of Illinois Press Urbana Illinois Hillis D M 1987 Molecular versus morphological approaches to systematics Annual Review of Ecology and Systematics 18 23 42 Hillis D M 1991 Discriminating between phylogenetic signal and random noise in DNA sequences Pages 278 294 in M M Miyamoto and J Cracraft ed Phylogenetic Analysis of DNA Sequences Oxford University Press New Yor
19. NO COLLAPSE NOCOLLAPSE requests that zero length branches not be collapsed to yield polytomies The default setting is to collapse zero length branches COLLAPSE can be used to reverse a previous NOCOLLAPSE setting NOJENFORCE ENFORCE requests that topological constraints be enforced i e trees that are not compatible with the constraint tree are not evaluated If the CONSTRAINTS option see below is not used to specify a constraint tree the current constraint tree is used CONSTRAINTS contraint tree name The specified constraint tree which must have been defined in a previous CONSTRAINTS command becomes the current constraint tree You must also specify ENFORCE if you want to search under constraints NO CONVERSE If CONVERSE is specified in conjunction with ENFORCE only trees that are not compatible with the constraint tree are evaluated NOCONVERSE reverses the effect of a previous CONVERSE specification NOJINCLUDEANC If INCLUDEANC is specified the current ancestor is included in the analysis and is used to root the tree You can use the ANCSTATES ancestral states name to choose a different ancestor The ASSUME command can be used to choose the current ancestor NOINCLUDEANC reverses the effect of a previous INCLUDEANC specification NO STATUS Ordinarily information on the progress of the search number of trees examined number of trees saved etc is output to the screen or to a window NOSTATUS su
20. P In press Missing data versus missing characters in phylogenetic analysis Systematic Biology Maddison W P M J Donoghue and D R Maddison 1984 Outgroup analysis and parsimony Systematic Zoology 33 83 103 Maddison W P and D R Maddison 1992 MacClade Analysis of Phylogeny and Character Evolution version 3 0 Sinauer Associates Sunderland Massachusetts Margush T and F R McMorris 1981 Consensus n trees Bulletin of Mathematical Biology 43 239 244 Meacham C A 1984 The role of hypothesized direction of characters in the estimation of evolutionary history Taxon 33 26 38 Meacham C A 1986 More about directed characters a reply to Donoghue and Maddison Taxon 35 538 540 Miyamoto M M 1985 Consensus cladograms and general classifications Cladistics 1 186 189 Page R D M 1989 Comments on component compatibility in historical biogeography Cladistics 5 167 182 Page R D M 1993 COMPONENT Tree comparsion software for Microsoft Windows version 2 0 Natural History Museum London Penny D and M D Hendy 1985 The use of tree comparison metrics Systematic Zoology 34 75 82 Robinson D F and L R Foulds 1981 Comparison of phylogenetic trees Mathematical Biosciences 53 131 147 PAUP 3 1 USER S MANUAL 279 Rohlf F J 1982 Consensus indices for comparing classifications Mathematical Biosciences 59 131 144 Sanderson M J 1989 Confidence limits on phylogenies the boo
21. PAUP may be controlled entirely by a command line system Commands may also be placed into input files for convenience i e to avoid repeated typing of complicated commands and to provide a simple batch facility COMMAND FORMAT The command descriptions below use the notational conventions specified in the Preface Remember that upper case items are to be entered as shown Italicized items e g user item represent variable items to be substituted by the user Items inside of square brackets e g OPTIONAL_ITEM are optional Items inside of curly braces and separated by vertical bars e g X Y Z are mutually exclusive options only one of the choices indicated may be used The default item if any is underlined as in ABC DEF Options in commands are generally specified as KEYWORD VALUE for options that have two or more potential values For options that represent simple on off Boolean switches options may be selected by simply specifying KEYWORD or deselected using NOKEYWORD Alternatively the forms KEYWORD YES and KEYWORD NO can be used to select or deselect options respectively 166 PAUP 3 1 USER S MANUAL Each command begins with a command name and ends with a semicolon Otherwise the commands are completely free format A command may span any number of lines and whitespace tabs and blanks may be inserted at will Input of PAUP commands is case insensitive so you may enter command names op
22. REPLACE NO FLUSH HE A If START is specified the default logging is initiated to the named file If FILE is not specified explicitly a default name is assigned If APPEND is specified subsequent output is appended to the previous contents if PAUP 3 1 USER S MANUAL 207 any of the file Otherwise subsequent output will overwrite the original contents of the file The APPEND REPLACE setting is retained between invocations of the LOG command unless the file is changed by a FILE directive If REPLACE is not specified explicitly and the file already exists you will receive a warning and will have the opportunity to cancel the command before the contents of the existing file are erased Specification of FLUSH causes the file s buffer to be flushed after every line of output Ordinarily this degrades system performance and is not recommended However there may be situations in which immediate flushing is useful If the file specification contains any of the characters equal sign semicolon colon or blank it must be enclosed within single quotes Examples LOG FILE MyOutput subsequent output is saved to file MyOutput LOG STOP subsequent output no longer saved to file LOG START APPEND resume logging appending new output to file MyOutput LOG STOP suspend logging once again LOG START resume logging again appending to MyOutput LOG FILE NewOutput begi
23. SANEASSU Mine dia 198 SAY ETRE inicio 198 SE Daria 199 SHOW ANG as 203 SHOW CONS UR ori io 203 SHOW DIST a 203 SHOWMATRIX ostia pias 204 SHOWTREBS a dia 204 SHO WUSER 0 204 TAXSE ts asi oeste 204 TURE DIS Lira 205 TREINTA 205 TS TATU ri aras 205 A EE EE oceanic 205 USERTRE Rca aii 205 USERT PES nadaa 206 WEI Sitio 206 E 206 WISE aia 206 CHAPTER 4 THE MACINTOSH INTERFACE cccococcoconoccnonnnonnncconocancnonnncnnnccanannnos 207 Installation onda 207 Th RAUL Edo to de dto 207 The Command Line muii ir eiiiai a ais 208 Selectie Items TAS tc aa Bae eee 209 Running under MultiFinder or System 7 0 and Later 209 The Apple Meta da 210 The Els Menta ind 211 Neill 211 DI e al e neta co 211 A Goesaumaaweraneesncseuensnaveueiabenenatens 212 ss io Gee 212 AVE A O OS 212 A say T A E EE 212 Page Setup inline ec 213 O E NS 213 Echo PAM ren 214 Print Selection a caice ea e E E A EE 214 Log Output to Disk oy evcsssscesaessosescenssaceasdionendcassavcavevedeeaters 214 Execute Filej ii a ei ed is 215 Export Fil en an e A a eaa 215 POEL m IE E E E TAE ee 216 Mii ini 218 The Edit Ment kemen na a E gated ey a aanceinenaes 218 Undo beha ls oe e a a gen a a O oes deme Eaa 218 tee a oe 218 A O dae AASE STS 218 PAS Le ld ae ence ee od Sashes 218 AIGA erenn lohtans dea atar eat 219 A Seca vance a O E 219 Clear Display Buffer iii ess ddeatesestacesdnsccaanneesaanels 219 Edit Display Butlers nscicdgretetnd
24. The best way to learn the difference between the various formatting options is just to experiment with different settings prior to issuing a SHOWMATRIX or Show Data Matrix menu command Searching This command controls background processing status window display and event checking while PAUP is performing a search Macintosh Specific Search Options E Show status window during searches 1 Check for mouse down and keyboard events during searches 1 Close status window automatically at end of search E Allow background processing XX Visual background notification i Alert box plus flashing icon O Flashing icon only XX Audible background notification Cancer x The Searching dialog box Ordinarily PAUP continues processing e g tree searches when it is switched to the background allowing other activities e g word processing in the foreground For large data sets however the responsiveness of the foreground application may be impacted If you disable background processing PAUP will simply sleep when it is switched to the background and will resume processing when it is returned to the foreground For example a run could conceivably continue over a period of several days by letting PAUP sleep during the day when you were using your machine for more important tasks but moving PAUP to the foreground when you left for the day If event checking is disabled PAUP will ignore all mouse cl
25. Up to 225 trees can be printed on a page with or without tree numbers You can also change how the trees are rooted when they are printed using outgroup Lundberg or midpoint rooting If outgroup rooting is chosen the outgroup can be rooted at an internal node with a basal polytomy or the ingroup can be made monophyletic relative to the outgroup If there is more than one outgroup present the tree can be rooted so that the outgroup is either paraphyletic relative to the ingroup or a monophyletic sister group to the ingroup Once you have selected the display options you prefer the Preview button allows you to see what the output will look like without printing It is useful to toggle back and forth between the Preview menu and the main selection menu to fine tune your selections You also have the option in Preview to copy the image to a PICT file which can be opened by any PAUP 3 1 USER S MANUAL 269 Macintosh graphics program or to the clipboard from which it can be pasted into an open window of a graphics application If you plan to import the image to an application which cannot import PICT files use Copy to Clipboard Neither PICT file nor Copy to Clipboard requires that you actually print the image PAUP currently makes no attempt to intelligently print files that are wider than the width of one page If this is a problem with your file some suggestions to reduce the size of the output are use a smaller font format using th
26. above The syntax is EXECUTE file specification If file specification contains any of the characters equal sign semicolon colon or blank it must be enclosed within single quotes The input file may contain any or all of the following DATA blocks ASSUMPTIONS blocks TREES blocks and valid PAUP commands Although not required you should place PAUP commands inside of a PAUP block so that other programs e g MacClade can use the same file Blocks other than DATA ASSUMPTIONS TREES and PAUP are permitted but are ignored by PAUP PAUP commands are processed exactly as if they had been entered from the command line with only a few exceptions e g the EDIT command cannot be issued from a file Commands are processed until the end of the file is reached or a QUIT command is encountered EXSET The EXSET command used to define exclusion sets is ordinarily issued from within the ASSUMPTIONS block However you may also issue it from the command line or from within a PAUP block See Commands used in the ASSUMPTIONS Block for the description of this command PAUP 3 1 USER S MANUAL 197 FILTER Use the FILTER command to filter trees according to length constraints or other criteria see Filtering Trees in Chapter 2 The syntax is FILTER NOT filtering criteria PERMDEL or FILTER OFF Any combination of the following filtering criteria can be specified MAXLENGTH maximum tree length
27. currently available in the Macintosh version only Taxon names have priority over taxon numbers in input commands This only becomes an issue if you use all digit taxon names For example if five taxa in a data file are for whatever reason given the names One 34 three 3 673 then the command delete 3 would cause deletion of the fourth taxon Because of the potential confusion you are strongly urged not to use all digit taxon names The name ALL is reserved for a taxon list containing all of the taxa in the data matrix see below You may not name a taxon ALL Character identifiers Ordinarily PAUP refers to characters by consecutive integers starting with 1 These integers are used both for input commands and in the output You may also assign alphanumeric character names if you wish The maximum length of a character name is 32 but PAUP truncates character 168 PAUP 3 1 USER S MANUAL names to 10 characters in its output Thus you should choose names that are unique up to at least the first 10 characters If you assign alphanumeric character names these will also be used to identify characters in PAUP output The rules for including punctuation underscores and blank characters are the same for character names as for taxon names see above As for taxa character names have priority over character numbers For example if you used all digit character names e g sequence positions or restriction map locati
28. either of these commands For example CHGPLOT 1 3 5 7 POSSPLOT will cause both commands to output information for characters 1 3 5 and 7 The trees below are the reconstruction of a single character using the command describe 1 chgplot PAUP 3 1 USER S MANUAL 131 The first tree optimized the character using ACCTRAN the second used DELTRAN You can see how different the two hypotheses are for character 1 The ACCTRAN hypothesis proposes an origin at the node leading to taxa 4 5 6 and 7 with a subsequent loss in taxon 5 The DELTRAN hypothesis prefers two independent origins of character 1 once in taxon 4 and once in taxon 7 Changes in character 1 on tree 1 AK EEE EEE EEE EE EEE EE EEE EEE EEE EEE EEE ELE EE ELLE EEL EEE 0 taxonl gt FEEL EE EEE EEE EEE EEE EEE EEE EEE ELE EEE LE EE EEK 0 taxon fic cc ccccccnd AEE EEE EEE EEE EEE EEE EEE EE 0 taxon3 ficcccccccccd AEE EEE EEE EEE EEK 1 taxond nad eds 1 a dj oO ea OND ficcccccccccl flaccccccccce 1 taxon6b Fororererral arce 1 taxon Changes in character 1 on tree 1 A EEE EEE EE EEE EE EEE EEE EEE EEE EEE ELE EEE ELL EEE LEE EE 0 taxonl gt A EEE EEE EEE EEE EE EEE EEE EEE EEE EEE LEE O taxon fic cc ccccccnd A EEE EEE EEE EEE EEE ELEC EEE 0 taxon3 ACRE REA Oe AAA AAA AAA 1 taxon4 ocre Marron n arras 0 taxon5 ficcccccccccd flaccccccccce 1 taxon Leccccccccce 1 taxon You can also display one tree summarizing the possible states for a given character This is d
29. find trees that are very different from a particular tree Use your imagination Remove Filter Remove Filter undoes the effect of a previous Filter Trees command making all trees in memory available once again Reverse Filter The Reverse Filter command simply reverses the filter status Trees formerly hidden by the filter become available and vice versa Show Trees Use this command to quickly output trees without associated character information PAUP 3 1 USER S MANUAL 265 Show Trees EJ Compress vertically Rooting TS Cancel ok The Show Trees dialog box If the trees are unrooted the rooting method currently in effect will be used or it can be set by clicking on the Rooting button Describe Trees Use this command to output trees with associated statistics and or character information There are a wide array of output options See the section Diagnosing Trees for detail on these options and their output Tree Description Options Select tree s to describe Optimization amp table of linkages states for interior nodes i list of changes O possible state assignments J character diagnostics O f value i O list of apomorphies O patristic distance matrix O states for terminal taxa J homoplasy matrix Y cladogram XX label interior nodes i O phylogram E Compress vertically Defaults Cancel Describe The Describe Trees dialog box
30. gt node_11 1 y gt 3 oth gt 12 gt 2 node_11 gt node_10 4 o gt 7 333 gt node_10 gt taxon4 6 429 gt 8 fa gt 12 2 gt 4 node_10 gt node 9 5 gt node 9 gt node_8 11 gt 12 A 2 gt 3 node 8 gt taxon6 3 nD gt 7 333 gt 8 NE gt node_8 gt taxon7 2 A gt 6 3 gt 10 0 gt Character Diagnostics In addition to information about where the characters change on a particular tree PAUP will display summary diagnostic information for each character This is obtained by selecting Character diagnostics in the Describe Trees dialog box or the DESCRIBE command with the DIAG option The output includes the minimum possible length for each character 1 e the length 1f a minimal tree were computed for each character taken separately the maximum possible length i e the length on a completely unresolved bush the length required on the tree being described and four goodness of fit statistics the unit consistency index CI homoplasy index HI retention index RI and the rescaled consistency index RC This output is obtained by the following command describe 1 diag and has the following format 130 PAUP 3 1 USER S MANUAL Character diagnostics Mi ni mum Tree Maxi mum Character Steps Steps Steps C H 6 DADO DN a 9 10 11 2 12 n a 13 n a 14 2 15 2 16 2
31. is collapsed 3 Collapse an interior branch if the maximum possible length of the branch is zero That is if all MPRs assign zero length to the branch for every character then the branch is collapsed I believe that only the third criterion can be justified and this rule is the one implemented in PAUP Comparisons of results from Farris s Hennig 86 program suggest that this rule is used by that program as well although his documentation does not address the issue The rationale is simply that if potential though perhaps ambiguous support exists for a particular resolution of a polytomy then the resolution should be retained However if no support exists for any resolution i e a branch length is zero under every possible reconstruction then a polytomy is in order For example if three trees found by PAUP consist of a trichotomy and two of the three resolutions of that trichotomy This means that there is potential support for two of the three resolutions of the trichotomy but for the third resolution there are no characters that provide even potential support Note that even when a branch is not collapsed it may still be assigned zero length under some reconstructions Thus if you have requested the option to collapse branches and yet zero length branches occur in the output you will know that there are other MPRs in which the branch has nonzero length The first criterion is flawed in that it may not be possible to collapse
32. parallel gains of the derived condition are not allowed As for ordered Wagner characters character states are linearly ordered according to their position in the symbols list specified in the FORMAT command of the DATA block However we now define a forward transformation as a change from a less derived state to a more derived state and a backward transformation as a change from a more derived state to a less derived state Rooted vs unrooted Dollo models PAUP as well as MacClade differs from some other implementations of Dollo parsimony e g Felsenstein s PHYLIP in that it can operate as an unrooted method The only requirement is that the reconstructed character states be consistent with the constraint that each derived state be uniquely derived Under this definition the position of the root affects neither the assignment of character states nor the length of the tree For example both of the trees below which differ only in the placement of the root require two steps under the unrooted Dollo model assuming that state 1 is the derived state 10 PAUP 3 1 USER S MANUAL T 1 0 YN Reconstructions for a Dollo character under two different rootings of an unrooted tree Terminal taxa are labeled according to their state for the character in question That is neither tree requires more than a single origination of state 1 In the tree on the right the derived state 1 is assumed to be ancestral with respec
33. taxon7 7 8 9 10 11 NEO 0 00 N NN ONO WNKNNHE WR Oo Ren Down An Change and Apomorphy Lists There are two ways of summarizing the character changes that have taken place on a particular tree The first is to describe the changes for each character the second for each branch If List of changes is selected in the Describe Trees dialog box or the command DESCRIBE CHGLIST is issued the output will show for each character where on the tree that character changed the state it changed from and the state it changed to If list of apomorphies is selected in the Describe Trees dialog box or DESCRIBE APOLIST is selected the output will show for each branch which characters changed the state they changed from and the state they changed to The following command will give character changes describe 1 chglist with the following output 128 PAUP 3 1 USER S MANUAL Character change lists Character ln Steps Changes 1 000 node_ 2 500 node_ node 3 500 node_ node 4 000 node_ 5 000 node_ 6 500 node_ node 7 333 node_ node_ node 8 500 node_ node 9 667 node_ node_ node_ 10 000 node_ node_ node _1l axon axon node _1l axon6 node_ 10 node 9 axon4 axon axon node_10 axon6 axon4 axon6 2 taxon2 axon4 2 taxon7 2 node 10 node 8 o 4 4 4 44 A uouo wou wo wou aw Wr OWN C a
34. use NORSTATUS to suppress this output The status report is very useful in evaluating the effectiveness of the heuristic search see the Stepwise Addition section for a description of random addition sequences HOLD n Specifies the number of trees to be help at each cycle of the stepwise addition procedure see the section Stepwise Addition By default HOLD 1 so that a single tree is held at each step However setting HOLD gt 1 sometimes improves the length of the tree found by stepwise addition SWAP TBR SPR NNIINONE Specifies the algorithm used by branch swapping TBR tree bisection reconnection SPR subtree pruning regrafting NNI nearest neighbor interchange NONE no branch swapping performed NNI rearrangements are a subset of those done by SPR and SPR rearrangements of those done by TBR Ordinarily you will use TBR but SPR or NNI can be used to reduce search times NO MULPARS Ordinarily PAUP saves all minimal trees it finds during branch swapping You can use NOMULPARS to save only one of the shortest trees found Use of NOMULPARS is not recommended 202 PAUP 3 1 USER S MANUAL as it can drastically reduce the ability of branch swapping to find the shortest tree see the section on Branch Swapping FROMTREE tree number TOTREE tree number Specify FROMTREE and or TOTREE to request use of a subset of the available trees for input to the branch swapping procedure Only trees with numbe
35. 31 23 P 122222 2 2 22 2 22223 Q01 222122121 22 1222222 3 R 212221 1 12 2 212 3 1 112212222 2 2 2 2 2 2 dsl 3 2 2 1222 212 222122 2 2 2 2 2 3 11 2 202 22 202 Vel 2 Pe 12 BolT 02 2 2 23 V 12111 121211222222202223 Wh 212221222 23221 222 2 0 2 1 3 Y 222121120230 L 2 22 2222013 y 212121221 2221112221103 33333333 ee 3 33303393 3 33 3 30 Stepmatrix for amino acid substitutions The stepmatrix gives the minimum number of amino acid replacement substitutions needed to convert one amino acid to another based on the genetic code used in nuclear genes of most organisms and chloroplast genes in plants The character distribution in the five taxa is matrix Alpha ABCDEFGHIK Beta AB EFGHIK Gamma BCDSFG Delta CIKDEFGHIK Epsilon DIKDEFGHIK where is a missing observation and is a deletion Reconstructing the third position on one of the unrooted trees gives the following when only the observed states are allowed at internal nodes Both C cys and K lys are possible assignments at the internal node Possible state assignments for character 3 on tree 1 BOE EEE OEE EEE EEE EEE EEE EEE EEE EEE EEE C Alpha LATA AAA RA AA ARAS o Beta CorenererenananarererCK flecccccccccccccccccccce K Delta gt gt fic Cee EEE EEK EEE eee EEC EK Epsilon FEE EEE EEE EEE EEE EEE EEE EEL EEE ELE EEE EEE EEL EEE KE KEKE EC Gamma Reconstructing internal nodes gives a different result when states that satisfy the 3 1
36. A node having degree greater than three e g one immediate ancestor and three immediate descendants see below corresponds to a polytomy or a multifurcation and trees containing one or more polytomies are sometimes called polytomous nonbinary trees A tree is rooted if there is a special node the root that imparts a direction upon the tree such that nodes lying on a path connecting the root to any other node are ancestors of ancestral to nodes in the path that are further from the root and descendants of descendant to nodes closer to the root Typically the root is an internal node having degree two however in PAUP the root is usually considered to be an additional terminal node positioned at the base of the tree A subtree is a connected subset of a tree A subtree consisting of all of the nodes and branches that descend from a particular internal node v on a rooted tree is called the subtree rooted at v or simply v s subtree Phylogeneticists often refer to the subtree rooted at an internal node as a clade A tree is ordered if the branches incident to each node are assigned in some nonarbitrary fixed order If the order in which the branches are connected is irrelevant or arbitrary as is generally the case for phylogenetic trees then the tree is said to be unordered Although the trees computed by PAUP are unordered in the sense that free rotation of subtrees around internal nodes does not affect the relationships implied fo
37. CharSets w 5 TypeSets v i ser Befined vi The Set Character Types dialog box Available character types are ordered Wagner unordered Fitch Dollo irreversible Camin Sokal and user defined See Character Types for a description of these character types 248 PAUP 3 1 USER S MANUAL When you choose Set Character Types a dialog box containing a scrollable list of character numbers and names if you have provided them will appear Select the character s whose type s you want to change and click on one of the character type buttons or menus at the bottom of the dialog box See the section Selecting Items in Lists for tips on making multiple selections The list will be updated to reflect the new character status You may then proceed to select other characters and set their types in the same manner If the Dollo or Irrev buttons are pressed a menu will pop down allowing you to choose an option for specifying character polarity See the section Character Types for the meaning of these options Similarly if the User Defined button is pressed a menu will pop down that shows the user defined types that were specified in the ASSUMPTIONS block Choose the desired type from this menu Set Character Weights Use this command to change the weight assigned to each character Character Weights Choose characters to be weighted CharSets w CER CLE Assign weight 2 Wt
38. Characters aoe Log subsequent output to 5 Taxon test out Cancel A output to screen Start Saving a Copy current contents of display buffer to file Step saving The Log Output to Disk dialog box To begin saving output name the output file then press Start saving All subsequent output generated by the program will be saved in this file If you want to stop saving the output to disk choose Log Output to Disk again and press the Stop saving button The Output to screen checkbox allows you to specify whether output is also written to PAUP s main display window If you want to send output to the file only without having it scroll by on the screen uncheck this box The Copy current contents of display buffer checkbox allows you to save the buffer to a file If you attempt to output to an already existing file you will be asked whether you want to append to the file or to replace overwrite it Execute File Use this command to request processing of the indicated file If you are in editor mode the file being edited will be executed Otherwise the most recently closed file will be executed For example you can open a file for editing make any desired changes save and close it and then execute it 230 PAUP 3 1 USER S MANUAL Export File Choose Export File to export a data matrix and or tree descriptions in one of three available formats 1 PHYLIP J Felsenstein s Phylogeny Infere
39. DATATYPE STANDARD DNA RNA PROTEIN GAP gap symbol OPTIONS IGNORE NONE INVAR UNINFORM MSTAXA UNCERTAIN POLYMORPH ZAP character list GAPMODE MISSING NEWSTATE CHARLABELS character name TAXLABELS taxon name STATELABELS charnum and state list charnum and state list MATRIX data matrix ENDBLOCK Every data block begins with a DIMENSIONS command specifying the size of the data matrix number of taxa number of characters It is followed by one or more optional commands that specify details of the matrix format option settings and character and or taxon names The last command in the DATA block MATRIX defines the data matrix itself A very simple example of a DATA block follows Note that the style of indenting shown here is not required but merely serves to enhance readability begin data dimensions ntax 4 nchar 5 matrix taxonl 00111 taxon2 0111 taxon3 11001 taxon4 10000 r endblock Remember that the data matrix like the rest of the NEXUS input file is entirely free format You may use any number of physical lines to represent each row of the data matrix We could just have well have entered the matrix above as follows PAUP 3 1 USER S MANUAL matrix taxo or even matrix taxonl 00111 taxon2 0111 taxon3 11001 n1 n2 11 n3 n4 taxon4 10000 NOTE In PAUP 2 4 one of the taxa in the data ma
40. David Cannatella Jonathan Coddington Joel Cracraft Mike Cummings Chris Darling Scott Davis Ron DeBry Doug Eernisse Bill Fink George Gutman David Hillis Kent Holsinger Bob Jansen Rick Mayden Chris Meacham David Mindell Mike Miyamoto Gavin Naylor Mark Norell David Penny Norman Platnick Winston Ponder Andrew Simons Greg Spicer Beth Stewart Sherman Suiter David Stock Rytas Vilgalys and Jim Woolley Three members of the staff of the Illinois Natural History Survey where all previous versions of PAUP have been developed deserve special recognition Angie Young has cheerfully and conscientiously taken care of the day to day distribution matters and Mary Lou Williamson has flawlessly kept the financial matters in order Larry Page director of the Center for Biodiversity provided encouragement and support throughout the development of earlier versions Finally I thank my wife Ruth Swofford for enduring many lonely evenings while I wrote and debugged code and for willingly and unselfishly providing assistance in many phases of the development of PAUP PAUP 3 1 User s Manual ABOUT THIS MANUAL Organization The manual is divided into four chapters the first three of which are common to all implementations of PAUP The first chapter presents background material relevant to the capabilities of the program We urge all users to become familiar with this material before using the program Chapter 2 provides basic instru
41. Heuristic Methods When a data set is too large to permit the use of exact methods we must resort to heuristic approaches that sacrifice the guarantee of optimality in favor of reduced computing time I like to apply the following analogy to the problem of searching for an optimal tree by approximate methods Consider the plight of a myopic pilot who loses his glasses when forced to parachute from his airplane into a mountainous region He suspects that there is a manned outpost at the top of the highest peak in the area but he must somehow grope his way there to have any hope of rescue Obviously simply walking uphill from the point of landing will not necessarily lead him to his goal since he may not have started on a slope of the highest peak Suppose that he reaches a summit and finds no outpost Two possibilities remain 1 he is in fact at the top of the highest peak but was wrong about the existence of the outpost or 2 he has climbed the wrong hill Unfortunately he will have no way of choosing between these alternatives Although rather silly the analogy is actually quite apropos Heuristic tree searches generally operate by hill climbing methods An initial tree is used to start the process we then seek to improve the tree by rearranging it in a way that reduces its length When we can find no way to further improve the tree we stop Like the downed pilot however we generally have no way of knowing whether we ended up at the t
42. J S 1970 Methods for computing Wagner trees Systematic Zoology 19 83 92 Farris J S 1972 Estimating phylogenetic trees from distance matrices American Naturalist 106 645 668 276 PAUP 3 1 USER S MANUAL Farris J S 1977 Phylogenetic analysis under Dollo s Law Systematic Zoology 26 77 88 Farris J S 1982 Outgroups and parsimony Systematic Zoology 31 328 334 Farris J S 1988 Hennig86 version 1 5 Distributed by the author Port Jefferson Station N Y Farris J S 1989a The retention index and homoplasy excess Systematic Zoology 38 406 407 Farris J S 1989b The retention index and the rescaled consistency index Cladistics 5 417 419 Felsenstein J 1978a Cases in which parsimony and compatibility methods will be positively misleading Systematic Zoology 27 401 410 Felsenstein J 1978b The number of evolutionary trees Systematic Zoology 27 27 33 Felsenstein J 1984 The statistical approach to inferring phylogeny and what it tells us about parsimony and compatibility Pages 169 191 in T Duncan and T F Stuessy ed Cladistics Perspectives on the Reconstruction of Evolutionary History Columbia University Press New York Felsenstein J 1985 Confidence limits on phylogenies An approach using the bootstrap Evolution 39 783 791 Felsenstein J 1991 PHYLIP Phylogeny Inference Package version 3 4 Distributed by the author Univ of Washington Seattle Washington Fishman
43. PAUP 3 1 USER S MANUAL SPECIFYING CHARACTER TYPES The Standard Character Types The standard character types are referred to in NEXUS and PAUP commands as ORD ordered UNORD unordered IRREV irreversible and DOLLO Dollo Dollo and irreversible characters may have an optional suffix DOLLO UP DOLLO DN IRREV UP IRREV DN See Character Types in Chapter 1 for a description of these types Assigning Character Polarities Dollo and irreversible characters require the specification of character polarities either implicitly or explicitly For Dollo characters there are three options for specifying the direction of forward less derived to more derived vs backward more derived to less derived transformations If the ancestor currently in effect has the missing state polarity can be either up or down Up specifies that states higher in the SYMBOLS order are derived relative to states lower in the SYMBOLS order i e for SYMBOLS 01 state O is ancestral and state 1 is derived Down specifies the opposite higher ordered states are ancestral to lower ordered states If the state in the currently chosen ancestor is non missing this state defines the most ancestral state with both lower and higher ordered states in the SYMBOLS list being relatively more derived Note that if the standard ancestor is in effect this state will be missing and the up vs down setting will apply Within a file or from the
44. PAUP is running in the foreground but the benefits of not having the computer totally tied up may outweigh the extra time needed for the search See the Macintosh System Software User s Guide for details on the use of MultiFinder I have tried very hard to minimize the impact of PAUP background processes on the foreground application e g keyboard delays etc that you may have seen with other programs Unfortunately however in some situations PAUP may cause an unacceptable degradation in the performance of the foreground application If this happens you may want 224 PAUP 3 1 USER S MANUAL to disable the Allow background processing under Multifinder option Searching command of the Options menu see below With the option disabled PAUP will simply sleep while it is in the background allowing the foreground application full control of the processor When PAUP is brought to the foreground its execution will resume For example you could start a time consuming search at the end of the day If 1t had not completed by the next morning you could move PAUP to the background while you used your machine for other activities during the day and then resume PAUP execution before leaving again that evening If a PAUP process finishes while some other application is currently in the foreground and you are running version 6 0 or later of the Macintosh operating system you will receive notification in two forms First an alert box w
45. Taxon test Ejea D Anacyclus D Angiosperms D Angiosperms trees D Berberidaceae D Cartoon Characters D cartoon phylip D command file O Show PAUP MacClade files only Edit Initial mode Execute 38H Edit 3E The Open dialog box Use the Close command to close the active window generally an editor window but it may be the command line or memory status windows This is equivalent to clicking the mouse in the close box of the window Some windows such as the main display and search status windows cannot be closed Use Save to save the file being edited in the active window to disk The file will be saved in exactly the same place it was when you opened it If you want to save a copy of the file under a different name or in a different place use Save As rather than Save Note that if the file has not previously been saved i e if it was created via New you will be requested to name it using the standard Save As dialog box see below Save As Use the Save As command to save the file being edited in the active window to disk Ordinarily you will use Save As only when you want to save the file being edited with a different name in a different folder or on a different disk Name the file using the standard file dialog box that is provided PAUP 3 1 USER S MANUAL 227 S Sample data files y amp Hard Disk D feren tes Eject D Raaeyc ius gt ROiosgerins Desktop O Ragiospenm
46. The primary outgroup is never included in the ingroup so the topology of the tree in that instance is directly affected by the choice of the primary outgroup SEARCHING FOR OPTIMAL TREES If the length of a tree under a particular set of assumptions character types and weights is less than or equal to the length of any other possible tree for the same data the tree is said to be optimal most parsimonious Depending on the data set there may be a single most parsimonious tree or two or more sometimes many equally parsimonious trees Ordinarily the goal of a search is to find all of the equally parsimonious trees that exist for a particular data set under the chosen assumptions In some cases you may also be interested in near optimal trees i e trees whose length is less than or equal to some specified cutoff criterion PAUP 3 1 USER S MANUAL 31 PAUP provides two basic classes of methods for searching for optimal trees exact methods and heuristics Exact methods guarantee to find the optimal tree s but may require a prohibitive amount of computer time for medium to large sized data sets Heuristic methods do not guarantee optimality but generally require far less computer time Note that the search algorithms described below consider only binary trees In many cases however the data will not be sufficient to determine a fully resolved i e binary tree however so that branches of zero length might occur In earlier versions
47. Using a subset of the characters eee 70 The Assumptions Block oooooconnnococconcccnoncnonancconanacinnnnos 71 The TREES Block tana ia 12 TAXA and CHARACTERS Blocks coooonoconociccccocconcconoos 13 MRAP BLOCKS E TE E EE 14 Batch Process viii lips deta 75 Error Messages and Input Files oooooonoccnnnccnocanooncnonnnnnos 76 Specifying Character Types rines 78 The Standard Character Ty p s is s0 0 2 cscseasaacabedacaeessness 78 Assigning Character Polarlti8S oooocnnncnnnnnnn 78 Defining Your Own Character Types cooooccnnncccinocccinnnoss 79 Character state es 79 DLEPINAMICES Ae eR ee es 82 Verifying USERTYPE definitions 83 Assigning Character Types oooococccooccnocnconncconacananonaconnncnns 84 Character Weighting ii ini 85 Assigning Weston 85 Excluding Characters viviste oia 87 a A eR 88 Defining Ancestral States aiii o 88 Defining and Using Outgroups cccoocccncccooncnonanannnonanononononcnnnnnnnnnnns 90 Simplifying Input with Sets oe eeeeseceseceseeeseeeeaeeenaeenes 91 Character Sets CHARSETS iii 91 Taxon Sets TAXSETS asiett sagi 92 ASTUTO a riencia 93 Type sets TYPESETS Ducati calsnaia s 93 Weight sets WITSETS iii lit 93 Exclusion sets EXSETS oooooocnccccnnonnnonancnonocnononns 94 Invoking assumption Sets oooccnoccnioconoccnonncnoncnanns 94 Multistate Taxa colar ip une 95 Deleting and Restoring Tala io ies 97 Distanes Matices dia 98 Searching A a a aeaa 99
48. WTSET EXSET and ANCSTATES commands may be provided The TREES Block The TREES block is used to input user defined trees to PAUP A single TREE or UTREE command is used for each tree any number of TREE or UTREE commands may be included in the block The syntax for the TREES block follows BEGIN TREES block name TRANSLATE token taxon name token taxon name TREE name tree specification UTREE name tree specification ENDBLOCK PAUP 3 1 USER S MANUAL 79 The TREE and UTREE commands are used to input rooted and unrooted trees respectively To input multiple trees you may use as many TREE or UTREE commands as you need However you may not mix TREE and UTREE commands all of the input trees must either be rooted or unrooted Tree descriptions require that taxa be referred to by the name assigned to them in the DATA block However the TRANSLATE command can be used to define a translation table that maps arbitrary tokens in the tree specification to valid taxon names Ifa TRANSLATE command is not present a default translation table maps the integers 1 through NTAX to the corresponding taxon names in the data matrix so that integer values rather than taxon names may be used in the tree specifications However 1t is usually best to avoid relying on a default translation table For example if you rearrange the order of the taxa in the data matrix this reordering will not affect the validity of th
49. You can use the OUTGROUP command to define an outgroup For example the command outgroup gar shark specifies that the taxa gar and shark are to be considered outgroup taxa relative to the remaining ingroup You can also assign outgroup taxa by number e g outgroup 1 18 21 The taxon numbers correspond to the row numbers of the taxa in the data matrix If you want to move a taxon previously assigned to the outgroup back to the ingroup you can use the INGROUP command For example ingroup gar returns the taxon gar to the ingroup 98 PAUP 3 1 USER S MANUAL Ordinarily the effect of the OUTGROUP command is cumulative i e any taxa already assigned to the outgroup by prior OUTGROUP commands remain as outgroups when a new OUTGROUP command is issued If you want only the taxa specified in a single OUTGROUP command to be assigned to the outgroup you can use the ONLY option outgroup 18 only This is equivalent to the pair of commands ingroup all outgroup 18 You can also use the Define Outgroup menu command to move taxa from the ingroup to the outgroup or vice versa see Chapter 4 for details SIMPLIFYING INPUT WITH SETS Sets in PAUP provide a way to refer to collections of objects with single names The use of sets can greatly reduce the amount of redundant typing needed to issue commands and can help to avoid mistakes when typing commands into the command line Ordinarily you will define sets wi
50. a DOS shell IBM PC and generic DOS versions only which do not exist at the time of this writing The syntax is DOS dos command If you do not specify a command PAUP will open a DOS shell Type exit at the DOS prompt when you want to return to PAUP If you specify a command control is returned to PAUP immediately after the command has finished You can type a command that contains a semicolon by enclosing the entire command within single quotes Use the EDIT command to edit a file using PAUP s editor The syntax is EDIT file specification See The PAUP Editor in Chapter 4 for more information on using the editor This command is only valid in implementations of PAUP that provide built in editing capabilities 196 PAUP 3 1 USER S MANUAL EXCLUDE Use the EXCLUDE command to exclude one or more characters from tree length calculations see Excluding Characters in Chapter 2 The syntax is EXCLUDE character list ONLY Unless ONLY is specified characters specified in the character list are simply added to the set of currently excluded characters If characters have already been excluded and you want only those characters specified in character list to remain excluded specify ONLY any currently excluded characters not explicitly specified in the list will be re included EXECUTE Use the EXECUTE command to request processing of an input file The input file should be a valid NEXUS file see The NEXUS Format
51. able to coexist on the same consensus tree If only two trees are being compared majority rule and strict methods are equivalent Adams Consensus The method of Adams 1972 1986 see also Swofford 1991 was the first consensus method to be proposed Adams trees typically preserve more structure than strict methods The example shown below illustrates the behavior of Adams consensus The strict consensus of trees a and b would be completely unresolved as there are no groups that are exactly the same on both of the rival trees The Adams consensus tree c shows more resolution but it must be interpreted carefully If this were a strict consensus we would infer that there is a monophyletic group ACDEF on PAUP 3 1 USER S MANUAL 59 all of the rival trees and that the relationships of B and G to that group are unresolved The Adams tree makes no such claim Instead of being defined on the basis of monophyletic groups it shows nestings shared among the trees one group is said to nest within a larger group if the most recent common ancestor of the smaller group is a descendant of the most recent common ancestor of the larger group which need not require monophyly of either group In this example the most recent common ancestor of taxa A and C is a descendant of the most recent common ancestor of taxa A C and D on both rival trees therefore AC is retained in the consensus The Adams consensus reflects a great deal of shared struct
52. automatically without a prompt DEROOT If the trees being input are rooted but all character types are currently undirected you will ordinarily be asked if you want to convert the trees to unrooted trees If you specify DEROOT explicitly this conversion will occur automatically without a prompt Use the HELP command to obtain help on using PAUP s command line interface The syntax is HELP COMMANDS command name If invoked with no arguments HELP produces a list of the available commands If COMMANDS or CMDS is specified a one line description of each command is output If a command name is specified HELP provides information of that command Examples HELP requests a list of available commands HELP COMMANDS requests a list of available commands with a one line description of each HELP EXEC requests help on the EXEC command 200 PAUP 3 1 USER S MANUAL HSEARCH Use the HSEARCH command to search for optimal trees using heuristic algorithms The syntax is HSEARCH options The following options are available KEEP keep length NO COLLAPSE NO JENFORCE CONSTRAINTS constraint tree name NO JCONVERSE NO INCLUDEANC ANCSTATES ancestral states name NO STATUS See Tree searching options under Options Affecting Multiple Commands earlier in this chapter NO STEPWISE Ordinarily starting trees for branch swapping are obtained via a stepwise add
53. coding gene as follows charset coding 25 99 138 234 251 298 charset introns 100 137 235 250 charset flanking 1 24 251 276 charset noncoding introns flanking charset lstPos 25 9713 138 23213 251 2963 e 26 983 139 23313 252 2973 c 27 99 3 140 234 3 253 298 3 harset 2ndPos harset 3rdPos You could then easily refer to these regions as in the following examples which are not necessarily intended to be biologically meaningful exclude flanking excludes flanking regions wts 4 1stPos 2ndPos 2 3rdPos 1 noncoding assigns double weight to 1st and 2nd positions half weight to noncoding positions ctype tv3 coding assign a user define character type to characters in the coding regions Character sets are also available in any dialog box that contains a character selection list Popup menus allow you to instantly select all of the characters in a set Taxon Sets TAXSETs The use of taxon sets allows you to refer to a group of taxa by a single name After defining a character set using a TAXSET command you can use the taxon set name in any place that you would ordinarily use a taxon name or number For example you could define two taxon sets as follows taxset myGenus 1 15 26 40 taxset otherSpp 16 25 and then use the taxon set names in subsequent commands 100 PAUP 3 1 USER S MANUAL delete otherspp deletes taxa other than those in mygenus outgroup otherSpp lassigns ta
54. command ASSUME Use the ASSUME command to invoke a type set weight set or exclusion set see Simplifying Input with Sets Invoking Assumption Sets in Chapter 2 or to select an ancestor see Specifying Ancestral States in Chapter 2 The syntax is ASSUME TYPESET 1ype set name WTSET weight set name EXSET exclusion set name ANCSTATES ancstates name BANDB Use the BANDB command to search for trees using the branch and bound algorithm The syntax is BANDB options The following options are available KEEP keep length NO COLLAPSE NOJENFORCE PAUP 3 1 USER S MANUAL 185 CONSTRAINTS constraint tree name NO JCONVERSE NO INCLUDEANC ANCSTATES ancestral states name NO STATUS See Tree searching options under Options Affecting Multiple Commands earlier in this chapter UPBOUND upper bound Use this option to specify an upper bound on the length of the shortest tree s If you do not specify an upper bound on the length of the shortest tree or if you specify UPPERBOUND 0 PAUP computes a starting upper bound via the stepwise addition algorithm ADDSEQ FURTHEST ASIS SIMPLE ADDSEQ specifies the way in which taxa are selected for next addition to the tree at the current node of the search tree FURTHEST is usually the fastest although it is not permitted unless all characters are of type ORD or UNORD it will automatically be overridden by SIMPLE in this case NOJM
55. command e g filter lessresolv only the more highly resolved trees will be retained Reversing this filter see below will keep the polytomous trees in lieu of more highly resolved versions of them Filters can be quite complex several criteria can be applied at the same time for example filter maxlength 50 constraint testcon permdel 148 PAUP 3 1 USER S MANUAL would filter all trees below fifty steps that are compatible with the constraint tree testcon and permanently delete them If you wish to permanently remove trees from memory which do not meet the filtering criteria use the PERMDEL option or select this in the Filter dialog box The command filter constraints testl permdel would remove all trees not compatible with the constraint tree test1 The set of trees in memory available to filter will then decrease Deleted trees cannot be recovered you must either repeat the search that produced them or reload them from a file if one exists Inverting a filter is done by using the Filter Trees menu command If trees have already been filtered you can select the Reverse Filter menu command Using the NOT option with the FILTER command achieves the same result filter not constraints testcon This will keep only trees that are not compatible with the constraint tree testcon When filtering is invoked PAUP will display the number of trees originally in memory and the number of trees retained by the
56. command line these options are specified by optionally appending a suffix to the DOLLO keyword If no suffix is provided DOLLO then up is assumed DOLLO UP and DOLLO DN can be used to explicitly request the up and down options respectively In versions of PAUP that provide a menu mode pop up menus in the Set Character Types dialog box may be used to choose one of the options Likewise for irreversible characters there are three options for specifying the direction of allowed less derived to more derived transformations and disallowed more derived to less derived transformations If the ancestral state defined by the current ancestor is not equal to missing then this state represents the most ancestral state with both lower and higher ordered states in the SYMBOLS list being relatively more derived If an ancestral state is missing then the polarity may either be up or down Up specifies that states higher in the symbols order are derived relative to states lower in the symbols order 1 e for SYMBOLS 01 PAUP 3 1 USER S MANUAL 85 state O is ancestral and state 1 is derived Down specifies the opposite higher ordered states are ancestral to lower ordered states Note that if the standard ancestor is in effect all ancestral states will be missing and the up vs down setting determines the polarity Within a file or from the command line these options are specified by optionally a
57. do not constitute a partition on the most parsimonious tree s you may wish to impose ingroup monophyly as a constraint Presumably you would do this only if additional evidence supporting ingroup monophyly exists but is not contained in the data set Simply define a constraint tree with a single group consisting of the ingroup taxa and enforce that constraint during a search Another problem that sometimes arises when using this approach is the lack of sufficient information to adequately resolve outgroup relationships with the result that a large number of equally parsimonious trees differing only in the relationships among outgroup taxa are found Again if there is support for some outgroup structure that comes from information not available in the data set these aspects of the tree can be enforced through topological constraints Finally note that if you use the option to collapse zero length branches see Zero Length Branches and Polytomies below PAUP will not collapse a branch if this action would result in violation of the constraints for a tree that would otherwise satisfy them Backbone constraint trees Backbone constraint trees differ from monophyly trees in that they contain only a subset of the study taxa A trial tree is compatible with the constraint tree if pruning the taxa not present on the backbone tree from the trial tree leaves a topology identical to the backbone Unlike the monophyly constraint trees describe
58. equal weights However you may choose any vector w of weights For instance if you had reason to believe that two characters were tightly coupled for reasons other than phylogenetic history then weighting each of the characters 1 2 would be appropriate Alternatively you may have prior knowledge or at least be willing to assume that some characters are more reliable than others These characters could be given higher weight than the rest Sometimes the number of states recognized for a character is highly arbitrary This would be the case for example if a character varying continuously in size were broken into a set of discrete states The character could perhaps be broken into four states a course grained approach or ten states a more fine grained coding Remember that under equal weighting the character will actually have three times as much influence on the analysis under the ten state coding than it would under the four state coding 10 1 4 1 Likewise under the ten state coding the character would have nine times as much influence on the analysis as an equally weighted binary presence absence character PAUP provides an option for scaling character weights so that the total influence of each character is the same regardless of the number of states Specifically a binary character is weighted 1 a three state character 1 2 a four state character 1 3 and so on NOTE Some workers have suggested that the problem o
59. excluding uninformative characters HI excluding uninformative characters retention index RI and rescaled RI See chapter 1 for a detailed discussion of these statistics Below is sample DESCRIBE command with no other options i e describe 1 Tree description Unrooted tree s rooted using outgroup method Character state optimization Delayed transformation DELTRAN Tree number 1 Tree length 23 Consistency index Cl 0 696 Homoplasy index HI 0 304 Cl excluding uninformative characters 0 667 HI excluding uninformative characters 0 333 Retention index RI 0 562 Rescaled consistency index RC 0 391 Table of Branch Lengths and Linkages When a Table of linkages in the Describe Trees dialog box or the LINKS option is specified PAUP outputs a table which shows for each terminal taxon and each node which node it is linked to the assigned length of the branch under the current reconstruction and the maximum and minimum possible lengths of that branch over all MPRs see Swofford and Maddison 1987 The command format would be describe 1 links PAUP 3 1 USER S MANUAL 127 Below is the format of the output The terminal taxa are listed first followed by the internal nodes Branch lengths and linkages for tree 5 unrooted Assigned Mi ni mum Maxi mum Connected branch possible possible Node to node length length length taxonl 1 2 2 E 4 taxon2 2 taxon3 3 taxon4 4 taxon5 5 taxon6 6
60. filter It doesn t show the trees you must do that yourself to see which trees survived the filter Trees filtered according to the following criteria constraint tree PAUP 2 Number of trees originally in memory 17 Number of trees retained by filter 6 You can also filter trees directly by tree number using the FROM and TO options of FILTER or by selecting the tree range in the Filter Trees dialog box Thus the command filter from 1 to 5 would retain trees one through five NOTE Filtering trees does not reset the tree numbers the numbers are those assigned before the invocation of the filter Removing the filter restores all undeleted trees This is done by the Remove Filter menu command or the command filter off PAUP 3 1 USER S MANUAL 149 Condensing Trees PAUP can either keep all dichotomous trees whether or not there is evidence to resolve all branches or it can collapse zero length branches This will have the effect of collapsing all branches for which there is no character support After collapsing duplicate trees are eliminated from the tree buffer if a tree is collapsed to a resolution that already resides in the buffer only one of those trees is retained Remember that PAUP will not allow trees to be condensed if this action would result in violation of the constraints for a tree that would otherwise satisfy them Collapsing is the default but 1t specified in the search dialog box or by using the COLL
61. has the same influence in discriminating among alternative tree topologies In general the greater the number of states observed for a character the greater will be that character s influence e g a character with five discrete states will always contribute at least four steps to the tree length whereas a binary character may contribute as little as one step This issue is particularly relevant if the number of states recognized is arbitrary as for a continuous character broken into an arbitrary number of discrete states In this case a more fine grained coding will lead to more character states and therefore more influence on the analysis than will a relatively coarse grained coding You can use the SCALE option to request that weights be assigned such PAUP 3 1 USER S MANUAL 93 that the minimum possible length for each character is the same for all characters in a group The simplest use of this option is as follows weights scale In this case weights are assigned to all characters such that the minimum possible length of each character is 1000 the default base weight That 1s binary characters are assigned a weight of 1000 three state characters a weight of 500 and so on Because version 3 1 of PAUP uses integers to store tree lengths and character weights fractional weights such as 0 5 0 33 etc can no longer be used in PAUP You can specify a base weight other than 1000 by using the BASEWT option For example the
62. how stable the resulting topologies are THE BOOTSTRAP PAUP s implementation of bootstrap analysis follows Felsenstein 1985 The method involves sampling the original data set with replacement to construct a series of bootstrap replicates of the same size as the original data set Each of these is analyzed and the variation among these replicate estimates is taken be an indication of the error involved in making estimates from the original data In Felsenstein s approach the taxa are held constant and the characters are sampled with replacement to build a series of new data sets the same size as the original These are then subject to a search either heuristic or branch and bound Finally a majority rule consensus is constructed for all of the bootstrap trees Ifa group appears in X percent of the bootstrap trees the confidence level associated with that group is taken as X percent This method gives the investigator the ability to assign statistical confidence to hypotheses of relationship There are a number of assumptions underlying the bootstrap which are discussed in Felsenstein 1985 and in more detail in Sanderson 1989 The utility of the method like all methods depends on the validity of these assumptions We will not go into them in detail here but perhaps the most important is the i d d assumption identified by Felsenstein Felsenstein 1985 which requires that the characters be identically and independently distribut
63. however that state 0 could be assigned to these nodes in a most parsimonious reconstruction But if state 0 were assigned to either of these nodes the reconstruction would require at least two steps rather than the one step implied by the indicated reconstruction Thus state 1 is the only permissible state for these nodes 26 PAUP 3 1 USER S MANUAL A B C D E il 1 E X 1 1 1 FE 1 1 N G 1 0 Noy 0 0 H A character state reconstruction that illustrating the inappropriateness of assigning missing values to internal nodes Similarly in the example below we must assign either state 1 or state 2 to all internal nodes even though non missing states are defined only for taxa A and B A B C D 1 2 j Je 1 NA 1 ae 1 A character state reconstruction that illustrating the inappropriateness of assigning missing values to internal nodes Sometimes internal node state assignments are truly ambiguous due to the presence of missing data For example either of the two reconstructions below require the minimum length of one step PAUP 3 1 USER S MANUAL 27 A B C D A B Cc D 1 1 Es 0 1 1 2 0 Nal No 1 1 Nod NS 0 1 0 0 0 0 PAUP always chooses one of the potential reconstructions according to the character types and optimization options in effect However it also provides capabilities for examining all possible state assignments
64. implied by the tree description is forgotten once the tree has been successfully stored other methods outgroup or Lundberg rooting must be used to root the tree for output purposes In this case the tree can be rooted at any convenient point for input purposes PAUP 3 1 USER S MANUAL 155 All nondeleted taxa should be included in the tree description If a taxon is omitted it will be joined to the basal node of the tree described by the remaining taxa The examples below should clarify some of these points In particular observe that trees a and b specify the same unrooted tree but distinct rooted trees Also note that the two descriptions shown for tree c are equivalent since in the second description omitted taxa 3 4 and 5 are joined to the basal node of the tree 1 23 4 5 1 23 4 5 1 a b c 1 2 3 4 5 1 2 3 4 5 2 34 5 1 2 3 4 5 1 2 Three trees and their associated tree descriptions DEFINING AND USING TOPOLOGICAL CONSTRAINTS User defined trees may also be used to enforce constraints during a search or as a constraint in filtering trees in memory There are two types of constraint trees those that include all taxa in the matrix monophyly constraints and those that include only a subset of those taxa backbone constraints Both of these types must be defined or loaded before they can be invoked in a search Defining constraint trees is done by using the CONSTRAINTS command The t
65. in the file This is a function of the way PAUP files are constructed The alternative is to require more rigid formatting which is acceptable neither to the programmer nor to users Below are some examples of more common errors The number of taxa and or characters often get changed as an analysis progresses but if the NCHAR or NTAX settings do not match the matrix PAUP will complain in an indirect way For example the following matrix has seven characters but setting NCHAR to 6 leads PAUP to complain about the t in taxon B in the following way Invalid character state t for taxon 2 character 1 This means that PAUP was expecting a character here not an alphanumeric taxon name PAUP did as it was PAUP 3 1 USER S MANUAL 83 told it read the first six characters of taxon 1 took the next 1 to be the label for taxon 2 the carriage return to indicate that data was coming and the following t to be the first character state about which 1t complained So the error is really three lines up in the file not where PAUP places its flag begin data dimensions ntax 5 nchar 7 format missing matrix taxonA 0000001 taxonB 0100000 taxonC 1110000 taxonD po es L taxonE 1101100 r endblock If you mistakenly set NTAX to four when it should be five PAUP will return the error expecting semicolon at end of data matrix This tells you immediately that PAUP had processed all the taxa you told it to and was not expecting
66. invoke steepest descent which will not abandon a round of swapping until all minimum trees from the previous round have been evaluated it then chooses the best available tree s for the next round Without steepest descent a round of swapping ends as soon as a shorter tree is found and this tree is used as the starting tree for the next round Because the rearrangements producing greater improvements are preferred over those leading to smaller ones use of steepest descent can sometimes reduce the problem of entrapment in local optima Unfortunately being overly greedy can also produce the opposite result in some cases rearrangements with smaller decreases in tree length ultimately lead to the shortest trees whereas the rearrangements preferred under steepest descent represent blind alleys local optima You can also potentially increase your chances of finding minimal trees by swapping on nonminimal trees The first way to do this is to KEEP trees above a certain length say one or two steps above the shortest tree found so far be aware that you may end up retaining enormous numbers of trees when you do this Then you can begin swapping on those trees This may have some of the same problems as using random addition sequence but may in fact lower the water level between islands enough to allow PAUP to find more of the optimal trees 112 PAUP 3 1 USER S MANUAL Another way to estimate optimal trees is to use the NCHUCK CHUCKL
67. is used For example format equate U T R AG X EQUATE macros may not be defined recursively That is you cannot equate A to B and B to C expecting A to be expanded to C If you equate the same symbol to more than one expansion the last definition applies RESPECTCASE By default PAUP does not distinguish between upper and lower case character state symbols in the data matrix If you want upper and lower case representations of the same alphabetic character to refer to different character states specify RESPECTCASE DATATYPE STANDARD DNA RNA PROTEIN If DATATYPE STANDARD the SYMBOLS list is taken from the SYMBOLS symbols list item above default 01 If DATATYPE is set to one of the molecular sequence types a predefined SYMBOLS list is used ACGT for DNA ACTU PAUP 3 1 USER S MANUAL 173 MATRIX for RNA and the standard one letter amino acid codes for PROTEIN In addition standard ambiguity codes are implementedby predefined EQUATE macros See Predefined formats for molecular sequence data in Chapter 2 for more information on these types GAP gap symbol The gap symbol refers to the symbol used to represent alignment gaps corresponding to insertions and or deletions For example gap would assign the hyphen as the gap character The GAP setting is ignored unless DATATYPE is DNA RNA or PROTEIN Alignment gaps may be treated either as missing data or as an additional ch
68. lt iteml item2 gt Why read the manual Two or more items enclosed in curly braces and separated by vertical bars indicate a set of mutually exclusive choices The underlined item if any indicates the default setting You should not include the braces or vertical bar when you enter the command Letters words and symbols shown entirely in lowercase italics represent variables for which specific information must be supplied by the user when the command is entered An ellipsis indicates that the preceding item may be repeated one or more times Ifa group of items may be repeated sequentially the group is surrounded by angle brackets These symbols are used to define the command format see above Unless otherwise indicated they should not be typed when the actual command is entered Other punctuation should be entered as shown in the command description Many users of microcomputer software take the position that the manual is something you refer to as a last resort It is sometimes suggested that well written software largely eliminates the need for a manual by providing an intuitive menu based interface that guides the user While this idealized state of affairs may be strictly true with respect to the mechanics of running the program a thorough reading of the manual is essential in order to understand many of the biological aspects of PAUP If you do not read the manual you will have no way of discovering many us
69. menu commands Files can be exported in Hennig86 or PHYLIP formats and you can choose to export the data matrix the trees or both Exported files can then be read directly by the other program You also have the option of exporting the PAUP 3 1 USER S MANUAL 159 file as tab delimited or plain text should you wish to import the file into a spreadsheet or word processing program The only things that are preserved when a file is exported are the matrix itself and any tree descriptions contained in the data file All other PAUP MacClade instructions are not included For example the following is a sample data matrix with one user tree and two taxa transferred to the outgroup NEXUS begin data dimensions ntax 8 nchar 16 format missing matrix taxonA 0000001001017000 taxonB 0100000011011111 taxonC 111000001101 000 taxonD 111110101101 000 taxonE 1101100111011000 taxonF 1111010111117111 taxonG 101111101011 000 taxonH 1011111010117000 end begin trees translate 1 taxonA 2 taxonB 3 taxonc 4 taxonD 5 taxonE 6 taxonF 7 taxonG 8 taxonH utree PAUP_1 1 2 3 4 7 8 5 6 end begin PAUP outgroup taxonA taxonB end When this is converted to Hennig86 format it appears as 160 PAUP 3 1 USER S MANUAL xread File test converted for Hennig86 by PAUP 16 8 taxonA 000000100101 000 taxonB 0100000011011111 taxonC 111000001101 000 ta
70. of 1 the search now finds 24 trees note that the ADDSEQ ASIS and SWAP NNI settings are automatically retained hsearch hold 10 PAUP 3 1 USER S MANUAL 111 Finally switching to TBR swapping yields 25 trees This is the maximum number that any other combination of settings will find in a heuristic search as you can verify yourself by exploring the data set with other settings hsearch swap tbr Of course with other data sets you may reach a stage where trying further rearrangements does not yield any new trees and takes up vast amounts of computer time In general the best way to identify all optimal trees is to aggressively vary the swapping and addition sequence options to explore as much of tree space as possible The single best strategy is use of the random addition sequence option which will present the rearrangement algorithms with a broad range of starting trees on which to swap The tradeoff is that random trees can be very far from optimal and as such are not very good places to start to find optimal trees It may drastically increase computation time and resources to find minimal trees from a random beginning On the other hand random addition frequently leads to the finding of more if not all islands than other addition sequences You can also hold more trees during stepwise addition which may increase the effectiveness of the search but also greatly increases the time spent on adding taxa Another strategy 1s to
71. of PAUP two or more trees that could be derived through different resolutions of polytomies were always considered to represent distinct binary trees containing zero length branches with the result that many more trees were sometimes saved than was actually necessary For example if a tree contained two trichotomies nine equally parsimonious trees would be saved three resolutions for the first trichotomy times three resolutions for the second Version 3 of PAUP provides an option to collapse branches having zero length to yield polytomies Because two or more distinct binary trees may collapse to the same nonbinary tree the program then must ensure that after collapsing a tree does not become identical to a tree already saved that was obtained by collapsing a different binary tree See Zero Length Branches and Polytomies below for further details on this point Exact methods Exhaustive search The conceptually simplest approach to the search for optimal trees is simply to evaluate every possible tree Assuming that exact methods exist for evaluating the length of any particular tree an algorithm that generates all possible tree topologies and evaluates each one is guaranteed to find all of the optimal trees The algorithm outlined in the figure below can be used for this purpose Initially the first three taxa in the data set are connected to form the only possible unrooted tree for these taxa row 1 In the next step the fourth taxo
72. one or more CONSTRAINTS commands in a PAUP block The Load Constraints command provides an alternative means of defining constraints that may be simpler in some cases When you choose Load Constraints the following dialog box will appear PAUP 3 1 USER S MANUAL 253 Load constraints from file Sample data files v D 5 Taxon test D Anacyclus D Angiosperms D Angiosperms trees D Berberidaceae D Cartoon Characters D cartoon phylip D command file Replace Add to existing constraints O Load as backbone constraints The Load Constraints dialog box Select a file that is known to contain a NEXUS format TREES block The trees in this block will be read from the file and converted to constraints specifications All other information in the file is ignored If constraints have already been defined prior to issuing the Load Constraints command you can choose Replace existing constraints to replace the existing constraint tree with the new ones loaded from the file or Add to existing constraints to add the new constraint trees to the ones already present If you are loading a tree as a backbone constraint check the Load as backbone constraints box The Load Constraints command is particularly useful if you are using PAUP and MacClade in conjunction with each other It is often easier to use MacClade to graphically edit the tree descriptions that you wish to use as constraints than it is to write out the tree
73. or IRREV DN The following option is used by MacClade but ignored by PAUP POLYTCOUNT MINSTEPS MAXSTEPS This option specifies whether polytomies are treated as hard or soft see Maddison 1989 when counting the number of steps required by a character and when reconstructing ancestral states PAUP currently supports only POLYTCOUNT MAXSTEPS PAUP 3 1 USER S MANUAL 177 hard polytomies and will ignore the POLYTCOUNT specification Future NEXUS conforming programs may define additional options the ERRORSTOP setting see SET command determines how PAUP will react to unrecognized option keywords TYPESET Use the TYPESET command to define a type set The syntax is TYPESET type set name character type character list lt character type gt character list gt The character type must be one of the standard character types ORD UNORD DOLLO DOLLO UP DOLLO DN IRREV IRREV UP or IRREV DN or the name of a user defined character type Any number of character type character list pairs separated by commas may be specified See Simplifying Input with Sets in Chapter 2 for information on how to use type sets including examples USERTYPE Use the USERTYPE command to define a new character type The syntax is USERTYPE character type name STEPMATRIX CSTREE character type description Character type description must follow the rules for character state tree or stepmatrix d
74. output generated by PAUP Key This font is used to represent a key on your keyboard PAUP 3 1 USER S MANUAL 111 COMMAND Menu Command Dialog item Command names are shown in boldface type Commands that are used from input files or typed on the command line are indicated using all upper case characters Commands available for selection from menus are shown in mixed upper and lower case This font is used to refer to buttons checkboxes and other items contained in dialog boxes or elsewhere on the screen In descriptions of data file and command formats the following notation is used ITEMTEXT Optional item Items typed entirely in uppercase are to be entered as indicated Input of PAUP commands is case insensitive however so you may enter command names keywords etc in any combination of upper and lower case characters In addition PAUP allows abbreviations of command names and keywords to the shortest unambiguous truncation Note that other NEXUS conforming programs MacClade in particular see below may not accept these abbreviations Brackets around an item means that the item is optional Square brackets may be nested as in Optionalltem AnotherOptionalltem In this case each level of nesting depends on the specification of the item at the next higher level You should not include these brackets when you enter the command PAUP 3 1 User s Manual variable item
75. polymorphic or uncertain will be manifested in some constant difference in tree length Trees calculated using the polymorphic option will be longer because they force extra change within the multistate taxon see above When the uncertain option is used PAUP chooses the state that leads to a minimal tree and does not invoke any further change within the multistate taxon Complications occur when a taxon is multistate for a Dollo or irreversible character In those cases PAUP will select the state which minimizes overall tree length and satisfies the demands of the character type With Dollo characters PAUP will allow a single derivation and multiple reversals if necessary With an irreversible character PAUP will allow multiple derivations but no reversals In the case of uncertainty PAUP chooses the state from the multistate taxon which satisfies the character type and minimizes tree length So PAUP ensures that the character type is maintained over the tree and can choose the best character state from the uncertain multistate terminal Itis even possible that a character for which one or more taxa is multistate may not require any extra steps if that character is treated as uncertain However when the multistate taxon is polymorphic for a Dollo or irreversible character the best fit will be one or more derivations of each state internally plus whatever is required within the terminal depending on the type of character remem
76. round of swapping until all input trees from the previous round have been examined by the swapping algorithm See the section Heuristic Searches for more information about these options Branch and Bound Use this command to request a search using the branch and bound method This algorithm is guaranteed to find all minimum length trees Branch and Bound Search Options p Initial upper bound A compute via stepwise minimal trees only Oset to lo O all trees lt length EJ Collapse zero length branches O Include ancestor p Addition sequence E Save all minimal trees MULPARS furthest i E Enforce topological constraints O as is simple Constraint tree Keep trees that i PAT Daistribution oniy Compatible with constraint tree i y Defaults Cancel search The Branch and Bound dialog box If you do not provide an initial upper bound one will be calculated via stepwise addition see Heuristics However the better the upper bound the faster the branch and bound search will proceed so for large data sets it will generally be worth your while to perform more extensive heuristic PAUP 3 1 USER S MANUAL 257 searches in search of a better bound prior to starting the branch and bound search The addition sequence specifies the way in which taxa are selected for next addition to the tree at the current node of the search tree FURTHEST is usually the fastest although it is not permi
77. status window has been completely obscured by other windows Search Status is only available when a search is actively being performed it will be disabled at other times PAUP Help Choosing PAUP Help will bring up the help system You can view the contents by topic only or read the help entry for a particular topic When you are done using help choose the Exit button to return to PAUP The help file must be in the same folder as PAUP when PAUP is started otherwise it will not be found when Help is selected Zoom This option zooms the current window to full size Reselecting Zoom will restore it to its original size Clean Up This option will restore all windows to their default location settings If the option key is held down when this is selected only the editor windows are affected Editor windows will be stacked so that all title bars are visible Close All This option will close all open editor windows PAUP 3 1 USER S MANUAL 237 Editor windows The Windows menu will ordinarily also contain the names of additional files that are available for editing Choosing one of these filenames will cause the window for that file to be brought to the front reading the file from disk if necessary You can also select a file for editing by clicking anywhere in an already open window However the window may be hidden from view by other windows in which case the Windows menu provides a means of bringing the window to the f
78. taxonl taxon2 3 taxon3 6 5 8 7 u Do doy taxon4 taxon5 0 una Pairwise homoplasy matrix 1 2 3 4 5 taxoni taxon2 taxon3 taxon4 taxon5 duna 0 0 0 2 Stepmatrix characters and unordered multistate characters are not included in patristic distance calculations and by extension homoplasy matrices Although it is probably not a limitation in practice you must choose costs and weights for stepmatrix characters such that the total length of the longest possible tree a completely unresolved bush is less than 2 31 1 2 147 483 647 LENGTHS AND FIT MEASURES Character information can be output in slightly different format by selecting LENFIT or the Lengths and Fit Measures menu command Typically these options are used to display the lengths and fit statistics of trees in memory but also to display summary information about characters relative to trees in memory You can choose to output information about all characters SINGLE ALL or only those which vary over trees SINGLE VAR This is a very good way to get a summary of the relative fit of various characters on different trees You can choose to output any combination of the length consistency index retention index 138 PAUP 3 1 USER S MANUAL or rescaled consistency index of characters The output will include the value for each character over each chosen tree as well as an entry for the best and worst value for that character The following com
79. the desired states within curly braces or parentheses as in the following example matrix taxl T I 0 0 tax2 1 12 1 0 tax3 0 2 1 01 0 0 11 tax4 r PAUP does not distinguish between the curly braces and parentheses However in MacClade multiple states enclosed in curly braces imply the uncertainty interpretation whereas those enclosed in parentheses imply polymorphism Thus 12 in the above example would imply state 1 or state 2 whereas the 01 would imply both states O and 1 For compatibility with MacClade and in anticipation of future versions of PAUP you may want to adopt this convention now In some situations it may be helpful to use equate macros to enter multistate taxa into the data matrix The replacement of character state specifications of the form abc by a single character allows easier alignment of columns of the data matrix For example the matrix above could be equivalently defined as follows 104 PAUP 3 1 USER S MANUAL format equate a 12 b 01 matrix taxl 1100 tax2 lal0 tax3 021b tax4 0011 nd Note to MacClade Users Remember that if you created your PAUP input file using MacClade s spreadsheet editor the and amp separators used to choose between the two interpretations of multistate taxa are ignored by PAUP You will have to set the interpretation used by PAUP using one of the methods described above In practice the difference between designating multistate taxa
80. therefore unavailable for other purposes Ordinarily 1f the number of trees found during a search reaches the value of MAXTREES you will be given a chance to increase MAXTREES before proceeding However if you uncheck the Prompt for a new value box the program will not stop when MAXTREES is reached Character Matrix Format Use this command to change the way PAUP formats character state matrices for output Format Options for Character Matrix Listings Character labels vs numbers Always use numeric labels O Use alphanumeric labels if available a Use equate symbols ambiguity codes for internal node state assignments O Include data for ignored and zapped characters O Include status information Column width Defaults Cancel ok The Character Matrix Format dialog box You may specify whether to use character numbers or alphanumeric names if available for the column headings Left to right vs top to bottom mode may be chosen for both alphanumeric names and character numbers 240 PAUP 3 1 USER S MANUAL You can also control the number of columns used for each character A column width of one character works well for nucleotide sequence data but for other kinds of data some extra spacing may be more pleasing Finally you may choose whether or not to include information on character status i e deleted characters uninformative and or constant characters are flagged
81. this purpose The syntax is SAVEASSUM FILE save file name REPLACE By default if the specified file already exists you will be asked if you want to replace it To suppress this warning specify REPLACE in this case the file will be quietly overwritten SAVETREES Use the SAVETREES command to write trees currently in memory to a file as a NEXUS format TREES block or as a treefile accepted by another program The syntax is SAVETREES options The following options are available FMT NEXUS ALTNEX FREQPARS PHYLIP HENNIG Specifies the type of treefile to be produced NEXUS requests a file containing the standard NEXUS TREES block using a translation table which greatly reduces the amount of disk space required to store the trees ALTNEX also specifies a NEXUS TREES block but no translation table is used the full taxon names are included in each tree description FREQPARS requests a treefile for the FREQPARS program described by Swofford and Berlocher 1987 PHYLIP requests a treefile for input to version 3 4 of Felsenstein s 1991 PHYLIP package 212 PAUP 3 1 USER S MANUAL HENNIG requests a treefile for version 1 5 of Farris s 1988 Hennig86 program NOJBRLENS If BRLENS is specified tree descriptions will include branch lengths if the program corresponding to the FMT setting supports them NOBRLENS reverses the effect of a previous BRLENS specification FILE tree file name Specifie
82. this terminology is unstandardized Meacham 1984 and others use directed vs undirected in the same sense that I use polarized vs unpolarized My usage of directed is more consistent with graph theoretic concepts see next paragraph PAUP 3 1 USER S MANUAL Ordered Wagner Characters Ordered characters are those typically associated with the Wagner method The character states are ordered according to their position in the SYMBOLS list See The Data Block Chapter 2 for information on the SYMBOLS list The list of symbols represents a linear transformation series For example if symbols ABCDE were specified in the FORMAT command of the DATA block PAUP would treat ordered characters under the assumption that to get from state A to state E the character must proceed progressively through states B C and D as indicated in the character state graph below 1 1 1 1 000 No numerical or alphabetical order is assumed If SYMBOLS 021 state 2 is assumed to be intermediate between states O and 1 Similarly if SYMBOLS ABQC state B lies between A and Q and state Q lies between B and C No polarity is implied by the symbols list however For example all of the following transformations are consistent with the symbols list 012 there is no requirement that 0 be the ancestral state state O ancestral 0 gt 1 gt 2 state 1 ancestral 221 gt 0 state 2 ancestral 2 gt 1 gt 0 Note to PAUP Version 2 Users T
83. to control the ordering Tree Order Options Standard O Ladderize right O Ladderize left O Alphabetical The Tree Order dialog box See the section Changing the Order of Taxa on Trees The standard and alphabetical orders are best for visual comparison of different trees for the same set of taxa whereas the ladderized orders sometimes provide a more 242 PAUP 3 1 USER S MANUAL pleasing appearance i e they reduce the monkey puzzle like appearance that sometimes characterizes the standard and alphabetical orders Stepmatrices This allows you to choose which character states to allow in internal nodes Stepmatrix Options Limit internal node state assignments to states observed in terminal taxa Allow assigment of states not observed in terminal taxa to internal nodes short cuts p Candidates for state assigment All states in stepmatrix Only those states that can be identified as potential shortcuts by the 3 1 test Cancer E The Stepmatrices dialog box See the section Stepmatrix Character Reconstructions for a detailed description of these options Ignore Characters Specify how to treat invariant and or uninformative characters O Keep all characters 8 Ignore invariant characters Ignore uninformative characters Concer x The Ignore Characters dialog box Ignoring characters effectively removes them from th
84. to effects from other characters in the data set since the alternative is likely to be many independent losses As a result taxa that would ordinarily be assigned to the outgroup may spring from within the ingroup Even when no basis exists for identifying outgroup vs ingroup taxa the rooting is still likely to be more artifactual than meaningful The assumption is that the ancestor of the full tree possesses the ancestral state for the character and that the derived states must evolve somewhere on the tree from this ancestral state Thus the tree tends to be rooted nearest the taxa that have the fewest derived states This may in fact be what you want but you should at least be aware of the reasons why the program places the root where it does 12 PAUP 3 1 USER S MANUAL To amplify on the points made in the preceding paragraph note that Dollo parsimony is sometimes recommended for restriction site data e g DeBry and Slade 1985 because of the asymmetry in the probabilities of losing an existing restriction site vs gaining a new site at a particular location If a site is present any substitution at any position in the recognition sequence causes a site loss whereas even if a particular sequence is one substitution away from being a site exactly the right substitution at exactly the right position is required to convert the one off site to a site Thus a rooted Dollo analysis in which the ancestral state is assumed to b
85. to tree 1 A EEE EEE EEE EEE EEE EEE EEL EEE LEEK EEE EEE EEE LE LEE EEE EEE EEE gt 0 00 D BWP iaa aaa a 2 FEE EEE EEE EEE EEE EEE EEL EEE LE EEE EEE EEE EEE EEE EEE EEL EE CONSENSUS TREES Consensus trees are hierarchical summaries of the information common to a set of rival trees PAUP provides the following consensus tree methods strict semistrict a variation on strict Adams and majority rule These methods differ in the criteria used for retaining groups subsets of taxa in the consensus Introductory descriptions of consensus methods can be found in Swofford 1991 and Wiley et al 1991 NOTE For all consensus methods other than that of Adams 1972 if 1 the rival trees are unrooted 2 outgroup rooting is in effect and 3 there are multiple outgroup taxa then PAUP first computes an unrooted consensus and then outgroup roots the consensus tree Algorithmically this is accomplished by rooting the tree at an arbitrarily chosen terminal taxon and applying the usual rooted consensus methods As long as the specified ingroup taxa always comprise a monophyletic group this approach is exactly equivalent to outgroup rooting the rival trees first and then computing a rooted consensus However when the tree cannot be rooted such that the ingroup is monophyletic the position of the root is not well specified and the resulting consensus would depend on the method used to resolve the ambiguity in locating
86. translation tables PAUP 3 1 USER S MANUAL 275 REFERENCES Adams E N II 1972 Consensus techniques and the comparison of taxonomic trees Systematic Zoology 21 390 397 Adams E N HI 1986 N trees as nestings complexity similarity and consensus Journal of Classification 3 299 317 Archie J W 1989 Homoplasy excess ratios new indices for measuring levels of homoplasy in phylogenetic systematics and a critique of the consistency index Systematic Zoology 38 253 269 Bremer K 1990 Combinable component consensus Cladistics 6 369 372 Camin J H and R R Sokal 1965 A method for deducing branching sequences in phylogeny Evolution 19 311 326 Cavalli Sforza L L and A W F Edwards 1967 Phylogenetic analysis Models and estimation procedures Evolution 32 550 570 Constantinescu M and D Sankoff 1986 Tree enumeration modulo a consensus Journal of Classification 3 349 356 DeBry R W and N A Slade 1985 Cladistic analysis of restriction endonuclease cleavage maps within a maximum likelihood framework Systematic Zoology 34 21 34 Donoghue M J and W P Maddison 1986 Polarity assessment in phylogenetic systematics a response to Meacham Taxon 35 534 545 Faith D P 1991 Cladistic permutation tests for monophyly and nonmonophyly Systematic Zoology 40 366 375 Farris J S 1969 A successive approximations approach to character weighting Systematic Zoology 18 374 385 Farris
87. tree 6 3 1 Transversion parsimony 1133 1313 1331 Times tree favored La Total counts favoring tree 6 2 Cumulative statistics E tree means d 3 000 s 3 000 rho s rl 1 000 r2 1 000 r3 1 000 r4 1 000 r12 1 000 r13 1 000 r14 1 000 r23 1 000 r24 1 000 r34 1 000 r123 1 000 r124 1 000 r134 1 000 r234 1 000 r1234 1 000 N4 c 2 000 X2 3 000 P 0 083265 N5 c 2 000 X2 3 000 P 0 083265 F tree means d 0 500 s 1 500 rho s rl 1 000 r2 1 000 r3 1 000 r4 0 000 r12 1 000 r13 1 000 r14 0 000 r23 1 000 r24 0 000 r34 0 000 r123 1 000 r124 0 000 r134 0 000 r234 0 000 r1234 0 000 N4 c 1 000 X2 0 333 P 0 563703 N5 c 1 000 X2 0 333 P 0 563703 G tree means d 0 500 s 0 500 119 120 PAUP 3 1 USER S MANUAL rho s rl 1 000 r2 1 000 r3 1 000 r4 0 000 r12 1 000 r13 1 000 r14 0 000 r23 1 000 r24 0 000 r34 0 000 r123 1 000 r124 0 000 r134 0 000 r234 0 000 r1234 0 000 N4 c 1 000 X2 1 000 P 0 317311 N5 c 1 000 X2 1 000 P 0 317311 The summary table shows that the human chimp tree is favored both by Lake s method and by the three parsimony variants Note that for a given sequence position transversion parsimony supports a tree if a pair of taxa on one side of the central branch both have purines and the pair of taxa on the opposite side both have pyrimidines Lake attaches a different meaning to transversion parsimony requiring that the two pyrimidines be the same base and the two purine
88. use the branch and bound method as it will generally run much faster However a nice feature of the exhaustive search is that it outputs a frequency distribution of tree lengths 258 PAUP 3 1 USER S MANUAL Exhaustive Search Options of tree lengths minimal trees only Class interval all trees lt length O Include ancestor J Collapse zero length branches J Eqforce topological constraints Constraint tree Keep trees that Are Are NOT Compatible with constraint tree Defaults Cancel search The Exhaustive dialog box Although PAUP no longer enforces a limit on the maximum number of taxa that can be used in conjunction with the exhaustive search expect long running times for 9 or 10 taxa For more than 10 taxa the number of trees becomes so astronomical that the exhaustive search will not generally be feasible Lake s Invariants This command requests Lake s method of invariants for nucleotide sequence data called evolutionary parsimony by him It is only available if DATATYPE RNA or DATATYPE DNA is specified in the FORMAT command of the DATA block Mode of analysis O Choose 4 taxa Dutput spectral components i Compute branch lengths i Summary tables only Partition taxa into 4 groups se exact binomial test for n lt All possible quartets Taxa to be used Excluded taxa ros R roo B roup E lt inetude lt oros B gt Esuchude gt
89. 1 Lengths and Fit Measures This outputs length and fit measures of all trees currently in memory and or lengths and fit measures of individual characters on these trees Tree measures include the consistency index CI homoplasy index HI retention index RD and rescaled consistency index RC See the section Lengths and Fit Measures for more detail on this output EX Show overall tree lengths and or fit measures i O All characters Only characters whose lengths vary over trees O None Show range ov Measures to show i J Tree length Consistency index M Retention index E J Rescaled consistency index i K Homoplasy index i Defaults The Lengths and Fit Measures dialog box For individual character output you may choose to see the fit measures for all characters or you may restrict the output to only those characters which vary over the trees in memory This latter information is useful in judging which characters or sets of characters provide support for which trees If you have a number of trees in memory from different searches this is a very good summary of character support for each rival tree however it does not provide detailed information on specific character change such as that output by a Describe Trees command Filter Trees Use Filter Trees to temporarily limit available trees to those that satisfy certain criteria These criteria can include 1 trees les
90. 10 4 942 gorilla 7 221 4 160 5 690 orang 31 291 21 322 26 306 central 1 567 Branch lengths for G tree Branch lst estimate 2nd estimate mean human 4 266 1 438 1 414 chi mp 3 726 7 110 5 418 gorilla 7 221 2 876 5 049 orang 24 840 31 291 28 066 central 0 000 If all possible quartets or four groups are used you have the option of requesting only summary tables For all possible quartets the summary tables show the number of times significant support for particular groupings of taxa is achieved For example the command lake mode allquart taxa 1 2 3 4 5 sumtabonly generates the following output 117 118 Number of times pairs supported rejected at P Number of times Number of times pairs supported rejected at P Number of times pairs supported rejected at P Below diagonal Above diagonal 2 3 4 5 1 human 0 0 1 1 2 chimp 0 1 1 3 gorilla 0 0 0 4 orang 0 0 0 5 gibbon 0 0 1 Below diagonal Above diagonal 2 3 4 5 1 human 0 0 0 0 2 chimp 0 0 0 3 gorilla 0 0 0 4 orang 0 0 0 5 gibbon 0 0 0 Below diagonal Above diagonal 2 3 4 5 1 human 0 0 0 0 2 chimp 0 0 0 3 gorilla 0 0 0 4 orang 0 0 0 5 gibbon 0 0 0 PAUP 3 1 USER S MANUAL 0 050 Below diagonal Number of times pair supported Above diagonal Number of times pair rejected 2 3 4 5 1 human 0 0 1 1 2 chimp 0 1 1 3 gorilla 0 0 0 4 orang 0 0 0 5 gibbon 0 0 1 pairs supported rejected at P 0 010 Numbe
91. 1970 treats binary or ordered multistate characters and permits free irreversibility Multistate characters may also be left unordered i e any character state is permitted to transform directly into any other state sometimes called Fitch parsimony after Fitch 1971 Other parsimony variants place additional restrictions on the types of character state changes that are allowed Dollo parsimony Farris 1977 permits each derived or apomorphic character state to originate only once Camin Sokal parsimony Camin and Sokal 1965 prohibits reversals from a derived state to a relatively more ancestral or plesiomorphic condition Also introduced with version 3 0 of PAUP is a ISome authors e g Wiley 1979 Nelson and Platnick 1981 prefer to distinguish between the terms tree a hierarchical statement regarding genealogical ancestor descendant and sister group relationships and cladogram a branching diagram depicting patterns of character distribution or nested sets of synapomorphies This distinction is purely terminological and probably inappropriate Hendy and Penny 1984 PAUP may be used to construct to construct either trees or cladograms the only difference being how the user interpets the output Tree will be used to refer to both evolutionary trees and cladograms thoughout this manual with apologies to those who find this usage unacceptable PAUP 3 1 USER S MANUAL procedure based on the work of David Sankoff and his coll
92. 29 Pa we 1 1 1 1 A ae yx y 1 1 C 1 1 0 A B Reconstruction for a hypothetical character with ancestral state missing a and 0 b See text PAUP provides commands for converting unrooted trees stored in memory to rooted trees and vice versa Note that you do not have to invoke a root trees command before showing describing or printing trees requesting consensus trees etc see command descriptions below The only time you would ordinarily use a root trees command is if you want to declare one or more characters to be irreversible or an asymmetric stepmatrix type when there are already unrooted trees in memory in this case PAUP must store the trees internally as rooted and you will have to convert them Otherwise the trees are stored as unrooted trees and are rooted automatically for output purposes using either the currently defined outgroup or the Lundberg method at your choosing Outgroup rooting of the unrooted tree A in the figure below can be done in three ways indicated by the three arrows leading to the rooted trees Pulling the tree down at each of the different arrows leads to a different rooting the tree can be rooted such that the outgroup is a monophyletic sister taxon to the ingroup tree B it can be rooted such that the outgroup is paraphyletic with respect to the ingroup tree C or it can be rooted such that the outgroup taxa form a basal polytomy with the ingroup Tree D th
93. 78 xx x 0 40 uae 0 33 A RE 0 30 A A 0 30 AX 0 17 RR 0 12 The frequencies indicate the number of bootstrap replicates in which the particular partition was found These are called partitions because the trees are unrooted either the group represented by periods or the group represented by asterisks would be monophyletic but not necessarily both each branch partitions the taxa into those on one side of the branch and those on the other Of course if the trees were intrinsically rooted due to the inclusion of an ancestral taxon or the presence of directed character types then the partitions would correspond directly to monophyletic groups on the tree If N multiple equally parsimonious trees are found in a given replicate each group found on each of these trees contributes 1 N to the total group count This explains the presence of fractional values in the partition frequency table The bootstrap consensus tree shows the relative partition frequencies expressed as a percentage corresponding to each branch Note that percentages on the consensus tree are rounded Bots trap S0tnajority ruleconsensustree BEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EE EEE EEE EEE EEE EE taXONAT COE EEE EEE EEE EEE EEE EEE ELE EEE EEE EEE EEE EEE EE EEE EEE EE EEE EEE EEE taKON AY MEE EEE EEE EEE EEE EEE EEE EE EEE EEE EEE EE EEE EEE EE EE CHAK OND fccceceeececccecceccceccccctaxondg Boro Bec cc facer cece dec ccecccecccctaxongy fcc c
94. 86 it supports only a subset of the Hennig86 tree description format If the file is successfully converted it will be opened as a new untitled editor document Ordinarily you will want to save the document to a file before proceeding Use the Quit command when you want to exit from PAUP and return to the Finder PAUP 3 1 USER S MANUAL 233 THE EDIT MENU The Edit menu contains the standard Undo Cut Copy Paste and Clear items which are available while in PAUP s editor You may also clear PAUP s display buffer from the Edit menu Undo Typing Cut 3H Copy C Paste U0 Clear Select All BA Clear Display Buffer The Edit menu Edit Display Buffer Set Tabs Font Find SF Find Again 36 Replace B Replace All Undo Use this command to undo your last text editing action Generally the word Undo will be followed by the undo able action e g Undo Typing Undo Copy etc Cut Cut removes the selected text and replaces it on the Clipboard Any previous contents of the Clipboard are replaced The Cut command is disabled if no text is selected Copy Copy puts a copy of the selected text on the Clipboard Any previous contents of the Clipboard are replaced Unlike Cut the selected text is not removed Paste Paste inserts a copy of the contents of the Clipboard at the insertion point If text is selected at the time Paste was requested it will be replaced Clear Clear removes th
95. APSE option during a search such as hsearch addseg asis swap nni hold 1 collapse Rooting and Derooting Trees Normally trees are stored unrooted in the tree buffer They are automatically rooted when output is requested based on the rooting criteria that are in effect You need not actively request rooted trees unless you wish to define an ancestor for the full tree see ANCSTATES command or if you are using directed for example irreversible characters Rooting is achieved by using the ROOT command or the Root Trees menu command Trees are rooted according to the currently specified outgroup There are no options Once trees in memory are rooted they can be derooted by the DEROOT command or the Deroot Trees menu command Trees cannot be rooted if a filter is in effect You must first remove the filter in order to root them Printing and Plotting Trees There are two ways to output trees plotting and printing Plotting involves invoking the PLOT option in a DESCRIBE command or with the Describe Trees menu command The output can be a cladogram phylogram or both The destination of the plotted trees can either be a log file when LOG or Log Output to Disk is invoked or a printer when ECHO or the Echo to Printer is invoked The trees can be output in compressed or uncompressed form when the TCOMPRESS option is chosen or by selecting it in the appropriate dialog box Here is an example of an uncompressed and compressed tree after plot
96. ATCNT treats each character weight as if it represents the number of times the character was observed For example the weights might represent the number of times a particular pattern of character states was observed with one column of the data matrix used to represent each pattern Obviously these two interpretations have very different implications and you should decide which one makes the most sense for your analysis Implied confidence will generally be higher with weights treated as repeat counts because there will be on average more characters supporting each retained group On the other hand if the weights are drawn from an especially appropriate biological criterion e g first and second vs third positions in protein coding DNA sequences the simple interpretation is probably more justifiable The output of the bootstrap procedure consists of 1 a table showing all partitions or groups that were found in the bootstrap replications and their frequencies and 2 a bootstrap majority rule consensus tree The numbers on the branches of the consensus tree indicate the percentage of the bootstrap replications that support the group descending from that branch 122 PAUP 3 1 USER S MANUAL Partitions found in one or more trees and frequency of occurrence 12345678 Freq m area 91 01 eee m 78 66 EEEE 11 66 DETERE 63 40 AAA 42 83 ES 25 80 ES MN 8 80 RESA RR 8 79 pie 4 34 e tod 4 07 Bh 3 09 Like 2 39 x kx 0
97. AUP 3 1 USER S MANUAL 89 Therefore we let sac 3 The stepmatrix could be defined in an assumptions block as follows begin assumptions usertype weird stepmatrix 4 ABCD 0132 1 00 21 oA e O ii2 0 endblock The name weird can be any name you choose 4 is the number of rows and columns in the stepmatrix and ABCD is the list of states corresponding to these rows and columns Also the STEPMATRIX keyword is not actually required as this is the default user defined type class NOTE The diagonal elements of a stepmatrix are always set to zero in PAUP so you can put any one character symbol you want there e g or to improve readability This is an extension to the NEXUS format however so other programs may not deal correctly with nonzero entries on the diagonal MacClade for example allows you to substitute periods or hyphens for zeros but other characters will generate an error Remember that defining a character type does not automatically assign that type to the characters in the data matrix Before a user defined character type will have any effect on subsequent analyses you must assign the type to one or more characters using one of the methods described in the section below Verifying USERTYPE definitions You can use the SHOWUSER command or the Show Usertypes menu command to verify that you have defined your character types in the way you intended The resulting output will have one of the following two f
98. Archie William Day Joseph Felsenstein Wayne Maddison Chris Meacham F James Rohlf and David Swofford meeting at Newick s a local lobster house on June 24 1986 152 PAUP 3 1 USER S MANUAL annual meeting of the Society for the Study of Evolution in Durham New Hampshire User defined trees are described using a parenthetical notation that defines the shape of the tree The terminal nodes of the tree correspond to the taxa included in the data matrix and the internal nodes correspond to hypothetical ancestral taxa Each pair of parentheses encloses all members of a monophyletic group Formally the system used to describe trees is essentially the same as that for character state trees see Defining Your Own Character Types above Once again think of the tree diagram as a set of roadways connecting nodes of the tree see figure below The itinerary is to visit all of the nodes of the tree both terminal and internal in a circuit beginning at the root node following two simple rules 1 when you come to an intersection or fork in the road internal node always bear to the left and 2 when you come to a dead end terminal node turn around The path indicated by the arrows shows the sequence in which the nodes would be visited in this example PAUP 3 1 USER S MANUAL 153 One Two Three Four Five a Tree to be input as a user defined tree b Circuit followed in writing its description To write the tree desc
99. CTYPE command to assign character types to characters The syntax is CTYPE character type character list character type character list The character type must be one of the standard character types ORD UNORD DOLLO DOLLO UP DOLLO DN IRREV IRREV UP or IRREV DN or the name of a user defined character type Each character list consists of one or more character numbers character names or character set names see Character lists earlier in this chapter for details The characters specified by character list are assigned the immediately preceding character type Any number of character type character list pairs separated by commas may be specified DEFAULTS Use the DEFAULTS command to specify default option settings for another command The syntax is DEFAULTS command name options Options for the following commands may be set ALLTREES BANDB BOOTSTRAP CONTREE DESCRIBE HSEARCH LAKE 192 PAUP 3 1 USER S MANUAL LENFIT RANDTREES SAVETREES For example DEFAULTS HSEARCH SWAP NNI ADDSEQ CLOSEST HSEARCH is equivalent to HSEARCH SWAP NNI ADDSEQ CLOSEST DELETE Use the DELETE command to delete a taxon from subsequent analyses see Deleting and Restoring Taxa in Chapter 2 The syntax is DELETE taxon list options The available options are Unless ONLY is specified taxa specified in the taxon list are simply added to the set of currently delete
100. Description to this Point One One One Two One Two One Two One Two Three One Two Three One Two Three One Two Three One Two Three Four One Two Three Four One Two Three Four Five One Two Three Four Five One Two Three Four Five Explanation leaving for internal node 9 s left descendant leaving for internal node 7 s left descendant leaving for internal node 6 s left descendant terminal node One visited leaving for internal node 6 s next descendant terminal node Two visited leaving internal node 6 s rightmost descendant internal node 6 visited for last time no node information written leaving for internal node 7 s next descendant terminal node Three visited leaving internal node 7 s rightmost descendant internal node 7 visited for last time no node information written leaving for internal node 9 s next descendant leaving for internal node 8 s left descendant terminal node Four visited leaving for internal node 8 s next descendant terminal node Five visited leaving internal node 8 s rightmost descendant visiting internal node 8 for last time no node information written leaving internal node 9 s rightmost descendant circuit completed no node information written for root node 9 Note that if the tree is defined as an unrooted tree i e a UTREE command see Chapter 3 the position of the root
101. E ELE LEE EEE outgroup TORDER RIGHT A EEE EEE EEE EE EEE EEE EEE EEE LEE EE EEE EEE EEL ELE KEELE outgroup gt EEE EEE EEE EEE EEE EE EEE EEE EEE ELE taxonA fic cccccccd A EEE EEE EEE EEE EEE EEE EEE EEE ELE EEE LE EEE taxonB ficccccccccd Ae EEE EEE EEE ELE EE EEE EEE EEE taxonE ficcccccccd Ae eee Eee EEK taxonD fic ccccccccg sR er axonc fic cccccccd EEE axonF ACRE terrace axonG TORDER LEFT EEE axonF AO EEE axonG FREER OEE OOECEME CECE EE axonc MELLA OEE CEE CCCL CAL EEA RARE axonD Ay OEE EEE EEE EEL EEE ELLE axonE MERECER y ELECO COE CLEC EERE OEM CLES IMEEM MEERA AMS RE axonB LEM IM ECR CRM EEE ELLER RMSE SES EERE EMER BASES E RIERA LEE REALAGE HIOEE LAE axonA KEE EEE EEE EEE EEE EEE RE RE RE RRE REE EEE LEE KEKE EEE EEE EL EE outgroup TORDER ALPHABET CERERE GG OG GG GG OG GG OG OG OG GG GG OG OG OG OG OG OG OG OG GG OG OG OG OG OG OG OG OG OOS outgroup gt AEE EEE EEE EEE EEE EEE EEE EEE EEE ELE LE EE taxonA gt gt AEE EEE EEE EEE EEE EEE ELE EEE EEE EE taxonB fic cccccccd gt fleccccccccccccccccce taxonC ficccccccccg flecccccccd fleccececce taxonF gt fleccccccccg fic ce ccccc eorcercec taxonG ficcccccccg fiecececececececececececeeeeee taxonD Fi EERE EEE EEE EEE EEE EEE EEE EEE taxonE User Defined Trees Tree descriptions follow the Newick s 8 45 standard agreed upon by an informal committee of phylogenetic software developers at the 1986 8The final draft was agreed upon by James
102. EN option pair so named for historical reasons in conjunction with random addition but see the note about the interaction of NCHUCK and KEEP in the section above If you set CHUCKLEN to some low value such as one step and set NCHUCK to ten you will save no more than ten trees for each random replicate before the next replicate begins Tf the next replicate finds ten different trees as short as the first ten they will be added to the ten found before and so on In that way you can build a set of trees in memory say one thousand with ten trees coming from each of one hundred random searches This is a better estimate of the universe of optimal trees than saving one thousand trees from a single search which probably gives a very biased estimate of the diversity of optimal trees For some data sets you may want to try a huge number of random replicates but avoid wasting too much time on replicates that start with very suboptimal trees You can specify ABORTREP along with CHUCKLEN and NCHUCK to abandon a replicate as soon as the chucking limits are hit In this case many replicates are likely to be abandoned early but many more perhaps thousands of replicates can be performed Branch and Bound Search The details of the branch and bound algorithm are given in Chapter 1 see also the section above describing options that affect all searches To begin a branch and bound search you must either specify the initial upper bound for examp
103. HE USE OR THE RESULTS OF THE SOFTWARE OR DOCUMENTATION IN TERMS OF THEIR CORRECTNESS RELIABILITY CURRENTNESS OR OTHERWISE IN NO CASE WILL THESE PARTIES BE LIABLE FOR ANY SPECIAL INCIDENTAL CONSEQUENTIAL OR OTHER DAMAGES THAT MAY RESULT FROM USE OF THIS SOFTWARE Suggested Citation PAUP 3 1 is in many ways comparable to a monographic article In its over 70 000 lines of code PAUP implements numerous original concepts and ideas and contains many new algorithms For these reasons citation of the program in a book format is recommended Swofford D L 1991 PAUP Phylogenetic Analysis Using Parsimony Version 3 1 Computer program distributed by the Illinois Natural History Survey Champaign Illinois Copyright Illinois Natural History Survey 1989 Copyright Smithsonian Institution 1993 All Rights Reserved Apple ImageWriter LaserWriter and Macintosh are registered trademarks of Apple Computer Inc Finder and Multifinder are trademarks of Apple Computer Inc Helvetica is a registered trademark of Linotype company Microsoft is a registered trademark of Microsoft Corporation IBM is a registered trademark of International Business Machines Inc Other brand and product names are trademarks or registered trademarks of their respective holders Table of Contents PRELIMINARIES ACEO WIE MENS ai da 1 About This Mana ia ii Organization isset e ici ii Typographical and notational conventions coooonnccnnncc 111 Why
104. MAXIMUM MINIMUM and MEAN For example to request reweighting using the mean over trees consistency index you would use the command reweight fit maximum index ci Ordinarily after reweighting the characters you will conduct another tree search using the new weights The process of reweighting and searching continues until the weights no longer change Farris 1988 or until identical trees or sets of trees are found in two consecutive searches You can use the CSTATUS command to examine the weights assigned to each character following a reweight command DEFINING ANCESTRAL STATES Ancestral states are used in PAUP for three purposes First some character types including irreversible and asymmetric stepmatrix characters automatically force a hypothetical ancestor to be included in the analysis the ancestral states assign a state to this ancestor for each character Second even when it is not required you may choose to include an ancestral taxon in a search i e to explicitly specify character polarities In this case the ancestor is treated as an additional taxon and trees computed during the search are automatically rooted at the point where the ancestor connects to the tree Third if you input user defined trees as rooted trees 1 e you use the TREE rather than the UTREE command character states must be assigned to the ancestor of the full tree 96 PAUP 3 1 USER S MANUAL Ancestral states are defined by an ANCS
105. OT command to request output of character state reconstructions for one or more characters on one or more trees The reconstruction is shown by superimposing the character states assigned to each node on a plot of the tree The syntax is CHGPLOT character list options The character list specifies the character s for which reconstructions are shown and consists of one or more character numbers character names or character set names see Character lists earlier in this chapter for details The following options are available TREES tree list The tree list specifies the tree numbers for which reconstructions are to be shown If this is the first CHGPLOT command and you do not specify a tree list reconstructions are shown for the first tree only OPT ACCTRAN DELTRAN MINF ROOT OUTGROUP LUNDBERG MIDPOINT OUTROOT POLYTOMY PARAPHYL MONOPHYL NO TCOMPRESS See Options Affecting Multiple Commands earlier in this chapter If no characters are specified for either a CHGPLOT or POSSPLOT command the characters are taken to be those plotted in the last invocation of either of these commands For example CHGPLOT 1 3 5 7 POSSPLOT will cause both commands to output information for characters 1 3 5 and 7 188 PAUP 3 1 USER S MANUAL CONDENSE Use the CONDENSE command to collapse zero length branches into polytomies for all trees and then keep only those trees that are unique after the colla
106. R S MANUAL Evaluating 10000 trees sampled randomly from the set of all possible trees Trees are unrooted Starting seed 445667 Time used 00 01 31 8 Frequency distribution of lengths of 10000 random trees HK EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE 4 3 1 1 3 gt 6 17 gt 6 gt 10 gt 14 gt 18 gt 18 gt 26 122 gt 2 EEE RE REE RE RRE RE RARE REE ERRE EEE LEE EEE EEE EEE EEE LEE EE EE mean 103 055500 sd 8 043744 gl 0 458140 g2 0 035357 It is interesting to compare these results with those from the branch and bound search where we obtained the number of trees of length less than or equal to 53 steps for the same data set see Branch and Bound Search above Even though there were 49 181 trees of length less than or equal PAUP 3 1 USER S MANUAL 125 to 53 steps 3 steps longer than the minimum length trees the shortest tree found by random sampling of 10 000 trees was 71 steps Obviously given that for this 16 taxon data set there are over 2 134 x 1014 possible unrooted trees random sampling of trees is not an effective way to search for optimal trees DIAGNOSING TREES To ouput diagnostic information for a tree or a set of trees use DESCRIBE option or the Describe Trees menu command The following options are available for tree description Cladograms and Phylograms Trees output by the DESCRIBE command may be drawn either
107. Sets vj Scale for equal character weighting y regardless of states i Base weight Cancel Hane The Set Character Weights dialog box By default all characters are weighted equally with a weight of unity To change the weights select all characters to be assigned a given weight enter the desired weight in the box provided and click on Apply When you are finished assigning weights click on Done In addition to providing explicit weights you may choose to scale character weights so that each character receives the same total weight regardless of the number of states observed For instance if a data matrix contains characters with 2 3 and 4 states the 4 state characters could be weighted 2 the 3 state characters weighted 3 and the binary characters weighted 6 In this way the total range of all characters is 6 This type of weighting is especially appropriate when the number of character states PAUP 3 1 USER S MANUAL 249 assigned to a character is essentially arbitrary as when continuous measurements are recoded into discrete characters Because Version 3 0 and later of PAUP uses integer weights and tree lengths you can no longer use decimal weights such as 0 33 0 5 and 1 0 for the above example Instead specify the base weight This can either be a common denominator or a large value like 100 or 1000 For example using a base weight of 100 provides results equivalent to using decimal w
108. TANDARD This option affects searching and character state reconstruction algorithms The command is a synonym for HELP with no arguments requests a list of the available commands See help on HELP for further information Use the command to execute a UNIX command from within PAUP The syntax is unix command PAUP 3 1 USER S MANUAL 183 You can type a command that contains a semicolon by enclosing the entire command within single quotes You can open a temporary UNIX shell by typing for example csh or sh When you are ready to resume your PAUP session type Ctrl D to exit the shell and return to PAUP As of this writing the UNIX version has still not been released ALLTREES Use the ALLTREES command to perform an exhaustive search of all possible tree topologies The syntax is ALLTREES options The following options are available KEEP keep length NO COLLAPSE NO JENFORCE CONSTRAINTS constraint tree name NO JCONVERSE NO INCLUDEANC ANCSTATES ancestral states name NO STATUS See Tree searching options under Options Affecting Multiple Commands earlier in this chapter NO ED Unless NOFD is specified a frequency distribution of tree lengths is output Normally obtaining the frequency distribution is the only reason for doing an exhaustive rather than branch and bound search so you will probably never specify this option HINTERVAL interval value Interval value s
109. TATES command in the ASSUMPTIONS block In addition to user defined ANCSTATES settings a standard ANCSTATES definition is defined to have the missing state for all characters If no other ANCSTATES definition appears in the ASSUMPTIONS block the default ancestor will be the standard all missing one For example to define an ancestor that assigns state O for all characters to the ancestor named allzero you would use a command such as ancstates allzero O all Of course you do not have to assign the same state for all characters For example the command ancstates mixed 0 1 3 6 10 1 2 4 12 assigns to the ancestor named mixed state O for the first third and sixth through tenth characters and state 1 for the second fourth and twelfth characters Any character numbers not explicitly assigned a character state e g characters 5 and 11 above retain the default missing state If two different states are assigned for the same character in an ANCSTATES command the last assignment takes priority This is convenient when you want to assign a state other than missing to most of the characters and a different non missing state to a few characters For example the command ancstates mostly_a a all b 2 5 assigns state a to ancestor mostly_a for all characters other than 2 and 5 which get state b Instead of using ranges of characters as above you can also explicitly specify each ancestral character stat
110. The first number of trees to retain will be retained in memory throughout the search Ordinarily all trees initially in memory will be replaced by trees found during the search The RETAIN option is not persistent you must reset it for every HSEARCH command if you want to continue retaining the same set of initial trees NO RESETDSEED When trees containing polytomies e g due to the COLLAPSE option are input to the swapping procedure they must first be dichotomized PAUP does this by randomly resolving each polytomous node in order to obtain a binary tree The seed for this randomization is ordinarily reset to the same initial value at the beginning of every search or every random addition sequence replicate if applicable so that a search will generate repeatable results However NORESETDSEED may produce alternative dichotomizations that get lucky in generating better rearrangements INCLUDE Use the INCLUDE command to re include characters that were previously excluded see Excluding Characters in Chapter 2 The syntax is INCLUDE character list ONLY Unless ONLY is specified characters specified in the character list are simply removed from the set of currently excluded characters If you want only those characters specified in the list to be included specify ONLY characters not explicitly specified in the list will then be excluded INGROUP Use the INGROUP command to return one or more taxa to the ing
111. UAL constraints input See the section Defining and using topological constraints See the section Manipulating Trees later in this chapter for more information on defining and using trees and Chapter 3 for a description of the syntax of the TRANSLATE TREE and UTREE commands TAXA and CHARACTERS Blocks An alternative to the use of the DATA block is the combination of TAXA and CHARACTERS blocks The TAXA block contains only a DIMENSIONS command and a TAXLABELS command Its DIMENSIONS command specifies only the number of taxa as NTAX n The CHARACTERS block is essentially identical to the DATA block except that the DIMENSIONS command specifies only NCHAR c and no TAXLABELS command may be present Thus the information contained in TAXA and CHARACTERS blocks is exactly complementary For example the data in the DATA block begin data dimensions ntax 4 nchar 5 matrix taxonl 00111 taxon2 0111 taxon3 11001 taxon4 10000 r endblock could be defined equivalently as begin taxa dimensions ntax 4 taxlabels taxonl taxon2 taxon3 taxon4 endblock begin characters dimensions nchar 5 matrix taxonl 00111 taxon2 0111 taxon3 11001 taxon4 10000 r endblock If you want to do analyses that only require information about taxa e g user defined tree input and consensus tree calculation you may omit the CHARACTERS block entirely Alternatively you can include one TAXA block and several CHARACTERS and PAUP blo
112. ULPARS Ordinarily PAUP saves all minimal trees it finds during the branch and bound search You can use NOMULPARS to save only one of the shortest trees found If you only want to know the length of the shortest tree s use this option The single tree found is guaranteed to be of minimum length and the search often runs much faster NO FD The FD option requests output of a tree length frequency distribution for all trees of length less than or equal to the value specified for KEEP KEEP must be set to a value gt 0 No trees are saved if FD is specified BOOTSTRAP Use the BOOTSTRAP command to perform a bootstrap analysis using either a branch and bound or a heuristic search The syntax is BOOTSTRAP bootstrap options search options PAUP 3 1 USER S MANUAL The options pertaining directly to the bootstrap analysis are BSEED starting seed Starting seed specifies an integer between 1 and 2147483646 2712 used to seed the random number generator If you do not specify a starting seed for the first bootstrap analysis 1 is used If you do not specify a starting seed for subsequent runs the seed defaults to the next number in the random number sequence initiated during the previous run NREPS number of replications Use NREPS to specify the number of bootstrap replications resamplings to be performed The default is 100 METHOD HEURISTIC BANDB If METHOD HEURISTIC a heuristic search is perf
113. a For example begin data dimensions ntax 4 nchar 5 format labelpos right matrix 00111 taxonl 0111 taxon2 11001 taxon3 10000 taxon4 i endblock You can also use LABELPOS RIGHT in conjunction with the TRANSPOSE option In this case the character names follow the data for each character Character state symbols You may choose any combination of digits alphabetic characters or other symbols to representing character states The only symbols not allowed are whitespace characters blank tab CR and LF and the following punctuation characters a aE al E UA The SYMBOLS list defines the set of symbols that you have chosen to represent character states in the data matrix and in other commands For data other than molecular sequences see below the default SYMBOLS list is 01 which means that the only non missing character state symbols permitted are 0 and 1 If you want to use any other symbols to designate character states you must explicitly define an alternate SYMBOLS list in the FORMAT command The format of a symbols list is a sequence of single character symbols enclosed within double quotes For example if your data matrix contains some two three and four state characters you might specify symbols 0123 Alternatively you could specify symbols abcd and use alphabetic characters rather than digits in the data matrix To specify a range of digits or alphabetic characters you can use the
114. a manner similar to the PROTPARS program for analyzing PAUP 3 1 USER S MANUAL 133 amino acid sequences in PHYLIP There amino acid sequences are analyzed using a stepmatrix specifying the cost of changing from one amino acid to another see the PROTPARS sample data matrix However not all amino acids are always present in the terminal taxa When this happens execution in PAUP is slowed significantly if all character states are allowed as candidates for assignment to internal nodes PAUP provides three options for dealing with this problem First you may choose to restrict the states for internal nodes only to those observed in terminal taxa This is done using the STEPMATRIX OBSONLY option of the SET command or selecting the appropriate box in the Stepmatrices dialog box Second you may choose to allow any state contained in the stepmatrix definition to be assigned to internal nodes regardless of whether it was observed in a terminal taxon This is done using the STEPMATRIX ALLSTATES option of the SET command or the Stepmatrices menu command For example the following might be issued either in a PAUP block or on the command line set stepmatrix allstates The third option provides a compromise between these two extremes It uses the 3 1 rule Applying the rule independently to each character all triplets of observed character states are examined in turn For each triplet each member of the triplet is assigned to one of the terminal n
115. a requires 2 steps but violates the Dollo constraint Reconstruction b satisfies the Dollo constraint but requires 3 steps Note that there is no possible rooting of the tree which does not require independent parallel gains of state 1 under reconstruction A whereas any rooting of the tree is consistent with a unique origination of state 1 under reconstruction B in accordance with traditional Dollo parsimony Unless an ancestor is included in the analysis either explicitly or due to the presence of irreversible or asymmetric stepmatrix characters see Outgroups Ancestors and Roots PAUP uses the unrooted Dollo method If on the other hand an ancestral taxon is included then PAUP performs a rooted Dollo analysis If all characters are binary and are assigned polarity up as noted below the rooted analysis corresponds to the implementation of Dollo parsimony in PHYLIP Since tree lengths will in general vary according to the position of the root the result of the analysis is an intrinsically rooted tree any specification of outgroups by the user is ignored While the ability to obtain rooted trees without assuming an outgroup may seem appealing it comes at a high price Suppose a derived character state appears both in our ingroup and in an undisputed outgroup taxon also included in the data matrix The analysis will place a premium on making all of the taxa possessing the derived state a monophyletic group subject of course
116. a slightly more complicated example AYO OO O G O A B C G D E F H I Another character state tree and its NEXUS format description In unusual situations you may want to use character state trees to define a linearly ordered rather than branching character but with a different ordering from that implied by the SYMBOLS list For example if SYMBOLS 012 and you want one character to be ordered as 0 2 1 rather than 0 1 2 you could declare a user defined type with the character state tree specification 1 2 0 Other ordered characters would retain the 0 1 2 ordering You are not required to assign a character state symbol to all internal nodes of the character state tree If you do not explicitly assign a symbol 88 PAUP 3 1 USER S MANUAL to an internal node an asterisk is used For example if none of the taxa in the data matrix actually possessed any of the states G H or I the description of the character state tree shown in the above tree could have been written as A B C D E F The states corresponding to the internal nodes of the character state tree would then be shown as an asterisk in the output This practice is not recommended however as it then becomes impossible to distinguish between different nodes of the character state tree that are all represented by the same symbol in the output User defined character state trees cannot include multistate taxa In that case you must code the
117. able stepmatrix The mean distance ine ee Wk keS where Wk wk if both xik and xjx are nonmissing and zero otherwise That is the absolute distance is divided by the total weight of the characters for which neither i nor j has the state missing Mean distances are more meaningful when some taxa have much higher proportions of missing data than others Distance matrix output has the following format PAUP 3 1 USER S MANUAL 107 Pairwise distances between taxa Below diagonal Absolute distances Above diagonal Mean distances 1 2 3 4 5 6 7 8 taxonA 0 429 0 357 0 429 0 500 097 571 SN taxonB 0 357 0 571 0 500 429 857 097 taxonC 0 214 0 286 500 500 500 taxonD 0 214 571 286 286 taxonE 500 500 500 taxonF 571 571 taxonG 000 taxonH NA E SN oon uono pe Nano ou gt Rw ROW ooooso oooooo ooooooo SEARCHING FOR TREES PAUP provides two basic classes of methods for searching for optimal trees exact methods and heuristics Exact methods guarantee to find the optimal tree s but may require a prohibitive amount of computer time for medium to large sized data sets Heuristic methods do not guarantee optimality but generally require far less computer time Either exact or heuristic methods may be employed by the bootstrap See Chapter 1 for a complete description of searching methods The options available to each type of search are discussed under the appropriate heading below but some options are
118. aborators Sankoff and Rousseau 1975 Sankoff and Cedergren 1983 that allows the user to specify the cost associated with a change from each character state to each other state Each of these methods is discussed in more detail under Character Types Minimization of the total tree length is equivalent to seeking trees that imply the least amount of homoplasy or similarity not directly attributable to common ancestry Homoplasies or extra steps reversals parallelisms and convergences constitute ad hoc assumptions required to bring observations into conformity with the simpler explanation that possession of the same character state in two or more taxa is due solely to inheritance from a common ancestor A word about the relationship between computerized minimum length tree approaches and the manual methods used in traditional cladistics e g Hennig 1966 Wiley 1981 is in order The latter Hennigian methods place a heavy emphasis on a priori assessment of character polarity the specification of which of the observed character states represents the ancestral condition in the group under study fcharacter polarities could always be reliably assessed and if there were no homoplasy in the data then phylogeny reconstruction would amount to nothing more than grouping taxa according to shared derived character states synapomorphies with no relevance being attached to the sharing of ancestral states symplesiomorphies That is
119. ainframe minicomputer versions of PAUP and are fully documented elsewhere in this manual If the command line is active type in the desired command and click on the Execute button or press the Return or Enter key Click Clear or press Escape to clear the command line You can retrieve the most recently executed command by clicking on the Recall button Some users may find the command line to be a more efficient interface particularly for certain operations that require a lot of pointing and clicking within dialog boxes 236 PAUP 3 1 USER S MANUAL Show Memory Status This command adds a line at the top of the main display window that provides constant monitoring of memory availability If the memory status window has already been opened Show Memory Status brings the window to the front Two pieces of information are available 1 the total amount of memory available 1 e not already allocated by the System by PAUP or by other programs and desk accessories and 2 the size of the largest free memory block Knowing how much memory is available can be useful if you are working near the limits of the memory capacity of your machine If you no longer want to see the memory status bring it to the front and choose Close from the File menu or click in the window s close box Search Status Choosing Search Status brings the status window for a heuristic branch and bound or exhaustive search to the front It is useful when the
120. all branches whose minimum length is zero and still obtain an MPR For instance two branches may each potentially have zero length but once one of the branches is reduced to zero length the other cannot be and vice versa An example is shown below Two MPRs exist A and B for the hypothetical discrete character on the binary tree shown requiring two steps It can be seen from these reconstructions that all three of the branches may potentially be assigned no changes zero length but only the interior branch indicated by a heavy line in tree C has zero length under both reconstructions If we collapse the zero length branches under 52 PAUP 3 1 USER S MANUAL reconstructions a and b we obtain the trees and reconstructions of trees D and E respectively which also require two steps However if all branches that are potentially of zero length 1 e of zero length in at least one MPR are collapsed an MPR on the resulting tree F would then require three steps whether state O or state 1 were assigned to the single remaining internal node This example also illustrates why PAUP uses the third criterion listed above rather than the second If branches having zero length under either reconstruction a or reconstruction b but not both are collapsed then either the group A B or the group C D would be retained Clearly the decision to retain one but not both of these groups would be arbitrary PAUP retains both groups since there i
121. an index The manual is still evolving too rapidly at this point to justify the considerable amount of time that indexing would require Please report any errors or other significant omissions Unfortunately I have not yet been able to incorporate all of the many useful features that users of PAUP have requested I will continue to plug away on the depressingly long wish list The most significant new feature that I did not get working in time for this release are Archie Faith Cranston permutation randomization tests PTP T PTP Randomization tests will almost certainly be included in the next release You re welcome to check the anonymous FTP server see below for test versions containing these and other new features but of course I can make no promises as to when these test versions will be made available David L Swofford 28 March 1993 ACKNOWLEDGMENTS The number of people who have had a significant impact on the development of PAUP continues to grow and I must apologize for failing to mention everyone who has suggested features or provided assistance I am especially grateful to Julian Humphries an important source of useful ideas and expertise during the development of earlier versions It was through Julian s prodding and assistance that the first interactive mainframe version of PAUP was developed He initially convinced me to port the program to microcomputers and was instrumental in the development of the first IBM PC version
122. an use a branch and bound search You may specify that compatible trees that are below the confidence level be kept by using the KEEPALL option For example bootstrap bseed 1 nreps 10 method bandb conlevel 50 keepall PAUP 3 1 USER S MANUAL 121 Be aware however that when KEEPALL is in effect some groups may appear in the bootstrap consensus tree that are incompatible with other groups that did not appear For example If two groups are different resolutions of a polytomy they will be incompatible with one another yet one of them may be included in the consensus and one not although they appear at similar frequencies The solution to this is to carefully examine the plot of partition frequencies for other incompatible groupings with similar frequencies there may be one group found in 32 of the replicate trees that appears in the bootstrap consensus and another conflicting group found in 31 that therefore could not be included The presence of unequal character weights adds an additional complication to the use of the bootstrap Version 3 0 of PAUP simply ignored any weights that might be in effect and weighted each character equally for the bootstrap analysis This option is retained in Version 3 1 and two new options have been added The first WTS SIMPLE simply assigns each character an equal probability of being sampled but then uses the weight attached to each character during the following search The second option WTS REPE
123. and The range of other Boolean choices are lees Traas from Mo Trees von EA mode 1 mode 4 Trees current Troos from Mo mode 2 mode 5 The MODE settings are as follows Let M the set of trees originally in memory and T the set of trees from the tree file The following mode values are then available 1 replace M by T M e keep trees from the file that are not originally in memory 2 replace M by T A M keep trees from the file that are also originally in memory 3 replace M by T e replace all trees in memory by all trees from the file 4 replace M by M T e keep trees in memory that are not also in the file 5 replace M by M O T e keep trees that are either currently in memory or in the file but not both places 7 replace M by M UT i e append trees from file to trees originally in memory with elimination of duplicates 272 PAUP 3 1 USER S MANUAL The modes are derived from the following system Each of the three areas in the Venn diagram has a designated value 4 for the leftmost crescent 2 for the middle area and 1 for the rightmost crescent The mode is then simply derived by selecting the regions in the Venn diagram that you wish to include and summing the values For example keeping all trees in the treefile encompasses the 1 region and the 2 region hence it is mode 3 Since mode 6 amounts to reading the trees in memory back into memory it accomplishes nothing and is not avai
124. and see how the characters fall out 1t may be that there is no PAUP 3 1 USER S MANUAL 23 difference Another strategy 1s to choose the method which best favors the null hypothesis For example if you are studying adaptation and want to discover instances of parallel derivations of a trait choose ACCTRAN which favors reversals If a pattern of repeated parallelisms still appears in spite of a bias against them the argument for adaptation is that much stronger A third option for character state optimization is available MINF Under MINF optimization character states are optimized so that the f value of Farris 1972 is minimized Two constraints are enforced the states assigned to an HTU must be present in at least one OTU and the tree length must be minimal it is possible to further reduce the f value by increasing the length of the tree The effect of minimizing the f value is that length is transferred from interior branches towards terminal branches whenever possible minimizing the risk that groups will be arbitrarily resolved internally This option will in many instances yield the same reconstruction as DELTRAN MINF is not available for rooted trees MISSING DATA Ordinarily you will assign a state for every character in every taxon However in two situations you may need to score a character state as missing 1 the data for a particular character are simply unavailable in a taxon but you want to include the taxon
125. any more Note that this is the same message you would get if you actually did omit the terminal semicolon from matrix This brings up an important point PAUP s free format works only when the semicolons are correctly placed so that it knows when to start and stop reading a particular part of the input file For example omitting the semicolon at the end of the following hsearch command in the PAUP block would return the error message keyword set not recognized hsearch swap nni hold 10 addseq asis set tcompress PAUP thought that the SET command was part of the preceding HSEARCH command which lacked a terminating semicolon You should be tipped off that PAUP thought SET was a keyword when you know it is a command There is nothing wrong with the SET command only with the preceding HSEARCH Adding the missing semicolon will cause both commands to be executed correctly Again PAUP will do its best to flag errors in an input file but if you do get a seemingly nonsensical error message the best strategy 1s to work backwards in the file to see where that message might be appropriate If you read carefully what PAUP thought it should find and think about what might have misled it earlier chances are you won t have to go far to find the cause If you use MacClade you can largely avoid these types of errors by using MacClade s built in spreadsheet editor which insulates you from having to understand the details of the NEXUS format 84
126. apping amp Enforce topological constraints ps Constraint tree Search Keep trees that Defaults O Are Are NOT i i Compatible with constraint tree O Stepwise addition i The General panel of the Heuristic dialog box Typically you will keep only the minimum length trees but you may also be interested in near minimal trees as well You can save all trees less than or equal to a length that you specify by clicking the appropriate radio button and entering the desired length into the adjacent field You can also choose to enforce topological constraints during the search and keep trees that do do not satisfy the constraint either backbone or monophyly When Constraint trees is selected a pull down menu will appear if there is more than one currently defined constraint tree The Include ancestral taxon checkbox allows you to specify that the currently defined ancestral states ANCSTATES are to be treated as corresponding to a hypothetical ancestral taxon used to root the trees during the search In most cases you will leave this box unchecked in which case trees are stored internally as unrooted trees and rooted only for output purposes only Use it only if you have specified an ancestor different from the standard all missing ancestor If you do not all rootings will be equally parsimonious and will be saved as distinct trees PAUP 3 1 USER S MANUAL 255 The second panel allows you to choose
127. approach was inconvenient and sometimes tedious and had the further drawback that the dummy characters needed to be deleted before tree lengths and consistency indices were interpretable Furthermore it is impossible to determine the number of steps required to break up a monophyletic group appearing on the most 46 PAUP 3 1 USER S MANUAL parsimonious trees by using dummy character techniques In addition to allowing you to force compatibility with the constraint tree PAUP allows you to request acceptance of trees only if they are incompatible with the constraint tree This information provides a crude index to the strength of support for a clade Of course you are then faced with the decision as to how many steps longer a tree must be before it the existence of a clade is considered to be insupportable One way to make this judgment is through the use of the T PTP randomization tests suggested by Faith 1991 Note that although constraint trees are input as rooted trees the criterion for satisfaction of topological constraints is the same whether rooted or unrooted trees are being determined For unrooted trees the rooted constraint tree is first derooted Then each tree being evaluated is compared to the unrooted constraint tree if the trial tree is equal to the constraint tree or the trial tree can be converted to the constraint tree by deleting one or more branches from the trial tree then the constraints are satisfied An e
128. aracter state fifth base or 21st amino acid using the GAPMODE option see the OPTIONS command below The MATRIX command defines the data matrix The syntax is MATRIX data matrix The form of the data matrix depends on the options specified in the FORMAT command of the DATA block see The DATA Block in Chapter 2 for details OPTIONS The OPTIONS command of the DATA block is used to set certain options pertaining to the treatment of the data matrix The syntax is OPTIONS option specification Any or all of the following option specifications may be specified IGNORE NONE INVAR UNINFORM If IGNORE NONE all characters in the data set are used If INVAR or UNINFORM are specified invariant constant and uninformative characters respectively are ignored See the section Using a subset of the characters in Chapter 2 for definitions of invariant and uninformative 174 PAUP 3 1 USER S MANUAL MSTAXA UNCERTAIN POLYMORPH If MSTAXA UNCERTAIN multistate taxa 1 e nonsingleton character state sets in terminal taxa are interpreted as representing uncertainty about the state found in the taxon If MSTAXA POLY MORPH multistate taxa are interpreted as polymorphism within the terminal taxon See Multistate Taxa in Chapter 2 for the differences between these two interpretations Note that MacClade allows MSTAXA VARIABLE as a third possibility PAUP does not support this option If you do not explici
129. arching under topological constraints for a full description of monophyly and backbone constraint trees as well as the difference between rooted and unrooted constraints A B F C D G H E A D F C B G H E Multiple constraint trees of either type may be defined by separate CONSTRAINTS commands although only one of them may be the current constraint tree This is conveniently done in either the data file or a command file so that many different constraints can easily be tested on a given data matrix Currently defined constraint trees are displayed using the SHOWCONSTR command or the Show Constraints menu command When this is invoked backbone trees are flagged and the number of trees compatible with all monophyly constraints are displayed this is not displayed for backbone constraint trees due to the difficulty of evaluating it 158 PAUP 3 1 USER S MANUAL Constraint tree bbl backbone A EEE EEE EEL ELLE axonA A EEE EEE EEL CELE EEE EEE EEE EEL EEE EEE EEE EE EE axonB gt A BRR ee SG OG OG Ge SG OG EG axonc PERRO cl ABELLA ECCER ER EEE EK OAOE EAA LAA AERA CK EAA ER ESE axonD Constraint tree full fle cc cccccce axonA flecccccceyccccccccce taxonB 9K 666666 erronea rara EK taxonC AKO OEE EEE EEE EEK taxonD AE COE EEE EEE EEE EEE EEE EEC EEE taxonE A EEE taxonF A OEE EEE EEE ELE EEE EEE EEE EEE EEE EEL EEE EEE axonG FEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE LE EEE EEE EE ELSE axonH Of the 135 135 distinct r
130. as 1 2 3 4 5 PAUP COMMANDS Commands described in this section are specific to PAUP They may be included in the PAUP block of a NEXUS file or typed from the command line Unless otherwise specified options specified in these commands are persistent i e they retain their values between successful invocations of the command Persistence of options simplifies typing of commands because one a command requiring a large number of options has been entered once subsequent invocations of the command need not respecify all of the options The drawback is that you can become temporarily confused if you forget that a previously specified option remains in effect until you explicitly override it Options Affecting Multiple Commands Several options apply to a number of commands Specification of these options on one command affects all commands that use the same options To minimize redundancy these options are described in the following sections rather than in the description for each individual command to which the option applies Tree searching options These options pertain to the commands that request searching for trees ALLTREES BANDB HSEARCH KEEP keep length If keep length is zero only the shortest trees found will be saved If keep length zero all trees of length lt keep length will be 180 PAUP 3 1 USER S MANUAL saved By default keep length 0 so that only the shortest trees found will be saved
131. as cladograms or phylograms Cladograms convey only branching information phylograms draw branch lengths proportional to number of inferred changes according to the current reconstruction This means that cladograms will not change when a different reconstruction is chosen but phylograms will Example of a phylogram obtained by describe 1 plot phylogram fleccccccce taxonl gt gt Farc res taxon2 gt gt gt gt flececce taxon3 fil2 y fle cccccccccccccccce taxond gt gt fle ccc B PI hee CLEC ET I gt fic cccccccccccccccce taxon gt Marrero gt gt fie OEE EEK taxonb ficcccccccccld fi taxon5 The command describe 1 plot cladogram gives a cladogram for the same taxa 126 PAUP 3 1 USER S MANUAL rr rr EEE rre LEE EEE EEE EEE taxonl gt gt FEE EEE EEE EERE EEE EEE EEE EEE EEE ELE EE ELLE EEE KEE taxon2 gt gt gt gt A EEE EEE EEE EEE EEE EEE EEE EEE ELLE EEE LEK taxon3 ficcccccccel gt HOG CCE axon4 gt gt flecccccccccc8 flicccccccccel l gt Re axon 7 gt Mco gt gt LARC er axon6 ficccccccccld MR LEER EA COE AEREO ESAS axon5 Consistency Indices and Goodness of Fit Statistics Diagnostic information for selected trees is always displayed even if no diagnostic character information is requested In addition to displaying how the tree was rooted and which optimization criterion was chosen the output lists the tree length consistency index CI homoplasy index HI CI
132. atatype dna gap interleave matrix One ATGCTATCCG 1 10 Two ATCCTAGCCG Three CTGCTAGCCG Four CTGAA CG One TCATGACCTA 11 20 Two T AGACGGA Three TGGAGTCCTA Four ACATAAGTCA i endblock Alignment gaps For molecular sequence data PAUP accepts only prealigned sequences That is if all of the sequences are not of the same length gaps corresponding to insertions and or deletions must be inserted into the sequences at appropriate locations If your data matrix contains gaps 7The issue of sequence alignment is not a trivial one Different alignments for the same set of sequences can generate markedly different results Because current versions of PAUP 3 1 USER S MANUAL 75 you must define the symbol used to represent them using the GAP option of the FORMAT command see the examples above Alignment gaps may be treated either as missing data or as an additional character state fifth base or 21st amino acid using the GAPMODE option of the OPTIONS command The default is GAPMODE MISSING If you want gaps to represent an additional character state specify GAPMODE NEWSTATE You should not use GAPMODE NEWSTATE when gaps longer than one or two bases occur because a single evolutionary event 1 e an insertion or deletion will then be treated as n independent events where n is the length of the gap If you believe that the gaps convey important phylogenetic information that would be lost by using GAPMODE MISSING one strategy 1
133. ating change will lead to a preference for a single origin followed by reversal Both hypotheses will always involve the same number of steps on the tree The only change will be where those steps are located The result will be that the character change associated with a particular node will vary with the choice of optimization method For that reason you have to carefully examine the output that PAUP gives you see Character Diagnostics In the figure below assume that the ancestral state is 0 and that taxa A B C and D have the state assigned them for that character There are two ways that this character can be optimized either the occurrence of state 1 in B and C is a reversal reconstruction a in which case arose at the level of ABC and was lost in A or the presence of 1 in B and C is due to parallelism reconstruction b or independent gain In either case this character requires two steps when it is optimized so the length of the tree does not change 0 1 1 0 0 1 1 0 A B Cc D A B Cc 1 gt 0 0 gt 1 0 gt 1 0 gt a b Character optimization using ACCTRAN and DELTRAN a ACCTRAN optimization of character with distribution shown b DELTRAN optimization of character Both require two steps This is especially important if an analysis is explicitly examining the levels of parallelism reversal in a data set In that instance there are several steps that one might take Most obviously use both optimization methods
134. atives as possible each of which may potentially provide a different starting point for branch swapping see below In particular although random addition sequences are not very effective in terms of the stepwise addition process they are exceedingly useful in obtaining different starting points for branch swapping For some data sets families or islands of trees exist Hendy et al 1988 Maddison 1991 Trees from the same island are much more similar to each other than trees from different islands Suppose we define an island as a set of trees such that for every tree in the set there is at least one other tree in the set that is exactly one rearrangement away Maddison 1991 see Branch Swapping for details on what constitutes a rearrangement 38 PAUP 3 1 USER S MANUAL Generally if we find any tree in the island we can recursively find all of the trees in the island by a process of swapping on every new tree we find However by definition a tree from a different island cannot possibly be obtained Fortunately by initiating branch swapping repeatedly from different starting trees we can increase the probability of beaching on more than one island and hopefully landing on all of them One way of obtaining different starting trees would be to start from randomly generated tree topologies however these trees are usually so far from the optimal trees that searches are slow and ineffective A compromise is to use trees o
135. available for all types of searches These will be described first followed by a description of each search type All of these shared options except STATUS are available under the General window of the Heuristic dialog box Although searches are designed to find optimal trees you can request that PAUP retain near optimal trees as well If you wish to do so you can specify a tree length below which trees should be kept by using the KEEP option of the HSEARCH BANDB or ALLTREES commands or by specifying the appropriate number in the Keep selection of the appropriate search dialog box Once you have done so the optimal and near optimal trees in memory can be saved or manipulated further You can also choose to include an ancestor with the INCLUDEANC option or the Include Ancestor item in a search dialog box The ancestor will be either the standard or one you have specified previously using the ANCSTATES command If this is done trees are rooted in memory using the designated ancestor Next you may choose during any search to collapse any zero length branches If you choose not to do so all resolutions of optimal trees will be retained even if there is no evidence for some of them This is done using the COLLAPSE option or the Collapse zero length branches selection in a search dialog box Next you may choose to enforce topological constraints during the search see the section Searching under topological constraints either keeping or discar
136. be set specifically the number of replicates NREPS or the reps selection in the search dialog box a seed for the random number generator by using the RSEED option or by entering this value in the Heuristic dialog box and whether or not a running status of the random addition should be displayed the RSTATUS option or by entering this value in the Heuristic dialog box HOLD values are ignored when random addition is used The following command will begin a search with ten replicates of random addition and the initial seed set to 123321 hsearch addseg random nreps 10 rseed 123321 Once the starting trees have been obtained the only thing remaining is selection of a branch swapping algorithm These include TBR tree bisection reconnection NNI nearest neighbor interchange and SPR subtree pruning regrafting These algorithms are described in detail in 110 PAUP 3 1 USER S MANUAL the section Branch Swapping The following command will begin a search with simple addition sequence and subtree pruning regrafting branch swapping hsearch addseg simple swap spr In addition to KEEP heuristic searches also have the option of limiting the number of trees NCHUCK option greater than or equal to a specified length CHUCKLEN option which will be kept see below This is different from KEEP which specifies that all trees less than or equal a certain length are kept Thus hsearch nchuck 25 chucklen 50 will begin a he
137. ber that PAUP must explain all of the states in a polymorphic multistate taxon In the case of a Dollo character unique derivation and reversal but not multiple derivation can be tolerated in the case of an irreversible character independent derivation but not reversal are allowed Because PAUP is required to account for all the states in a polymorphic taxon and to adhere to the demands of the character type over the entire tree the states PAUP assigns internally can be very different from those that would be assigned under a condition of uncertainty This of course can lead to different tree topologies so the difference between treating multistate taxa as polymorphic versus uncertain is certainly non trivial Keep this in mind when you include multistate taxa in your analysis PAUP 3 1 USER S MANUAL 105 DELETING AND RESTORING TAXA PAUP permits interactive deletion of taxa using the DELETE command as in the following example delete salamander frog You can also delete taxa by number delete 1 10 12 21 The taxon numbers correspond to the row numbers of the taxa in the original data matrix i e before any taxa have been deleted If you want to restore a previously deleted taxon to the analysis you can use the RESTORE command For example restore frog restores the taxon named frog to the analysis Ordinarily the effect of the DELETE command is cumulative i e any taxa already deleted by prior DELETE commands
138. bers referring to one or more trees currently stored in memory If two tree numbers are separated by a hyphen this indicates that the range of trees between the first tree number and the second tree number inclusive are to be included in the list To include all trees currently in memory specify ALL as the tree list COMMANDS USED IN THE DATA BLOCK CHARLABELS Use the CHARLABELS command if you want to define alphanumeric character names to supplement the default numeric identifiers The syntax iS CHARLABELS character name list The character name list is of the form character name character name where each character namej is a valid character identifier for the ith character see Identifiers above The underscore _ character can be used as a placeholder if you do not want to assign names to some characters It is permissible to supply fewer than NCHAR names the remaining characters will then be identified by number only The length limit on these labels is 10 characters due to line length constraints in PAUP output However on input the full character label is needed Any CHARLABELS command issued from the command line overrides labels provided in the data matrix DIMENSIONS The DIMENSIONS command specifies the size of the data matrix The syntax is DIMENSIONS NTAX number of taxa NCHAR number of characters where number of taxa and number of characters are integer values Technically the limits on numbe
139. btained from random addition sequences which are often different enough to be from distinct islands but still not too far from optimal Another strategy would be to effectively lower the water level between islands and swap on nonminimal trees In that way you don t have to try and drain the ocean and swap on random trees Even when there is only one island random addition sequences can be used to circumvent the problem of entrapment in local optima Ifa nonminimal tree cannot be rearranged to produce a shorter or equal length tree the rearrangement process ordinarily terminates even though the globally optimal trees have not been found By trying several different starting trees however the chances improve that at least some of these starting trees will lead to globally rather than locally optimal trees Finally random addition sequences can be used to evaluate the effectiveness of the heuristics If you do 100 different replications under the random addition sequences and get the same 15 trees in every case you can be reasonably confident that these trees are in fact all of the optimal trees By examining the running status report available as an option you can gain some idea of whether additional replications are likely to be effective If for example the first 10 replications produce 50 trees in three different islands and the next 20 replications terminate due to finding one of those 50 trees the chances are reasonably go
140. c searching above there may be distinct families of trees that are similar with a family but very dissimilar to trees in other families Hendy et al 1988 Maddison 1991 Random addition sequence searches provide one way to discover islands Another way to examine this phenomenon is to use one of several metrics that measure the dissimilarity of pairs of trees PAUP provides only one of these the symmetric difference or partition metric Penny and Hendy 1985 which is equivalent to the Robinson and Foulds s 1981 ontraction decontraction metric Several others are calculated by the COMPONENT program Page 1993 The symmetric difference metric simply counts the number of groups that are on one tree or the other but not on both Trees that are most similar will have the 54 PAUP 3 1 USER S MANUAL lowest distance value The output is potentially useful in identifying classes of similar trees By default a matrix of pairwise distances between all trees is computed as well as a frequency distribution of the distance values You may choose to suppress one or both of the options You can also choose to compare all trees in memory to a reference tree rather than comparing all pairs of trees Here is a sample of the output generated from the 25 most parsimonious trees for the Menidia sample data set The command treedist with no options requests the full matrix of pairwise comparisons PAUP 3 1 USER S MANUAL 55 Symmetric
141. cations where character changes were assigned character changes option or where an unambiguous character changes must occur Possible states assignments option For the possible states plots a drawn at an internal node means that any of the observed states could be assigned to that node Y Print Trees This command brings up a complicated dialog box that gives you wide control over the printing of trees in memory PAUP 3 1 USER S MANUAL 267 O Show tree numbers Plot type Slanted Cladogram vj pTaxon labels O Show branch lengths s Font Helvetica v i Font Helvetica v i Page Setup Trees per Page Rooting Preview Defaults More Options Cancel Print The Print Trees dialog box Under the Plot type menu you can select the basic form of the plot whether a phylogram circletree or cladogram rectangular or slanted You must also choose a tree or range of trees from those in memory and optionally include the tree number and branch lengths in the output If there are multiple trees you may wish to save space by printing more than one tree on a page This is done using the Trees per Page option icate number of trees per page a E Rows 4 Columns 5 Trees page 20 Positioning mode Horizontal trees Vertical trees The Trees per Page dialog box Select the number of rectangles that correspond to the nu
142. cccnnncncnnnnoss 167 TRANSEATE ta os 167 TREESUTREE unas 167 PAUP Commands 3 sccccisssersacsassascleatessatensvaaccesnceecsannconstesebaaeeeenets 168 Options Affecting Multiple Command eee 168 Tree searching OPtlONS cooococnoncccnonccononcnonananinnnnos 168 Tree rooting OPtlONS oooonocononnconnnonancnoncconcnannninnno 169 Tree output options said ia 170 Options for character matrix listings 170 Other options stilo o dina 171 A a ita 171 A ci nau adc uate sinan dni sl elcid 171 ALTRES 00 nl 171 ANCS TA TES it ia 173 ASSUME spin alii ala 173 BANDE aio 173 BOOTSTRAP ccciansaets sane suas 174 CHARSE oiean a tendon gueqetoda R Raa 175 CHGPLOT soriana aida 175 CONDENSE usa ita 176 CONSTRAINTS daa ida 176 CON TREE 0 Sid esse Sale oss cede des iat 177 ESTATUS vecina nannan 179 CTYPE unas ao 179 DEFAULTS tnia noros Ae e ai t ie 179 DELETE tnea aa onea R E aE 180 DEROO Tiro cinco ici ii dd 181 DESCRIBE v eee i E AE E E OR 181 DOS o 183 EDT ai sadouceboat 183 EXCLUDE ainia ie 183 EXECUTE ias 184 EXSE nta 184 PIETER unan siii 184 GETTREBS siii iii 185 HELP A Aa 187 HSEARC Horadada it 187 INCLUDES Ios 190 INGROUP casual aiii 190 PAKE oean sabes eines a rete A debt bia at haativgtnend as 191 LEAVE ota 192 MEMA VAT Dorso 194 OUTGROUP ironia tt dit 194 POSI cai eee 195 QUito 195 RANDTRE ES da 196 RES PORE it iia ts 196 REPITE aia alo eaten contra nie 197 REW EIGHT sii ria drid 197 ROO 197
143. ccessive approximations character method Farris 1969 as implemented in the Hennig86 program Farris 1988 This method is discussed in more detail in the section A Posteriori Character Weighting CHARACTER STATE OPTIMIZATION The reconstruction of character states at internal ancestral nodes on a given tree is called character state optimization or character mapping Character optimization does not come into play at any time during the search for optimal trees only when character reconstructions are requested Under the maximum parsimony criterion the goal is to assign character states so as to minimize the total number of change required by a particular character on a given tree The set of such assignments at internal nodes is called a most parsimonious reconstruction MPR or simply an optimal reconstruction Swofford and Maddison 1987 Swofford and Maddison 1992 Characters do not have to be polarized in order for these reconstructions to be made The optimal reconstruction for a given character is a function of tree topology but also of any other assumptions about character order and or irreversibility that have been made Specifically this involves the cost of changing from one character state to another This is a function of the character type whether one of the standard types or user defined The character optimization algorithm operates by making two passes over a tree The first pass proceeds from the tips toward the r
144. ch and bound search NEXUS begin trees Treefile saved Tuesday February 16 1993 10 18 AM gt Branch and bound search settings gt Initial upper bound unknown compute via stepwise gt Addition sequence furthest gt Initial MAXTREES setting 100 gt Branches having maximum length zero collapsed to yield polytomies gt Topological constraints not enforced gt Trees are unrooted gt Shortest tree found 23 gt Number of trees retained 5 utree PAUP_1 taxonA taxonB taxonC taxonD taxonG taxonH taxonE taxonF utree PAUP 2 taxonA taxonB taxonC taxonD taxonG taxonH taxonE taxonF utree PAUP 3 taxonA taxonB taxonC taxonD taxonE taxonF taxonG taxonH utree PAUP 4 taxonA taxonB taxonC taxonD taxonF taxonG taxonH taxonE utree PAUP_5 taxonA taxonB taxonC taxonD taxonE taxonF taxonG taxonH endblock Recovering Trees from Files You can also retrieve trees from a file using the GETTREES command or Get Trees from File menu command One or more of these imported trees may then be used as starting trees for searches or the current data PAUP 3 1 USER S MANUAL 143 matrix may be optimized onto imported trees PAUP gives you the option of importing some or all of the trees in a particular treefile or only those which meet certain conditions This last category is available through Boolean operations When GETTREES or Get Trees from F
145. character state tree using a stepmatrix see below Stepmatrices To define a stepmatrix character first draw the character state graph corresponding to the assumptions that you wish to impose for the character type The character state graph should have arrows connecting all allowed transformations this is where you can incorporate the changes associated with a multistate taxon with a cost weight associated with each transformation see Chapter 1 for examples Then assign each element sjj of the stepmatrix according to the sum of the costs of all the transformations required to convert state i stepmatrix row to state j stepmatrix column If there is more than one path from state i to state j 1 e there are one or more cycles in the character state graph choose the path the allows the smallest total cost Several example stepmatrices are given in Chapter 1 Here is one more example intentionally made rather unusual in order to illustrate a few points A BCD A 0132 B 102 1 C D 1101 i i 20 A character state graph and its associated stepmatrix First note that state C can only transform into state D and vice versa since the B to D and B to C arrows are unidirectional Hence the matrix entries for C and D to A and B are infinity represented by the character in the stepmatrix definition Also observe for example that there are two paths from A to C A gt B gt C 3 steps and A B gt D gt C 4 steps P
146. character tree lengths when calculating the total length of the tree PAUP 3 1 USER S MANUAL Assigning Weights You can use the WEIGHTS WTS command or the Set Character Weights menu command to specify a priori character weights For example the command weights 2 4 10 15 18 20 3 11 12 assigns a weight of 2 to eleven characters and a weight of 3 to two characters If a weight is not explicitly assigned to a character the character retains the weight that was previously in effect i e the effect of the WEIGHTS command is cumulative The sequence of commands wts 2 1 5 more commands wts 2 10 results in the first five and the tenth characters having a weight of 2 If a more than one weight is assigned to a given character the last assignment applies This precedence rule can be used to advantage if you want to assign a nonunit weight to most of the characters For example the command weights 2 all 1 2 6 11 assigns weights of 2 to all characters other than 2 6 and 11 which receive a weight of 1 Likewise if you want to be sure that only the characters specified in a single WEIGHTS command receive nonunit weight you could issue a command such as weights l all 2 5 15 31 45 Any nonunit weights previously in effect are then reset to unity before a weight of 2 is assigned to some of the characters It is important to realize that the assumption of equal weights does not necessarily mean that each character
147. chose may not lead to all minimal trees in an island It may be that another resolution would actually have done better but was not chosen There are two solutions to this problem First you can save all of the trees found by branch swapping and input them into a second round of swapping This greatly increases the chances of finding all of the trees in an island The second solution is not to use the COLLAPSE option in the first place With COLLAPSE off you can save all dichotomous trees and use the CONDENSE command at the end of branch swapping The problem with this strategy 1s that there may be too many dichotomous trees and huge amounts of memory and long search times may be required The effect of using COLLAPSE has special significance when random addition sequences are used to construct starting trees In the normal case if PAUP finds a tree that it has found in a previous replicate it assumes that branch swapping in the previous round has already found all trees in the island to which that tree belongs therefore it abandons the current round But that assumption may not be true since there is no guarantee that the original dichotomization allowed full exploration of the island So the trees that resulted are not necessarily all of the minimal trees In that case it may be prudent not to use COLLAPSE in conjunction with random replicates This may greatly increase computation time but will avoid the problem of arbitrarily missing
148. cify which fit measure to use when calculating the new character weights RC rescaled consistency index CI consistency index RI retention index FIT MAXIMUM MINIMUM MEAN The FIT option is used to specify whether the new weights are based on the maximum minimum or mean of the fit values for each character over all of the trees in memory NO TRUNCATE Because PAUP uses integers to represent weights they are ordinarily scaled to the nearest integer by rounding Use the TRUNCATE option if you want to simply discard the fractional part e g 8 7 goes to 8 rather than 9 I can think of no reason to use this option other than to PAUP 3 1 USER S MANUAL 211 duplicate the weights obtained using the Hennig86 program Farris 1988 ROOT Use the ROOT command to convert all trees in memory from an unrooted to a rooted representation Trees are rooted according to the currently specified outgroup There are no options The circumstances under which you would need to use this command are rather unlikely see Rooting and Derooting Trees in Chapter 2 SAVEASSUM Use the SAVEASSUM command to save the current character type character weight character exclusion ancestral states taxon deletion and outgroup status to a file in ASSUMPTIONS and a PAUP block You can restore the settings in effect at the time the SAVEASSUM command was issued simply by executing the commands stored in the file use the EXECUTE command for
149. cks see below to PAUP 3 1 USER S MANUAL 81 analyze different data matrices for the same set of taxa Note that taxon labels in the characters block are just placeholders the taxon labels defined in the TAXA block are used for the remainder of the data file PAUP supports the TAXA block mainly for compatibility with Rod Page s COMPONENT program Page 1993 the latest version of which uses the NEXUS format PAUP Blocks One or more PAUP blocks may be included in the data file in addition to the standard blocks Any valid PAUP command may be placed in these blocks in fact any command outside of NEXUS blocks is assumed to be a PAUP command A typical PAUP block would look like begin paup log file my output replace endblock begin data endblock begin paup ctype ord all bandb savetrees file my trees endblock See Commands Used in the PAUP Block or from the Command Line for a description of all PAUP commands In many cases it may be easier to include commands in the file than to enter them from the command line or to make the corresponding requests via the menu system An example follows NEXUS begin paup exclude 1 5 10 12 wts 2 1 10 3 11 20 set maxtrees 200 hsearch keep 120 describe all chglist apolist endblock The advantage of keeping separate command file is that the data matrix need only be executed a single time When each command set is needed simply open it and execute it Yo
150. cology and North American Freshwater Fishes Stanford University Press Stanford Swofford D L and G J Olsen 1990 Phylogeny reconstruction Pages 411 501 in D M Hillis and C Moritz ed Molecular Systematics Sinauer Associates Sunderland Massachusetts 280 PAUP 3 1 USER S MANUAL Templeton A R 1983a Convergent evolution and non parametric inferences from restriction fragment and DNA sequence data Pages 151 179 in B Weir ed Statistical Analysis of DNA Sequence Data Marcel Dekker New York Templeton A R 1983b Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to the evolution of humans and apes Evolution 37 221 244 Wiley E O 1981 Phylogenetics The Theory and Practice of Phylogenetic Systematics Wiley and Sons New York Wiley E O D Siegel Causey D R Brooks and V A Funk 1991 The Compleat Cladist A Primer of Phylogenetic Procedures University of Kansas Museum of Natural History Special Publ No 19 Lawrence Kansas
151. command wts scale basewt 100 requests a base weight of 100 provides results equivalent to using decimal weights recorded to 2 decimals A slight problem with the use of a number like 100 or 1000 as the base weight is roundoff error For example if your data contains a mixture of two three and four state characters the scaled weights resulting from a base weight of 1000 are 1000 500 and 333 respectively Suppose two trees are being evaluated and that exactly one character is homoplastic i e requires extra steps on each tree Further suppose that the first tree requires three extra steps homoplasies in a four state character and that the second tree requires two extra steps in a three state character Even though the two trees should be considered equally parsimonious the first tree is deemed shorter since 3 x 333 999 is less than 2 x 500 1000 In this case using a base weight of 6 solves the problem the scaled weights then become 6 3 and 2 with 3 x 2 2x 3 You can mix scaled and unscaled weights in the same command For example you might want to scale the weights for just a few characters but use equal weighting for the remaining characters In this case the syntactically valid command wts scale 3 9 11 14 would probably not produce the desired results since the specified characters would receive weights of 1000 500 333 etc but the remaining characters would retain their initial weights probably 1 Inst
152. configurations Click on the boxes for Show command line and or Show memory status if you wish these to be active when PAUP starts Finally you can have PAUP remember the location of windows so that the next session will start up with the windows in the same place 246 PAUP 3 1 USER S MANUAL Restore Option Settings Use this command to set the current option settings to those that were in effect at an earlier time Restore all option settings to Current startup defaults O Factory settings cancer ok The Restore Option Settings dialog box If you request the Current startup defaults the option settings will be restored to their settings at the time the program was started If you choose Factory settings the option settings will be reset to the default values in effect at the time PAUP was originally shipped THE DATA MENU The Data menu contains commands for setting Include Exclude Characters S F Set Character Types y character types and weights excluding taxa Set Character Weights i Reweight Characters and characters from the analysis and EA displaying information concerning the current Define Outgroup data set Show Character Status Show Taxon Status Show Usertypes Show Data Matrix Show Distance Matrix Show Ancestral States The Data menu Choose Assumption Sets Save Current Assumptions Include Exclude Characters This command
153. ct methods cannot be used See the section Heuristic Search Strategies below There are too many options to give exhaustive examples At some point it is incumbent on the user to explore how changing options affects the analysis of her his data set The reason that PAUP has all these features is to make it possible for you to explore the data set in as many ways as possible Like any analytical program PAUP can only give you the tools you must decide which tools to use and when to use them Unfortunately the bottom line is that no two data sets are the same so no two heuristic searches will ever be the same Searching involves changing the search options from the default values By default zero length branches are collapsed all most parsimonious trees are kept MULPARS and steepest descent is not invoked You can find out the current search settings at any time by using the command hsearch Which will list all options available during a heuristic search and their current settings Options which are nonpersistent are marked with an asterisk PAUP 3 1 USER S MANUAL 109 Once you have selected the general options listed above that will be in effect during the search you have other decisions to make where to get starting trees for branch swapping 1f by stepwise addition how to add taxa and how to branch swap Starting trees for a search can be some or all trees in memory using FROMTREE and TOTREE options or by selecting tree
154. ction on the use of PAUP s primary functions and other features Chapter 3 provides a detailed reference to PAUP commands and input file organization Chapter 4 is specific to each implementation i e Macintosh Unix etc and describes details of the user interface that are specific to that implementation Typically you will refer to Chapter 2 to find out what PAUP can do and get a general idea of how to do it When you need more specific information you will then refer to Chapters 3 and or 4 Admittedly this organization is not ideal From a user s point of view it would be much nicer if for example the instructions for using the menu dialog box interface for setting character types were presented at the same time that the equivalent command based method was described Unfortunately maintaining n different copies of the manual for n different implementations would then become a nightmare from our point of view By isolating the implementation specific portion of the manual in a single chapter we only need to maintain a single version of the remainder of the manual Typographical and notational conventions The following typographical conventions are used throughout the manual Important term Boldface italics are used to highlight important terms the first time they are defined ER INPUT This font is used to represent input supplied by the user either from a file or the command line PAUP output This font is used to identify
155. d above backbone trees force a relative topology and other taxa may be added at any point on the backbone tree as long as the backbone is not violated For example a typical backbone topology might look like tree a below and the corresponding unrooted backbone topology would look like tree b Trees c and d are compatible with the backbone while tree e is not when taxa F G and H e those not present on the backbone tree are pruned Tree fresults which differs from Tree b PAUP 3 1 USER S MANUAL a jaa EST A BFCODGHE SY ABFCDEGH Y B A E Cc f D Backbone constraints a Rooted backbone b unrooted equivalent of rooted backbone c and d Rooted trees compatible with backbone backbone is highlighted e Rooted tree incompatible with backbone Tree connecting taxa A E is highlighted f unrooted tree for taxa A E derived from e Backbone constraints force a relative pattern of relationships In the above example the relative positions of A B C D and E on the tree must be the same Using backbone constraints is less restrictive than using monophyletic constraints in that any topology is compatible with the constraint tree as long as the relative backbone topology is preserved This means that a backbone does not force an inclusive monophyletic group as does a monophyly constraint tree The interesting feature of backbone trees is that they can also be unresolved partially as long as there are two child
156. d except for irreversible Stepmatrix characters defined below are undirected if and only if the stepmatrix is symmetric i e d dji for all pairs of character states i j see below The directedness of the characters bears a close relationship to whether PAUP stores trees as rooted or unrooted trees When directed characters are in effect trees must be treated as rooted trees because the position of the root may affect the length of the tree On the other hand if all characters are undirected PAUP ordinarily considers the trees to be unrooted since the length of the trees is independent of the position of the root Unrooted trees need not be explicitly rooted before requesting subsequent output PAUP roots trees automatically using the currently defined outgroup or via Lundberg rooting whenever necessary In addition characters are either polarized or unpolarized Polarized characters are those for which the state ancestral to all other states 1s prespecified If the ancestral state is not specified or is designated as missing the character is said to be unpolarized A character type may formally be described as a weighted directed graph weighted digraph The vertices of the graph correspond to character states and the edges of the graph are arrows corresponding to permissible character state changes The weight of each edge is the cost associated with the transformation from one character state to another 3Unfortunately
157. d taxa If taxa have already been deleted and you want only those taxa specified in taxon list to remain deleted specify ONLY any currently deleted taxa not explicitly specified in the list will be restored The following three commands pertain if there are trees in memory that will become invalidated by the deletion of taxa PRUNE If PRUNE is specified newly deleted taxa will be removed pruned from the trees currently in memory which otherwise remain unmodified CLEAR If CLEAR is specified any trees currently in memory are simply deleted CONDENSE If CONDENSE is specified any duplicate trees that result from the removal of taxa PRUNE option are deleted If you do not specify PRUNE or CLEAR the program will prompt for your desired action PRUNE and CLEAR are PAUP 3 1 USER S MANUAL 193 mainly useful for batch file processing where you do not want the program to stop and wait for a response from the user before continuing DEROOT Use the DEROOT command to convert all trees in memory from a rooted to an unrooted representation There are no options The circumstances under which you would need to use this command are rather unlikely see Rooting and Derooting Trees in Chapter 2 DESCRIBE Use the DESCRIBE command to output tree diagrams and associated information The syntax is DESCRIBE tree list options The tree list specifies the numbers of the trees you wish to describe If this
158. dentical to that for a heuristic search except for when the FD is selected For example using the sample Menidia data set the command bandb fd keep 53 outputs the following Frequency distribution of tree lengths Mr E 50 gt 70 51r 959 IIA ASA EA ARA A 40627 Dr EEL EEE EEE EE EEL EEE LEE EEE EEE EEE ELE LE mean 52 803725 sd 0 453089 gl 2 354361 g2 5 528415 This tells us that there are over 8 500 trees within two steps of the shortest and over 49 000 trees within three steps of the shortest The statistics for this distribution are ordinarily not meaningful due to the fact that it has been artificially truncated but they are output anyway Note that no trees are saved when this option is in effect Exhaustive Search Exhaustive methods are not generally useful for more than 10 or 11 taxa since there are over 2 million trees for 10 taxa and 34 million trees for 11 taxa Computation of all trees for larger numbers of taxa is impractical at present With very large data sets heuristic methods may be the only avenue available for analysis If however you do have a data set small enough for exhaustive analysis there are several useful options besides the standard search options see the section above on options affecting more all search algorithms Obviously addition sequence is irrelevant in an exhaustive search since all trees are going to be evaluated anyway The most useful option is the ability to calculate a
159. description in the parenthetical notation used by PAUP MacClade can save the trees to a file from which constraints can be loaded by PAUP Show Constraints This command causes a tree representing the currently chosen constraint definition to be output You may also choose to see other constraint definitions as well Select constraints to show backbonel a backbone2 monophyly3 EJ Compress vertically Cinca Toe The Show Constraints dialog box 254 PAUP 3 1 USER S MANUAL Backbone and Monophyly constraints will be flagged in the output Heuristic Use Heuristics to request a search using a heuristic approximate algorithm A heuristic search ordinarily includes two components 1 stepwise addition of taxa to a developing tree until all have been connected and 2 rearrangement of these trees via branch swapping techniques There are too many heuristic search options to include them all in a single dialog box Consequently the controls for these options are divided into four panels To select a different panel of options simply click on one of the radio buttons at the left of the dialog box The first panel of options can be used to specify several general options Heuristic Search General Heuristic Search Options Options i Minimal trees only General o i All trees lt length i i Starting trees i O Include ancestor i XX Collapse zero length branches Branch sw
160. difference distances between trees 1 2 3 4 R10 0 DU Y Na gt WD 010 a 00 JN wv Donnu Ol UN Ud RN a 00 UD UN A DM MN OM UI O o W 10 010 EN WWHNYHE DUH od a dr Symmetric differ 25 25 Distribution of O44 ECREKECEEE 0 rra ETB rra LAA AAA ERRADA rra rra rra ld gt gt gt gt gt gt gt gt gt gt gt 0 1 2 3 4 5 6 7 8 9 0 1 5 6 Po Dow0YRvAaN OU ooo Doa BNW eooc a NN 001 IN Y o nu e A As ence d sy mmet LEER 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Hi RVWHYOCWHNWE DDH VYAWR HUE FOWCOHFWUN o e O Oe O Dun OS OS Os Y W UNID Ya o e NS TO AT istances between ric difference d ELEOEECECOLESERES reraaaaa 20 boo uno ono wn ww sono pa NOU SWUE NY ROY RUI on an ws DN uo ooo Poo uw Ronson won woon trees continued stances between trees COOOL EK 33 rra rca 45 rra ra 147 rrraraaara cr 39 cia ci nn era 38 32 FEE EEE EEE EEE EEE EEE LEE EE EEE EEL EEE LEEK EEE EE EEE EEE LEE To limit the analysis to comparisons of one tree against all others specify a tree number for the COMPARE option treedist compare 1 56 PAUP 3 1 USER S MANUAL Symmetric difference distances to tree 1 Tree number D 3 6 7 8 91011 1243 14 15 16 17 18 19 20 21 Distance ee AP Qe be a Dee I BS O SOB De IP E A A Tree number 22 23 24 25 Distance 9 8 9 3 Distribution of symmetric difference distances
161. ding trees that are 108 PAUP 3 1 USER S MANUAL compatible with the constraint tree This is done using the ENFORCE option with CONVERSE or NOCONVERSE or the Enforce topological constraints selection in the appropriate search dialog box Remember that you must have previously defined a constraint tree in order to invoke this option Finally you can request that PAUP output the status of the search using the STATUS option One option that affects only heuristic and branch and bound searches is MULPARS or save all minimal trees in a Heuristic dialog box Searches without MULPARS in effect can sometimes be significantly faster because less of the search tree needs to be explored However at the end of the search you will know the length of the shortest tree s but you will have no idea how many trees exist at that length Heuristic Searches Heuristic searches generally provide the fastest way to find optimal trees but the results being approximate may depend on the way in which the search is conducted Unlike the exact methods heuristic methods may find only a subset of the optimal trees for a given data set and a given set of search parameters or may fail to identify minimum length trees at all In addition no one combination of settings will provide the best results for all data sets You must be prepared to spend some time exploring different options to find optimal trees This is especially true for large data sets for which exa
162. e and increases the size of several punctuation characters for improved readability You can set font and tab settings for individual windows using the Set Tabs and Fonts dialog under the Edit menu Font tab settings chosen there automatically override the default settings You can change the font tab settings for all open windows by checking the box Apply to all open windows in the Editor menu command To edit a data file choose the Edit option in the Open dialog box When the file is opened a window will appear with the data file in it You can then make changes to the data file as you like In order to use the file you must then execute it whether or not you saved your editing Alternatively if you have opened and executed the file already it will appear in the Windows menu Simply select the file and the editing window will appear If you close the window without saving your changes PAUP will prompt you for the appropriate action If you wish to create a new file choose the New dialog box An untitled window will appear You can enter a new data matrix or paste in data from another file When you are done save your work and execute the file as above THE COMMAND LINE All of PAUP s capabilities may be invoked using either the menus or the command line PAUP 3 1 USER S MANUAL 223 Anda for command mae E The Command Line To use the command line choose Show Command Line from the Windows menu and type in the de
163. e ANCSTATES command or if you are using directed characters see Character Types in Chapter 1 you should define rooted trees instead In any case you can convert rooted trees to unrooted trees and vice versa using the ROOT DEROOT commands or the Root Trees Deroot Trees menu commands Note that in order to write the tree descriptions unrooted trees may be rooted at any convenient point including a terminal taxon or internal node the position of the root is simply ignored when the tree is stored PAUP currently ignores the tree names trees are referred to subsequently only by number in the order in which they are presented in the TREES block MacClade uses the tree names however If the name is preceded by an asterisk the tree becomes the default tree E g a DESCRIBE command with no tree list will result in the description of the default tree PAUP 3 1 USER S MANUAL 179 If a taxon is omitted from the tree specification it is assumed to descend from the root of the tree described by the remaining taxa See Manipulating Trees User Defined Trees in Chapter 2 for instructions on writing tree descriptions Note to PAUP Version 2 Users The new standard requires commas in places in which versions 2 4 and earlier allowed them to be omitted In particular commas are now generally required between a right parenthesis and a left parenthesis or taxon number E g the old description 1 2 3 4 5 must now be written
164. e site absent will tend to root the resulting tree s near the taxa that have the fewest sites Although one could argue that this approach to rooting is reasonable it seems a bit arbitrary to me There are ways of circumventing the problems of using rooted Dollo parsimony for characters like restriction site data One approach is to infer character polarity via traditional outgroup analysis and then use a mixture of Dollo and Camin Sokal irreversible characters see below parsimony If the site occurs only in some members of the ingroup then we would designate the ancestral state as absent and allow a single gain of the site followed by as many losses as would be required to explain the character 1 e traditional Dollo parsimony But if a site occurs in the ingroup and in the outgroup the assumption that the site was already present in the common ancestor of the ingroup plus outgroup would be reasonable In this case we would designate the ancestral state for this character as present and then treat the character as an irreversible rather than Dollo character allowing only losses of the site in accordance with the Dollo model A somewhat simpler but logically equivalent approach is to constrain the ingroup to be monophyletic either through the use of a heavily weighted dummy synapomorphy or by using the topological constraints feature of PAUP and use Dollo parsimony with an all absent all zero ancestral taxon for all o
165. e Filter Trees menu command For example filtering trees below a certain length would be accomplished using the following command filter maxlength 50 There are several different filtering criteria you may select Filtering trees within three partition metric units from a specified tree tree 3 would be done by filter sd 3 from 3 Filtering with constraints would be invoked using the command filter constraints testcon where the constraint tree testcon will be used This can be either a backbone or monophyly constraint tree Constraint trees must be defined before they can be invoked as is the case when searching under constraints Although several constraint trees may be defined only one may be invoked in a particular filter You also have the option of keeping polytomous trees only if there are no other compatible resolutions of them in memory For example two resolutions of a trichotomy might receive potential support but the third resolution might have no conceivable character changes supporting it in which case the zero length branch would collapse to the trichotomy In this case a search will retain trees containing the two resolutions of the trichotomy as well as trees containing the unresolved trichotomy resulting from the third resolution If you check the Do not retain a polytomous tree for which a more highly resolved compatible tree exists item in the Filter Trees dialog box or specify the LESSRESOLV option of the FILTER
166. e Me cccg ccc ccce cece taxonkg horario cer recen cer erecererecerececcceeectaxonB FCCC EEE CELE EEE EEE EEE EEE ECE EE EEE EC EaKONAG The consensus tree that is ouput for the bootstrap will in general be longer than the minimal tree for the full data set although this is a function of consensus trees in general and is not unique to the bootstrap This is because in PAUP polytomies are treated as representing simultaneous multiple splitting hard polytomies Maddison 1989 rather than as indicating uncertain resolution soft polytomies PAUP 3 1 USER S MANUAL 123 Random Trees If you are interested in the tree length frequency distribution see Exhaustive Search above but the number of taxa is too large to perform an exhaustive search you can approximate the shape of the distribution by randomly sampling trees under a model in which every possible tree is equally likely As in the case of an exhaustive search you may save the tree length distribution to a file for input into other programs Random trees may be generated using the RANDTREES command or the Random trees menu command For example the following will evaluate the lengths of 10000 random trees with the seed set to 445667 randtrees tseed 445667 nreps 10000 Output consists of a tree length frequency distribution which may be saved to a file as in an exhaustive search For the sample Menidia data set the resulting output is as follows 124 PAUP 3 1 USE
167. e analysis See the section Using a subset of the Characters Semigraphics PAUP uses special characters in its internal font to draw trees and other items These characters are neither in the standard 128 ASCII characters nor in the set of special characters normally included with Macintosh fonts Thus although the trees look nice when drawn in the display PAUP 3 1 USER S MANUAL 243 buffer they will not look right when printed on most printers Ordinarily PAUP translates these semigraphics characters to standard ASCII substitutes when sending output to a printer file or document window The Semigraphics command allows you to alter this behavior via the following dialog box How should nonstandard semigraphics characters in output files editor windows and printer output be treated Translate to standard ASCII substitutes Leave alone The Semigraphics dialog box Editor The Editor command allows you to change options that affect PAUP s editor Default tent settings for editor windows evsesssnsessnnsesennessonsesssneessnnesssnuessensesssnecssnnesssnnesssnensesnee Font PAUPMonaco w Size Tab Spacing J AutoIndent Apply to all currently open windows Creator field for files saved by PAUP that were created by i another application determines application launched when document icon double clicked Leave unchanged Always reset to PAUP
168. e by using the VECTOR format option This is useful if not necessary if the character states for the ancestor are very heterogeneous For example an ancestor with five characters would be specified by the following VECTOR statement where name is the name of the ANCSTATES set ANCSTATES name VECTOR 0 1 2 0 2 Any number of ancestors may be defined via multiple ANCSTATES commands but only one is in effect at any one time You may select the current ancestor in any of three ways e By preceding the name in an ANCSTATES command with an asterisk For example the commands PAUP 3 1 USER S MANUAL 97 ancstates allzero O all ancstates allone l all ancstates mixed 0 1 10 1 11 20 define three ancestors with ancestor allone being the currently chosen ancestor By using the ASSUME command For example the command assume ancstates allzero would make ancestor allzero the current ancestor e By using the Choose Assumption Sets menu command ANCSTATES commands are typically issued from within an ASSUMPTIONS block however you may also issue them from within a PAUP block or via the command line DEFINING AND USING OUTGROUPS Unless you are using directed characters or you choose to include an ancestral taxon the trees found by PAUP searches are unrooted see Outgroups Ancestors and Roots in Chapter 1 The most commonly used method for rooting these trees is to include an assumed outgroup in the analysis
169. e interleave option or use more than one line for each taxon print from within MacClade or open the file in an application that automatically wraps long lines most word processing programs Compute Consensus Use the Compute Consensus command to compute consensus You will usually compute the consensus for all trees currently in memory but you can select a subset 1f desired Consensus Tree Options E Strict O Semistrict combinable component O Adams OMajority rule O show other compatible groupings Trees to include Y O Output to treefile in consensus nae O Consensus indices Rii EJ Compress vertically Charca Toe The Compute Consensus dialog box Consensus indices may also be computed These are described in Rohlf 1982 and Swofford 1991 Available consensus techniques are semistrict strict Adams and majority rule You can include consensus indices and output the consensus to a treefile of your choosing Print Consensus This allows you to specify the format of the consensus when it is printed 270 PAUP 3 1 USER S MANUAL EI Show consensus type Li idth A Plot type Slanted v ine width e IP pt rTaxon labels o BI Show group frequencies Font Helvetica e Helvetica y Trees per Page Le DEN in 1 8 0 5 Jin Defaults Page Setup Print The Print Consensus dialog box The options are generally the same as for
170. e it as a constraint or perform other tree manipulations on it you must first save it to a treefile using the SAVE option and then load it into memory as you would any other tree In fact you may wish to calculate consensus trees for different subsets of trees in memory and store them in separate files for further comparison contree 1 2 3 strict save file summary con 146 PAUP 3 1 USER S MANUAL will calculate the strict consensus for trees 1 2 and 3 and save it in the file summary con You may also specify ADAMS for an Adams consensus MAJRULE for a majority rule consensus or SEMISTRICT for a semistrict consensus Any or all of these may be specified on the same command line or on the same line of a PAUP block Consensus indices will be printed following each tree if the INDICES option is included or they are selected in the Compute Consensus dialog box For example contree 1 2 3 strict adams indices file summary con will calculate and store the strict and Adams consensus trees and include consensus indices for both in the file summary con For majority rule consensus trees a table of partition or group frequencies can be requested the percentage cutoff specified and compatible trees occurring in less than fifty percent of the trees can be included by specifying contree 1 2 3 majrule cutoff 50 dotplot le50 The format of the partition frequency table is Partitions found in one or more trees and frequency of occurr
171. e locally parsimonious within the ingroup but not globally parsimonious If the trees being evaluated are rooted either implicitly because directed characters are present or explicitly due to the user s decision to include an ancestral taxon the character states taken by the ancestor of the full tree must be specified By default these ancestral states are set to missing the standard ancestor however the user may make alternative assignments using ANCSTATES definitions see below The default setting of all missing values was chosen for two reasons First if all characters are undirected trees may be converted from rooted to unrooted or vice versa with no change in tree length Second if directed characters are present the program has the freedom to assign a character state to the first fork of the tree i e the internal node immediately descending from the root corresponding to the initial branching event that permits the most parsimonious reconstruction subject to the constraints imposed by the character type Consider for example an irreversible character with polarity up If for whatever reason the state for all terminal taxa included in the analysis is 1 then the character will contribute no length to the tree as expected for a constant character A below But if the ancestral state had defaulted to 0 rather than missing a step would have been assigned to the basal branch B below PAUP 3 1 USER S MANUAL
172. e selected text without altering the contents of the Clipboard unlike Cut which deletes the selected text but places it on the clipboard 234 PAUP 3 1 USER S MANUAL Select All This will select all text in the open window You can then perform the above actions on all selected text Clear Display Buffer Use this command to clear all output in PAUP s display buffer from memory allowing you to start with a fresh display window Edit Display Buffer This command causes the current contents of the display buffer to be placed into a new untitled editor window for editing or printing For example you might want to save only a portion of the output currently stored in the display buffer for subsequent retrieval You can do this by choosing the Edit Display Buffer command deleting the portions of the output you do not want to save and then requesting the Save command File menu to save the remaining output The current semigraphics mode is respected see the Semigraphics command of the Options menu If you have disabled the translation of semigraphics characters to standard ASCII characters you will not be able to print these characters in any font other than the special Monaco font included in PAUP Set Tabs and Font This allows you to choose the tab stops and font that PAUP will use while editing Find This option allows you to find a specified text string within the document and to specify other text with whic
173. e tree descriptions so long as you either retain the original translation table or use taxon names rather than numbers in the tree descriptions There are two advantages to using a translation table rather than using the taxon names directly in tree descriptions First tree descriptions are generally much more compact when integers rather than taxon labels are used especially if a large number of trees are being described Second if you decide to rename a taxon only one element of the translation table need be changed Without a translation table the old taxon name would have to be changed to the new one in every tree description TREES block processing is much faster 1f you use the consecutive integers 1 2 NTAX for the arbitrary tokens in the translation table Unless you have a good reason for not doing so you should follow this convention especially if a large number of trees are being defined A sample TREES block for four taxa named one two three and four 1s shown below begin trees translate 1 one 2 two 3 three 4 four tree a one two tree b one two tree c a o tree d 1 74 endblock three four three four r uses translation table If one or more taxa are omitted from a tree specification they are joined to the root node of the subtree described by that specification It is recommended that taxa be excluded from tree description only for 80 PAUP 3 1 USER S MAN
174. e two interpretations at the same time 1 e some multistate taxa interpreted as uncertainty others as polymorphism This current limitation will hopefully be eliminated in a future version IMPORTANT Polymorphism refers only to variability within a terminal taxon Multistate taxa do not provide a mechanism for dealing with characters that are polymorphic in a population genetic sense In particular there is no provision for polymorphism in hypothetical ancestral taxa internal nodes Ifa taxon is coded as having multiple states it is assumed that this taxon represents a monophyletic collection of subtaxa each of which are themselves monomorphic The program then assumes that the ancestor of this monophyletic group possessed one of the observed states from which the other states were subsequently derived Thus for example if two sister taxa were both coded as being polymorphic for states 0 and 1 the ancestor of this pair of taxa would be assigned either 0 or 1 in order to minimize the overall number of changes on the tree and either the 0 s or the 1 s would be interpreted as parallelisms While it might seem reasonable to assign both states to the ancestor with the polymorphism being retained in the two descendant taxa this would require additional assumptions about the relative probabilities of retention of polymorphism vs character transformation that are beyond PAUP s scope To assign multiple states to a taxon enclose all of
175. ead you could use a command such as wts 100 all scale basewt 100 3 9 11 14 See Chapter 4 for information on the Set Character Weights menu command PAUP 3 1 USER S MANUAL Excluding Characters Excluding characters is actually a specialized version of weighting characters excluded characters are simply assigned a weight of zero Excluded characters therefore do not contribute to overall tree lengths However they are still used in other contexts for example 1t possible to examine character changes in excluded characters with the DESCRIBE and CHGPLOT commands An increasingly common approach is to draw inferences about the evolution of one set of characters based on a phylogeny computed from an independent second set of characters In PAUP this can be accomplished simply by excluding the first set of characters searching for trees based on the second set and then using the standard facilities for interpretation of character changes in the first set The EXCLUDE command is used to specify a list of characters to be excluded For example the command exclude 1 25 requests exclusions of the first 25 characters If you want to restore previously excluded characters to the analysis you can use the INCLUDE command For example include 5 15 restores ten characters to their original nonexcluded status Ordinarily the effect of the EXCLUDE command is cumulative i e any characters already excluded by prior EXCLUDE commands r
176. ed This is separated into two somewhat less restrictive assumptions by Sanderson 1989 namely that characters are independent and that the observed character set is a representative sample of the universe of characters It is especially important to remember that if the sample of characters does not accurately reflect the larger underlying distribution of characters then the bootstrap confidence intervals may be very poorly estimated Also the bootstrap cannot overcome systematic biases such as long branch effects In any event it is wise not to take the bootstrap confidence values as absolutes there are many factors that might lead to over or underestimates of confidence PAUP 3 1 USER S MANUAL 63 LAKE S METHOD OF INVARIANTS PAUP provides an implementation of Lake s 1987b method of linear phylogenetic invariants or evolutionary parsimony for RNA or DNA sequence data In brief it is an attempt to estimate phylogeny accurately in the presence of highly unequal rates of change in different lineages When substitution rates are high in some branches and low in others parsimony may lead to an incorrect tree grouping the most substituted branches together even if they are not in reality closest relatives also see Felsenstein 1978 Lake s method is an attempt to circumvent the problems of unequal rates of substitution in four distantly related taxa It examines each of the three distinct unrooted trees labeled the E tree F t
177. ed in the Lake s Invariants dialog box or by using the LAKE command If no options are specified the command only calculates the invariants lake mode choose4 taxa 1 2 3 4 with the following output for the sample Hominoid mtDNA data Note that this data set is used for example purposes only the number of positions considered informative by Lake s method is too small for meaningful application of the method to this data set Lake s method of Phylogenetic Invariants Evolutionary Parsi mony E human chimp gorilla orang F human gorilla chimp orang G human orang chimp gorilla Invariants E tree x 0 0 0 3 0 25000 binomial test F tree y 0 0 1 1 1 00000 binomial test 3 0 0 0 0 0 0 noo noo G tree z These results would suggest that there is slightly more support for a human chimp clade the E tree than for human gorilla or chimp gorilla but that the result is not statistically significant Note that a binomial test is used rather than a chi square see Holmquist et al 1988 due to the low expected frequencies The following command will begin a search for a single quartet and output the spectral distribution and branch lengths lake mode choose4 taxa 1 2 3 4 spectdist brlength with the output 116 PAUP 3 1 USER S MANUAL Lake s method of Phylogenetic Invariants Evolutionary Parsimony Output of spectral distributions reque
178. ees in the file that are also currently in memory this will replace trees in memory with those trees in the treefile which are the same as those in memory MODE 2 replace all trees in memory by all trees in the file this is the default MODE 3 keep trees in memory that are not also in the treefile this excludes trees in memory that are also found in the treefile MODE 4 144 PAUP 3 1 USER S MANUAL keep trees that are currently in memory or the file but not in both places this excludes trees which are shared between memory and the treefile MODE 5 add trees in the file to the trees in memory all trees in the treefile are added without duplication MODE 7 PAUP will echo the Boolean selection and display a summary of how many trees in memory and or the file have been kept For example adding all trees in the file to the trees currently in memory gives this output Processing TREES block from file manual trees Keeping trees in memory plus trees in file without duplication 6 trees originally in memory 11 trees read from file 5 trees from file kept 11 trees now in memory A GETTREES command specifies either the range of trees to be imported or the boolean MODE Setting the MODE to different values invokes the different boolean operations described above IMPORTANT Getting trees using MODE 2 can be dangerous If no trees in memory match trees in file all trees in memory will be lost You may want to save trees
179. eful features and shortcuts available in PAUP More importantly you may be unaware of assumptions that the program is making during its calculations We cannot emphasize strongly enough the importance of reading the User s Manual carefully painful as this may be PAUP 3 1 USER S MANUAL TECHNICAL SUPPORT Assistance with the use of PAUP and interpretation of output will be provided only to licensed users If you are unable to resolve a problem by experimentation or need information that is not available in the User s Manual you may reach me in any of the following ways E mail Standard Mail Express courier FAX response paup onyx si edu Internet David L Swofford Laboratory of Molecular Systematics MRC 534 MSC Smithsonian Institution Washington DC 20560 USA David L Swofford Laboratory of Molecular Systematics Smithsonian Institution Museum Support Center 410 Silver Hill Road Silver Hill Road Suitland Maryland 20746 301 238 3059 Note that I cannot guarantee a FAXed Use of e mail is vastly more likely to generate a quick response than the other methods Please provide the following information in your communication 1 2 The exact wording of any error messages that you have received If a crash occurred the sequence of events prior to the crash e g commands issued etc The exact version number of your program You can obtain the version number from the opening screen that is d
180. eie cated Gaietenss 219 A O 219 A ET 219 IN epitsenter as E AE N e 219 Replace aare oo ae o E eS 220 Replace Mi a ina 220 The Windows M t ee as Se a hate Ro A G 220 Main Displays eninin nunn dd ira 220 Show Command Lie di 220 Show Memory Status sesseesseeeeseeesseesseesseesseeessressresse 221 Search Static dicta aida 221 LO O 221 Clean A cesses ga hcy Sect asta Get ges ast E dee Eas 221 A nn a S 222 Editor WiNOWS cisma daa iris 222 Nhe O prions Mt di 222 Multas Ta AO e o 222 OPIO E a aS E E taecnseaseeeaesies 223 Set Maktree Sern ii E a eas a aa ias 223 Character Matrix Format a 224 Searchin gd 225 ROGUE eere E E E N RE 225 Tree Onder iia aii 226 A ceS enan A ee eee ees 226 Tanjore Characters en e a fue iaeee iS 227 aTe O A T ag 227 Editke TN 228 Warnings amp ErrOTS sin tii 228 NEXUS Poda 229 Start p Preferences eo 230 Restore Option Settings ooooooccccoccccnoncnononcnonanaconanacinnnnos 230 The Data Mentee A oe ances ase nove 231 Include Exclude Caracter orion 231 Set Character Type AAA r a a a a Li 232 Set Character Weightss ciscc jssccsasacanesdecdgacesusrscsvensscetenedes 233 Rewelent Characters oi na A TE 234 Delete Restore Dad ns 234 Defme Q tgto p to iia 235 Show Character Status cocinas 235 Show Taxon SAS ra a i ead aera 236 Show UsertypeS A A e a 236 Show Data MA a a Rain 236 Show Distance Matrix cci cccsiessesdcctesesecassasdederesessaceonsndes 236 show Ancestral a 236 Choose Assum
181. eights recorded to 2 decimals Reweight Characters This provides the option of a posteriori weighting of characters You must first calculate some trees using equal weights Then reweight the characters depending on consistency index retention index or rescaled consistency index If there is more than one tree you can choose to weight by the best worst or average fit of the character over all trees You can also specify whether weights are truncated or rounded Hennig 86 truncates weights and if you wanted to replicate a Hennig86 search you would have the weights truncated This means that a character with a CI of 87 with a base weight of 10 would receive a weight of 8 instead of 9 the value obtained by rounding Reweight characters according to 8 Rescaled consistency index RC Consistency index C1 Retention index RI If more than one tree use Maximum value best fit Minimum value poorest fit Mean value Base weight E Truncate rather than round like Hennig86 1 5 ance Ex The Reweight Characters dialog box You also have the option of specifying the base weight for reweighting After assigning the new weights a new search is run The cycle continues until either the weights do not change over two successive runs or the same set of trees is obtained in two successive runs Delete Restore Taxa Use this command to temporarily delete taxa from the analysis or to rest
182. els amnion appendages thermoreg nostrils teeth assigns character names to the first five characters You can then use these character names in any context in which a character number would be used PAUP will also use these names to identify characters in the output Ordinarily the CHARLABELS command will supply a name for each character in the data set If you provide fewer than NCHAR names a 12 PAUP 3 1 USER S MANUAL warning will be issued a mistake may have been made in entering the names only the character numbers will then be available for the missing elements You may use a single underscore _ as a placeholder if you want to name some characters but use only numbers for others For instance if you wanted to name only the second and fourth characters in the above example you could use the command charlabels _ appendages _ nostrils _ The remaining characters could then be identified only by number 1 3 and 5 You should use a CHARLABELS command only if the TRANSPOSE option is not in effect If you enter the data in transposed format the character names will be obtained from the data matrix itself See Identifiers in Chapter 3 for rules pertaining to character names Predefined formats for molecular sequence data The predefined formats RNA DNA and PROTEIN are available for nucleotide and amino acid sequence data These are selected using the DATATYPE option For example begin data dimens
183. emain excluded when a new EXCLUDE command is issued If you want only the characters specified a single EXCLUDE command to be excluded you can use the ONLY option exclude 6 8 11 only This is equivalent to the pair of commands include all exclude 6 8 11 You can also use the Include Exclude Characters menu command to exclude or re include characters see Chapter 4 Successive weighting In addition to a priori weighting PAUP allows a posteriori weighting based on the fit of the characters to the trees currently in memory see A Posteriori Character Weighting in Chapter 1 The simple command PAUP 3 1 USER S MANUAL 95 reweight corresponds to the successive weighting method used in Hennig86 Farris 1988 The weight assigned to each character is proportional to the maximum rescaled consistency index over all trees in memory The weights are scaled between O and the base weight which is initially set to 1000 The BASEWT option allows you to specify a different scaling For example if you issue the command reweight basewt 10 weights will be scaled between 0 and 10 You can request that other fit measures be used via the INDEX option Possible values for INDEX are RC rescaled consistency index the default CI consistency index and RI retention index You can also use the FIT option to request that the minimum worst fit or mean fit values be used rather than the maximum best fit Available choices for FIT are
184. emember that character state trees like linearly ordered characters are undirected The character state tree only specifies available paths for character transformation it does not imply anything about polarity In the example above states B C or D could just as easily have been placed at the base as shown below PAUP 3 1 USER S MANUAL AY OM Q D Q B B 8 A C D B A D B C A C B D Three equivalent rootings of a character state tree See The ASSUMPTIONS block for details on the specification of character state trees in an input file Stepmatrices A stepmatrix is a square matrix specifying the distance from every character state to every other state These distances represent the cost in tree length units of the corresponding transformations Transformations may be completely forbidden by coding 1 for infinity as the transformation cost For each stepmatrix character PAUP uses dynamic programming algorithms Sankoff and Rousseau 1975 Sankoff and Cedergren 1983 to determine the minimum possible length on each tree it evaluates and to reconstruct hypothetical ancestors consistent with this length Stepmatrices may be either symmetric or asymmetric corresponding to the usual definition of matrix symmetry i e a matrix d is symmetric if dj dj for all ij Symmetric stepmatrices imply free reversibility of characters because the cost of reversing a given transformation is equal to the cost of t
185. ence i Re This output is very useful if there is a very large number of trees in memory If you are interested in all trees in which a particular group appears it would be tedious to output all trees to find the ones of interest A better strategy might be to compute a majority rule consensus with partition frequency table output identify groupings of interest construct a constraint tree that incorporates those groupings and finally filter the trees in memory to output only those trees on which the groupings occur Filtering Trees Trees in memory can be selectively hidden by a process known as filtering The analogy is to a physical filter which has certain characteristics that cause some elements to be retained leaving a residue on the filter itself PAUP s filters are of this type with the residue being the trees that meet the filtering criteria You build the filter by specifying its components for example so that it will keep trees shorter or longer than a given length within a certain topological distance from a reference tree or trees compatible with a constraint tree Unlike a physical filter though PAUP s filters can be inverted that is the residue will be trees that do PAUP 3 1 USER S MANUAL 147 not meet the filter criteria Filtering criteria are not cumulative each time you filter trees the filtering criteria are applied to all undeleted trees in memory Filtering is invoked by the command FILTER or th
186. ents are only locally optimal given the state assignments from the previous iteration Thus a taxon might be connected to a branch to which it would not connect if a full optimization of the tree were performed Branch swapping Because of the excessive greediness and susceptibility to local optima problems stepwise addition algorithms generally do not find optimal trees unless the data are very clean However it may be possible to improve the initial estimate by performing sets of predefined rearrangements a technique commonly referred to as branch swapping In general any one of these rearrangements amounts to a stab in the dark but if a better tree exists and enough rearrangements are tried one of them is likely to find it 40 PAUP 3 1 USER S MANUAL PAUP uses three branch swapping algorithms In order of increasing effectiveness these are 1 nearest neighbor interchanges NNI equivalent to the local procedure used in Versions 1 and 2 of PAUP 2 subtree pruning regrafting SPR approximately but not exactly equivalent to the global procedure used in earlier versions of PAUP and 3 tree bisection reconnection TBR a new procedure In NNI swapping each internal branch of the tree defines a local region of four subtrees connected by the internal branch Interchanging a subtree on one side of the branch with one from the other constitutes an NNI Two such rearrangements are possible for each internal branch as sho
187. epmatrix a b a 1 b 1 ES 3 I BRO If a stepmatrix contains a triplet that violates the triangle inequality PAUP will display a warning the stepmatrix is internally inconsistent but will still allow it to be used in an analysis In that case you must decide if you really want to disallow shortcuts such as bc and what the biological meaning of that stepmatrix really is The implications of having stepmatrices which violate the triangle inequality are discussed in more detail in Maddison and Maddison 1992 TREE LENGTHS AND CHARACTER WEIGHTS Parsimony analysis operates under the assumption that characters are independent of each other so that the length L of a full tree can be computed as the weighted sums of the lengths of the individual characters C L Yi gt jel 20 PAUP 3 1 USER S MANUAL where C is the total number of characters Consequently for the purpose of computing tree lengths we can treat each character in isolation from the rest It is this property that allows the mixing and matching of character types For example nothing prevents you from assuming that one subset of the characters is ordered Wagner another is unordered and yet another Dollo For any tree the minimum length required by each character is evaluated under the assumptions assigned to that character and the full length of the tree is obtained by summing over characters Typically wj 1 for all characters
188. equence of one or more character identifiers names and or numbers If two character identifiers are separated by a hyphen this indicates that the range of characters between the first character and the second character inclusive are to be included in the list The second character must have a higher number than the first character If the second specification in the range is followed by a backslash Y then an integer value immediately following the backslash represents an increment For example the list 3 24 3 consists of the characters 3 6 9 21 24 If a range plus increment is used only those characters contained within the range are included in the list e g the list 2 5 2 includes only characters 2 and 4 The special identifier consisting only of an unquoted period refers to the last character in the data matrix F NCHAR A character list composed of the reserved name ALL specifies all of the characters in the data matrix Examples of valid character lists assuming that corresponding character identifiers have been defined in the DATA block are as follows 8 amnion appendages gizzard teeth 1 two 5 7 TIS 1 37 16 31 28 LLSCALES CPDSCALES IENS all Character states A character state is a single digit alphabetic character or other symbol that represents a valid character state as defined by the SYMBOLS list 170 PAUP 3 1 USER S MANUAL Tree lists A tree list is simply a list of tree num
189. er in this chapter for details The characters specified by character list are assigned the immediately preceding character weight Any number of character weight character list pairs separated by commas may be specified WTS The WTS command is a synonym for the WEIGHTS command For some reason I have a hard time typing weights WTSET The WTSET command used to define a weight set is ordinarily issued from within the ASSUMPTIONS block However you may also issue it from the command line or from within a PAUP block See Commands used in the ASSUMPTIONS Block for the description of this command PAUP 3 1 USER S MANUAL 221 Chapter THE MACINTOSH INTERFACE This chapter describes aspects of running PAUP that are specific to the Apple Macintosh PAUP has a standard Macintosh interface including pulldown and pop up menus dialog boxes and scrollable lists This chapter assumes that you are familiar with these basic interface elements It also assumes that you already know how to use your Macintosh computer If this is not the case you should read or review the documentation that came with your computer PAUP may be started by double clicking on the PAUP application or document icon or under System 7 by dragging a PAUP or MacClade document icon onto the PAUP application icon INSTALLATION PAUP uses no special installation procedure Just copy the application and help files from the distribution diskette to your hard disk
190. escription outlined in the section Defining Your Own Character Types Chapter 2 WTSET Use the WTSET command to define a weight set The syntax is WTSET weight set name character weight character list character weight character list or WTSET weight set name VECTOR wt wt Wtz W NCHAR Character weight must be a nonnegative integer value In the first form any number of character weight character list pairs separated by 178 PAUP 3 1 USER S MANUAL commas may be specified In the second form a single weight is provided for every character m See Simplifying Input with Sets use welght sets in Chapter 2 for information on how to COMMANDS USED IN THE TREES BLOCK TRANSLATE Use the TRANSLATE command to define mappings of arbitrary tokens appearing in TREE and UTREE commands to valid taxon names Ordinarily the tokens are the integers 1 through NTAX TRANSLATE token taxon name token taxon name See The TREES Block in Chapter 2 for more information on the use of translation tables TREE UTREE Use the TREE and UTREE commands to define user specified rooted and unrooted tree topologies respectively The syntax is UJTREE tree name tree description You will usually define unrooted trees These trees are then rooted for output purposes using the outgroup or Lundberg rooting procedures However if you wish to define an ancestor for the full tree se
191. estral States in Chapter 2 No options are available SHOWCONSTR Use the SHOWCONSTR command to show one or more constraint tree definitions The syntax is SHOWCONSTR constraint tree name ALL If you do not specify a name or ALL the current default constraint tree will be shown Specify ALL to show all constraint trees that have been defined SHOWDIST Use the SHOWDIST command to output a matrix of distances between taxa see Chapter 2 There are no options SHOWMATRIX Use the SHOWMATRIX command to list the current data matrix The syntax is SHOWMATRIX options The options available are listed under Options for character matrix listings earlier in this chapter SHOWTREES Use the SHOWTREES command to request a diagram of one or more trees with no other information see also the DESCRIBE command The syntax is 218 PAUP 3 1 USER S MANUAL SHOWTREES tree list options TCOMPRESS This option is described in the section Options Affecting Multiple Commands earlier in this chapter ROOT OUTGROUP LUNDBERG MIDPOINT OUTROOT POLYTOMY PARAPHYL MONOPHYL ANCSTATES ancestral states name See Options affecting multiple commands above ANCSTATES is ignored unless LUNDBERG rooting is in effect SHOWUSER Use the SHOWUSER command to show all user defined character types see Verifying USERTYPE definitions in Chapter 2 for examples There are no options
192. etention index r is defined as g s g m Thus when a character fits the tree as poorly as possible its retention index will be 0 Note that for uninformative e g autapomorphic characters m g so that r is undefined Farris 1989a 1989b recommends using r as a factor for scaling c between 0 and 1 defining the rescaled consistency index as the product of r and c rc An ensemble overall consistency index C Farris 1989b for a suite of characters is calculated as M S where M and S are the sums over all characters in the suite of the individual character m and s values respectively The ensemble retention index R is defined analogously to the ensemble consistency index i e G SAG M Lg LsVXig Lm Archie 1989 independently proposed an equivalent index calling it the homoplasy excess ratio maximum HERM The product of R and C is referred to as the ensemble rescaled consistency index PAUP 3 1 USER S MANUAL 61 In general the homoplasy index HI is equal to 1 C But when multistate taxa are treated as polymorphic the homoplasy index has a slightly different meaning In that instance PAUP identifies an ancestral character state in the multistate taxon from which all observed states in that taxon must be derived see Multistate Taxa Thus the homoplasy index will not be the same for a character in which multistate taxa are treated as polymorphic character change allowed within t
193. ew running status report Defaults Hold tree s at each step The Stepwise Addition panel of the Heuristic dialog box The final panel has options for branch swapping You can choose the swapping option you want tree bisection reconnection subtree pruning regrafting or nearest neighbor interchange If you wish you can also select no swapping performing the stepwise addition phase only 256 PAUP 3 1 USER S MANUAL Heuristic Search Branch Swapping Options Options p Swapping algorithm tree bisection reconnection TBR General O subtree pruning regrafting SPR i O nearest neighbor interchanges NNI O Starting trees Ono swapping Stepwise addition f amp Save all minimal trees MULPARS Branch swapping O Steepest descent search H E Save no more than CO 2 length l each rep Defaults i a ee ey eh i o Abort current replicate if this limit hit under random addition sequence The Branch Swapping panel of the Heuristic dialog box PAUP will save all minimal trees if MULPARS is selected Selecting Save no more than ___ trees length ____ is handy when searches get bogged down finding huge numbers of trees that are far from optimal especially when used in conjunction with random addition sequence You can also choose to have PAUP abort the current replicate if the Save limit is hit during random addition Steepest descent will not abandon a
194. f special characters normally included with Macintosh fonts Thus although the trees look nice when drawn in the main display window they may not look right when printed on some printers Therefore PAUP ordinarily translates these semigraphics characters to standard ASCII substitutes when output is directed to 214 PAUP 3 1 USER S MANUAL a printer file or document window If you want to override this behavior specify SEMIGRAPH E g many IBM PC printers can print the high ASCII characters and the Apple LaserWriter can create a bit map version of PAUP s internal PAUPMonaco font This option is relevant only for IBM PC and MACINTOSH versions only The following three options affect the setting of the maximum number of trees that PAUP can store at any given time MAXTREES n The MAXTREES parameter specifies the maximum number of trees that can be saved Setting MAXTREES to a large value will reduce the likelihood that the tree buffer will become full during a search or tree file operation at the expense of a larger chunk of memory being tied up and therefore unavailable for other purposes Ordinarily if the number of trees found during a search reaches the value of MAXTREES you will be given a chance to increase MAXTREES before proceeding This behavior can be altered using the INCREASE option see below MAXTREES is initially set to 100 INCREASE PROMPT AUTO NO The setting of the INCREASE option deter
195. f the above paragraph can be circumvented by breaking the multistate character into a set of additive binary characters but this is not the case If you break a ten state character into nine binary characters it is indeed true that each of the resulting binary characters has the same weight as does a truly binary e g presence absence character but there are now nine of them In fact for a linearly ordered multistate character it makes no difference to the analysis whether you treat it as a single character or break it into a set of additive binary characters Doing the latter serves no useful purpose and merely complicates the interpretation of the output Note that most implementations of PAUP 3 0 do not allow fractional or decimal character weights Thus to assign a weight of 1 2 to a character you must instead assign a weight of two to all of the other characters The reason for this design decision is that integer arithmetic is vastly faster on microcomputers than is floating point or real arithmetic PAUP 3 1 USER S MANUAL 21 When scaling weights it is desirable to choose a smallest common multiple of the desired weights in order to avoid roundoff error For example if your data set contains a mixture of two three four and five state characters and you want to scale weights use 60 as the base weight so that the weights 1 2 1 3 1 4 and 1 5 correspond to 30 20 15 and 12 respectively PAUP also implements the su
196. f the characters Then if the site occurs in both the ingroup and the outgroup a site gain will be forced along the basal branch of the tree terminating at the common ancestor of the ingroup plus outgroup and all subsequent character changes will be losses If on the other hand the site occurs only within the ingroup then a single origination will be assigned within the ingroup perhaps with one or more subsequent losses Fortunately the unrooted Dollo approach available in PAUP and MacClade renders these more complicated approaches unnecessary but users should understand the logical connections between the alternative methodologies 4An all missing ancestor with polarity up see Direction of transformations polarity for all characters could also be used with equivalent results The same character state would be assigned to the node corresponding to the common ancestor of the ingroup plus outgroup in either case The only difference is that with an all absent PAUP 3 1 USER S MANUAL 13 NOTE Felsenstein 1984 has described a variation on rooted Dollo analysis that he calls the unordered Dollo method This terminology is a bit unfortunate unpolarized Dollo would have been a better name Unlike the Dollo method implemented in PAUP which assumes that the polarity is known the unpolarized Dollo method evaluates each character on a given tree under all possible polarity assignments and chooses that polarity which allows the m
197. frequency distribution of tree lengths Normally obtaining the frequency distribution is the only reason for doing an exhaustive rather than branch and bound search so you will probably always specify this option The frequency interval can be set using either the HINTERVAL option or the Class Interval selection in the Exhaustive search dialog box By default HINTERVAL 1 so that the number of trees at each length is output If HINTERVAL gt 1 then adjacent tree length values are pooled into tree length classes of width 114 PAUP 3 1 USER S MANUAL interval value An exhaustive search is begun by using the ALLTREES command or the Exhaustive menu command for example alltrees hinterval 5 will begin an exhaustive search and use an interval of five steps in the tree length histogram If you wish to save the interval distribution to a text file for input to other programs you must use the SAVEFD option or the Save tree length frequencies to file selection in the Exhaustive search dialog box The SAVEFD option is not persistent You must also specify a filename to receive the output either by using the FILE option or by specifying a filename in the dialog box that appears when the save option is selected For example the following command will begin an exhaustive search use an interval of five steps on the histogram save the distribution to the file test out replacing it without prompting if it already exists alltrees hinterval 5
198. g zapping it does not Note that you can also exclude characters ignoring and excluding are very different operations see Excluding Characters later in this chapter An uninformative character is defined as one which contributes exactly the same length to every possible tree topology or equivalently one whose minimum possible length is equal to its maximum possible length For example an unordered character is informative only if at least two character states each occur in more than one taxon otherwise the singleton states can always be explained as single changes on terminal branches regardless of the tree topology In general uninformative PAUP 3 1 USER S MANUAL 77 characters simply add a constant amount to the tree length and do not otherwise affect the results However the above definition of an uninformative character is somewhat ambiguous when missing data are present In this case uninformative characters can nonetheless provide potential support for certain groupings and can therefore affect whether zero length branches are collapsed see the sections Missing Data and Zero Length Branches and Polytomies Consequently unless you decline the option to collapse zero length branches tree searches may find different numbers of trees depending on the setting of the IGNORE option Invariant characters are those in which only one non missing state was observed Such characters contribute zero length to any possible
199. ge shares one vertex in common with the preceding edge in the sequence and the other vertex in common with the next edge in the sequence A path connecting a pair of vertices can also be described as a sequence of vertices with each vertex adjacent to the preceding vertex in the sequence A graph is connected if there is at least one path from any vertex to any other vertex A cycle is a path that connects a vertex to itself in which no vertex is repeated Most importantly for our purposes here a tree is a connected acyclic graph Vertices and edges on trees are often called nodes and branches respectively and this terminology will be used throughout the manual Nodes are called terminal nodes if they have degree one and internal nodes otherwise Unfortunately the above terminology is not standardized and much synonymy exists Terminal nodes are also called tips or leaves branches edges are also called links segments intervals or internodes Terminal nodes corresponding to biological taxa are often called Operational Taxonomic Units OTUs or simply terminal taxa Similarly internal nodes are sometimes referred to as Hypothetical Taxonomic Units HTUs or hypothetical ancestors A tree is binary if none of its internal nodes has degree exceeding three If a binary tree has at most one vertex of degree two it is also full Full binary trees are sometimes called strictly bifurcating or fully dichotomous PAUP 3 1 USER S MANUAL 5 trees
200. ght set mywts and the exclusion set noncoding all of which must previously have been defined in TYPESET WTSET and EXSET commands respectively 3 By using popup menus in the dialog boxes associated with Set Character Types Set Character Weights and Include Exclude Characters see Chapter 4 MULTISTATE TAXA Ordinarily you will assign a unique singleton character state to each taxon for each character However two situations may necessitate the assignment of multiple character states to a taxon 1 You may be uncertain about the state that a particular taxon possesses but some of the potential states can be excluded as possibilities This condition is called partial uncertainty 2 A terminal taxon may be an assemblage of lower level taxa which vary in the character state possessed This condition is loosely referred to as polymorphism When multiple states are interpreted as uncertainty PAUP will choose a state from the set of available states that allows minimization of the tree length When multiple states are treated as polymorphism PAUP assumes that the terminal taxon is actually a heterogeneous group In this case all but one of the states in the polymorphic terminal taxon must be derived from a monomorphic ancestral taxon in the most parsimonious PAUP 3 1 USER S MANUAL 103 way possible In PAUP you can choose either of the above interpretations of multistate taxa However it is not possible to mix th
201. go into effect when PAUP starts Display buffer memory Default starting mode when double click document icon in Finder Execute Edit EJ Remember window locations O Show memory status O Show command line Set startup defaults for other options to Current settings O Factory settings 8 Current startup defaults The Startup Preferences dialog box You may specify the amount of memory allocated for the display buffer in the range 32K to 128K The more memory that is allocated for the buffer the greater the number of lines can be retained for subsequent recall but the less memory that will be available for other purposes You may also set the default settings for most of the options in the program to either the values that are currently in effect or to the values that are in effect at the time PAUP is originally shipped Factory settings If the Save button is pressed a file called PAUP Preferences will be created or modified in the System Folder of the disk from which the computer was booted If a previous version of this file already exists 1t will be replaced If you instead press the Save As button a standard file dialog box will appear that allows you to create a file with any name and folder location you choose This new file can then be used to start PAUP by double clicking from the Finder with the chosen default settings In this way you can have a variety of default
202. h are output The following options are available SINGLE ALLI VARI NO The single option specifies the type of single character output If SINGLE ALL lengths or fit measures are output for all trees specified by tree list If SINGLE VAR lengths or fit measures are output only for those characters whose lengths vary among the trees specified by tree list If SINGLE NO no single character tree lengths or fit measures are output 206 LOG PAUP 3 1 USER S MANUAL NOJRANGE If RANGE is specified only the minimum and maximum tree lengths or best and worst fit measures are output for each character NO TOTAL If NOTOTAL is specified overall tree lengths and fit measures are not output NO TL Unless NOTE is specified tree lengths are output The default is to show tree lengths and no other fit measures NO CI Use CI to request output of consistency indices NO RI Use RI to request output of retention indices NO RC Use RC to request output of rescaled consistency indices NO HI Use HI to request output of homoplasy indices By default output generated by PAUP goes only to the display buffer a region of memory set aside exclusively for this purpose PAUP s main display window is used to view this information The LOG command may be used to request direction of PAUP output to a file e g for subsequent printing The syntax is LOG FILE file specification START STOP APPEND
203. h to replace it the actual replacing is done using the Replace command below Find Again This option repeats a previous Find command Replace This replaces highlighted text with the text specified in the Find command Use Find Again and Replace to cycle through all occurrences of the text specified in the Find command PAUP 3 1 USER S MANUAL 235 Replace All This option finds and replaces all instances of the text specified in the Find command THE WINDOWS MENU The Windows menu allows you to open available windows or bring them to the front if they are already open v Main Display 0 Show Command Line K Show Memory Status 3M Search Status PAUP Help Bf Zoom Clean Up Close All backbone 2 1 manual 2 2 Angiosperms 23 The Windows menu Main Display Choose Main Display to bring PAUP s main display window to the front The same effect may be achieved by simply clicking on the main display window but it may be entirely covered by other windows The width of the main display is the same each time it is called up You may shrink it by holding down the option key while clicking and dragging on the lower right hand corner It cannot be expanded beyond the standard size Show Command Line Choose Command Line to show the command line at the bottom of the main window The command line provides an alternative way to control PAUP s actions The available commands are identical to those found in generic m
204. h up to 31 characters may included in graphical tree output This is due to line length constraints on PAUP output Any TAXLABELS command issued from the command line overrides any labels provided in the matrix MacClade ignores TAXLABELS unless the data matrix is transposed COMMANDS USED IN THE ASSUMPTIONS BLOCK ANCSTATES Use the ANCSTATES command to define a set of ancestral states corresponding to an ancestral taxon The syntax is ANCSTATES ancestor name character state character list character state character list Any number of character state character list pairs separated by commas may be specified You may also repeat the ANCSTATES command to define multiple ancestors although only one ancestor can be in effect at any given time If you precede the ancestor name with an asterisk that ancestor becomes the current ancestor You can also assign the current ancestor with the ASSUME command or the Choose Assumption Sets menu command ANCSTATES can also be used to explicitly specify the state for each character in the ancestor This is done using the VECTOR format option e g if NCHAR 5 ANCSTATES name VECTOR 0 1 2 0 2 would define an ancestor name with the specified states for the five characters See Defining Ancestral States in Chapter 2 for information on how ancestral state specifications are used CHARSET Use the CHARSET command to define a character set Character sets are sim
205. haracters that are to be excluded from the analysis see Excluding Characters under Character Weighting above For example the command exset dontwant 4 11 20 31 34 specifies a list of 15 characters that can be excluded by invoking the exclusion set dontwant Remember that you can use character sets see above to refer to groups of characters in the EXSET command Simply specify a character set name anywhere a character name or number could be used Invoking assumption sets The TYPESET WTSET and EXSET commands merely define assumption sets they do not cause the specified types weights or exclusion status to go into effect An assumption set must be invoked before it actually takes effect There are three ways to invoke an assumption set 102 PAUP 3 1 USER S MANUAL 1 By preceding the type set weight set or exclusion set name with an asterisk For example if the following commands were issued in sequence typeset one ord 5 8 typeset two dollo all typeset thr irrev 1 5 dollo 6 three type sets would be defined and all characters would currently be assigned type DOLLO 2 By using the ASSUME command This command allows you to invoke any combination of type sets weight sets exclusion sets and ancestors For example assume typeset one wtset mywts exset noncoding sets the current character types weights and exclusion status to the settings defined in the type set one the wei
206. he correct punctuation If for instance you omitted the first comma in the above example the character type dollo would be interpreted as a character identifier and would generate an error message Remember that you can use character sets see above to refer to groups of characters in the TYPESET command Simply specify a character set name anywhere a character name or number could be used PAUP 3 1 USER S MANUAL 101 Weight sets WTSETs Weight sets assign a weight to each character For example you might define a weight set as follows wtset mywts 2 2 4 8 12 3 6 14 20 Any character for which a weight is not explicitly assigned receives a weight of one Be careful to use the correct punctuation If for instance you omitted the first comma in the above example the next 3 would be taken as a character number and assigned weight 2 However the immediately following colon would generate a syntax error message An alternate format vector allows you to specify the weight for each character sequentially rather than by character lists For example the following command is equivalent to the one shown above wtset mywts2 vector 12121312222213353 313 53 3 Remember that you can use character sets see above to refer to groups of characters in the WTSET command Simply specify a character set name anywhere a character name or number could be used Exclusion sets EXSETs Exclusion sets allow you to specify a set of c
207. he default character type is now unordered rather than ordered as in earlier versions Unless ordered characters are explicitly requested PAUP will assume that multistate characters are unordered Unordered Characters Unordered characters are defined such that any state is capable of transforming directly to any other state with equal cost For example a five state unordered character with states A through E has the character state graph PAUP 3 1 USER S MANUAL 9 Character state assignments are made to internal nodes of the tree so as to minimize the total number of character state transformations steps using an algorithm based on that of Fitch 1971 NOTE The ordered vs unordered concept does not pertain to binary characters The order of a character refers to the potential pathways of character transformation and there is only one possible path between the two states of a binary character Consequently it makes no difference whether freely reversible binary characters are defined as ordered or unordered Dollo Characters A Dollo character is one that is consistent with the requirement that every derived character state be uniquely derived If a hypothetical ancestor is included in the analysis this definition corresponds to the traditional Dollo model in which each character state is allowed to originate only once during evolution and all homoplasy takes the form of reversals to a more ancestral condition i e
208. he original transformation NOTE Asymmetric stepmatrices force a rooted tree If the stepmatrix is asymmetrical the taxon designated ancestor is very important If you want to have asymmetrical transition penalties but remain agnostic regarding the ancestral condition of the characters you should include an ancestor with all missing values See the section Character Types for the relationship between character type and tree rooting The generalized parsimony approach provided by stepmatrix characters is extremely powerful Researchers with nucleotide sequence data can assign different weights to transitions vs transversions and even to different kinds of transitions and transversions PAUP 3 1 USER S MANUAL 17 ACGT A 0515 C 50 51 G 1505 Ta 0 90 Character state graph and corresponding stepmatrix for a character type that assigns five times more weight to transversions than to transitions For restriction map data site gains can be given higher weight than site losses so that parallel loss and gain loss events are preferred over parallel gains and loss regains Templeton 1983b Templeton 1983a avoiding the perhaps overly severe strict prohibitions imposed by the Dollo and Camin Sokal models DeBry and Slade 1985 AR i QS E Character state graph and corresponding stepmatrix for a character type that assigns three times as much weight to a gain than a loss ro OW Stepmatrices may be used to define
209. he terminal and uncertain no character change allowed within the terminal All of these measures are useful not only in comparing characters on a single tree but on multiple trees as well If several different analyses of a data set produce different trees when different assumptions are made these indices are a quick way to estimate which characters support which hypotheses of topology When you use the LENFIT command or the Lengths and Fit Measures menu command you can output a summary of length and fit measures for trees and for characters These can then be used to evaluate which characters fit which trees You can choose to output information only for those characters which vary over the trees in memory see the section Lengths and Fit Measures in Chapter 2 A POSTERIORI CHARACTER WEIGHTING PAUP also allows you to assign weights to characters based on their fit to trees in memory These a posteriori weights are then used as input for another successive analysis You can continue to reweight and reanalyze until the weights do not change for two consecutive analyses or until identical trees or sets of trees are found in two consecutive searches By doing this you may or may not converge on fewer tree topologies depending on the data matrix If you choose this route however tree lengths will no longer be comparable between successive searches as they are directly dependent on the weights which are in place The basic idea is t
210. he tree file If file specification contains any of the characters equal sign semicolon colon or blank it must be enclosed within single quotes REPLACE APPEND Ordinarily if SAVE is requested and the specified file see FILE option above already exists you will be prompted for confirmation that the existing file should be replaced Explicit specification of REPLACE suppresses this prompt the existing file will be quietly overwritten by the new data Alternatively you may specify APPEND in which case a new TREES block will be concatenated to the end of an existing file The APPEND option is mainly useful only for archival purposes as current versions of PAUP and MacClade can process only the first TREES block in a file For obvious reasons this option is not persistent ROOT OUTGROUP LUNDBERG MIDPOINT OUTROOT POLYTOMY PARAPHYL MONOPHYL NO TCOMPRESS See Options Affecting Multiple Commands earlier in this chapter PAUP 3 1 USER S MANUAL 191 CSTATUS Use the CSTATUS command to request a listing of character status information for all characters There are no options For each character the following information is output 1 The number and name if any of the character 2 If the character is constant invariant excluded or uninformative this is indicated 3 The character s current type and weight 4 A list of the states observed for the character CTYPE Use the
211. hed at each of three possible places The shortest tree resulting from the addition of taxon D is retained and taxon E is added resulting in five trees Branch swapping can then begin on the shortest tree s from these five BG His Stepwise addition of five taxa Of course the above description is oversimplified We must have some way of determining which three taxa will be joined initially and which one of the unplaced taxa will be connected to the tree at each step PAUP provides four options for specifying the addition sequence 1 Asis The taxa are simply added in the same order in which they are presented in the data matrix starting with the first three and sequentially adding the rest This method is usually not very effective 2 Closest Initially the lengths of all possible three taxon trees formed by joining a triplet of terminal taxa to a single internal node are evaluated The three taxa yielding compose the PAUP 3 1 USER S MANUAL 37 starting tree At each successive step all remaining unplaced taxa are considered for connection to every branch of the tree and the taxon branch combination that requires the smallest increase in tree length is chosen Obviously the closest approach requires considerably more computation than does as is In the latter the number of tree lengths that must be evaluated at each stage is simply equal to the number of branches on the tree i e one for each potential connection
212. hich taxa have been assigned to the outgroup No options are available TYPESET The TYPESET command used to define a type set is ordinarily issued from within the ASSUMPTIONS block However you may also issue it from the command line or from within a PAUP block See Commands used in the ASSUMPTIONS Block for the description of this command USERTREE Use the USERTREE command to input a single user defined tree The syntax is USERTREE tree specification Ordinarily you should use a TREES block to input one or more user defined trees This command merely provides a mechanism for quickly specifying a user defined tree from the command line which may be useful in certain situations 220 PAUP 3 1 USER S MANUAL USERTYPE The USERTYPE command used to define a user defined character type 1s ordinarily issued from within the ASSUMPTIONS block However you may also issue it from the command line or from within a PAUP block See Commands used in the ASSUMPTIONS Block for the description of this command WEIGHTS Use the WEIGHTS command to assign weights to one or more characters The syntax is WEIGHTS character weight character list character weights character list The character weight must be a valid character weight see Tree Lengths and Character Weights in Chapter 1 Each character list consists of one or more character numbers character names or character set names see Character lists earli
213. hival purposes as current versions of PAUP and MacClade can process only the first TREES block in a file For obvious reasons this option is not persistent The SET command is used to set a variety of options whose scope extends beyond single commands The syntax is SET option specification The available options are as follows NO DISPLA Y NODISPLAY suppresses output to the main display window or terminal screen and is useful when you want to send output to the log file and or printer only DISPLAY reactivates the main display At least one output destination must be active at all times Consequently if no log file is active or the echo to printer ECHO option is not set output will be sent to the main display even if NODISPLAY has been requested MSTAXA UNCERTAIN POLYMORPH The MSTAXA option specifies how PAUP treats multistate taxa see Multistate Taxa in Chapter 2 This option is set at the time the data matrix is processed see description of the OPTIONS command under Commands used in the DATA Block earlier in this chapter You can change the interpretation of multistate taxa at any time by specifying MSTAXA in a SET command NO SEMIGRAPH PAUP uses special characters in its internal font to draw trees and other items On the IBM PC these characters are nonstandard high ASCII characters On the Macintosh these characters are neither in the standard 128 ASCII characters nor in the set o
214. hoal bass AMNION 23 x21 02 myType PAUP imposes limits on the lengths of identifiers as defined below If you use identifiers that exceed these limits PAUP simply truncates them to the maximum acceptable length If you use names that are longer than PAUP 3 1 USER S MANUAL 167 the maximum lengths e g for other NEXUS programs that allow longer identifiers you should make sure that identifiers in the same class will be unique after truncation by PAUP MacClade will not accept all digit character or taxon names PAUP allows them with a warning that taxon character names have precedence over numbers e g the if the tenth character is named 5 the command EXSET no_5 5 causes character number five to be excluded by MacClade and character number ten to be excluded by PAUP Do not use all digit names in PAUP if you intend to also use the data file in MacClade and it s not advisable to use them in PAUP in any case Taxon identifiers For input taxa may be referenced either by name or by number The numbers are simply the row number of the taxon in the input data matrix Note that if you delete taxa the original taxon numbers are still used to refer to taxa Taxon names may be up to 32 characters in length however for most output they are truncated to 16 characters Thus you should choose names that are unique up to at least the first 16 characters The full 32 character names are used only for high resolution tree plotting
215. how the starting trees will be obtained by stepwise addition or by using some or all of the trees already in memory You can also choose to swap on all the trees in memory including nonminimal trees or on minimal trees only by selecting the appropriate checkbox in the Swap on option Heuristic Search E Starting Trees for Branch Swapping Options i General i Get by stepwise addition Starting trees i E O Use tree C through C O Use all trees in memory n 5 O Stepwise addition O Branch swapping o All starting trees including Search nonminimal ones lane 8 Minimal trees only Cancel JRetain all trees initially in memory The Starting Trees panel of the Heuristic dialog box The third panel has options for stepwise addition the addition sequence can be simple closest asis or random You can also select how many trees are held at each step For random addition you may change the number of repetitions or the seed tree You can also select a running status report of the random search Using the random addition sequence option is the best way to find islands of trees that may not be discovered by other addition sequence options Heuristic Search Stepwise Addition Options Options E i p Addition sequence O General A simple reference taxon 2 i i O closest i Starting trees i o I Oasis Stepwise addition l i O random Branch swapping i reps Search J i O sk
216. icks and key presses Because event checking steals processor time from PAUP s computations disabling it will provide a speed improvement However you give up the ability to abort a search other than by restarting the computer or to activate a different program under MultiFinder PAUP 3 1 USER S MANUAL 241 Rooting The Rooting command allows you to specify whether unrooted trees will be rooted using the outgroup or the Lundberg methods Choose method for rooting unrooted trees Outgroup rooting o Root tree at internal node i i with basal polytomy Define Outgroup l E H Ta Make ingroup monophyletic i i If more than one outgroup taxon present i EE Make outgroup paraphyletic E 1 i 1 with respect to ingroup Ez Make outgroup a monophyletic 1 sister group to ingroup O Luadberg rooting Encsiale O Midpoint rooting The Rooting dialog box See the section in Chapter 2 for a description of the different rooting options If outgroup rooting is selected the outgroup may be defined from within this dialog box by clicking on the Define Outgroup button This will bring up the standard Define Outgroup dialog box For Lundberg rooting the outgroup ancestor is set according to the current ANCSTATES setting You can use the pop up menu to change the outgroup ancestor to be used Tree Order The order of the taxa in a tree drawn by PAUP has no physical significance four options are available
217. ict e and 50 majority rule f consensus trees Semistrict Consensus This method corresponds to the combinable component consensus of Bremer 1990 essentially the same idea was earlier proposed by Hillis 1987 Under this method if all trees have either an A B C trichotomy or an A B C dichotomy 1 e A B is never contradicted just not always supported then A B is retained in the consensus In the example above trees a and c are unresolved while tree b supports A B The semistrict consensus e preserves A B while the strict consensus f does not When there is conflict semistrict behaves the same as strict E Majority rule consensus In contrast to strict consensus it may be of interest to find groups that appear on a certain pre specified percentage of the rival trees Thus a group may be preserved in the consensus even if there are some trees that support conflicting groups see Margush and McMorris 1981 and Swofford 1991 In the example above the majority rule consensus d preserves EF because it is found in two of the three rival trees a and c Likewise ABC is unresolved because two of the three trees do not resolve it Typically the majority rule percentage is set to 50 so that the consensus will retain all groups found in over half of the rival trees The reason that groups must be present in over half the rival trees to be included is that two groups occurring in exactly half the trees might not be
218. ig86 or PHYLIP Once these have been imported into PAUP they can either be stored as NEXUS files or exported to the original or a different format Comparing Trees Once trees are in memory either as the result of a search or from the reading of a treefile they can be compared using the TREEDIST command or the Tree to Tree Distances menu command There are no options This calculates the symmetric difference or partition metric Penny and Hendy 1985 which is equivalent to the Robinson and Foulds 1981 contraction decontraction metric Trees that are most similar will have the lowest distance value The output will appear in a matrix of tree to tree distances For example Symmetric difference distances between trees LE 7 3d 5 OY 8 9 0 1d RA DH RW ee ROUSR RN RRR CONTR ON BRNO NRW HAN OD o co MON a Doo co o oo Nao w Calculating Consensus Trees If multiple trees are in memory you can calculate the consensus for those trees in four ways Adams strict semistrict and majority rule The details of each of the consensus methods are discussed in the section Consensus Trees You may calculate one or all of the different consensus types for any given set of rival trees you do not have to issue repeated requests for each type Consensus trees are obtained by using the CONTREE command or the Compute Consensus menu command Once a consensus tree is calculated it is not stored in the tree buffer If you wish to us
219. ile is invoked the default is to import all the trees in a particular treefile replacing all trees in memory For example gettrees file trees out will import all of the trees in the file trees out However you don t have to import all of the trees in the file you can also specify the tree numbers if you know which number corresponds to which topology For example gettrees file trees out from 1 to 5 will replace any trees in memory with trees one through five in the file trees out These trees may be the products of a previous search of the current data matrix they may be derived from a different data matrix for the same taxa or they may have been created by a program such as MacClade You can then test how certain starting trees affect the search and explore how the characters are reconstructed on different trees There is also a way to import trees without knowing which tree number corresponds to which topology PAUP has the option of performing Boolean operations on tree files using the MODE option or by making the appropriate selection in the Get Trees From File dialog box You can choose these operations with the following command gettrees file trees out mode n where n is an integer taken from the MODE value for one of the choices below keep trees which are in the file but not originally in memory this means you will import only trees which are different from those currently in memory MODE 1 keep tr
220. ill be presented notifying you that the process has completed Then a small PAUP icon will alternate with the Apple logo in the menu bar until PAUP is returned to the foreground Remember that each application running under MultiFinder reserves a certain amount of memory for its own use Once that memory is exhausted the application is not allowed to access other memory even though there may still be some free memory available Unlike some applications the amount of memory needed by PAUP is a function of the size of the data set Therefore in setting PAUP s default memory size I was forced into a compromise between requesting enough memory to handle larger data sets while not tying up more memory which is then unavailable to other applications than PAUP really needs for typical data sets If you get an out of memory message while running under MultiFinder you will have to increase the memory size as follows 1 Select the PAUP icon in the Finder and choose Get Info from the File menu 2 Change the value displayed in the Application Memory Size to a larger value 3 Click the close box on the information window If you run out of memory and you are not using MultiFinder you will either have to add additional memory to your machine or take additional steps to reduce the amount of memory used by the system e g remove unneeded INITs from the System Folder etc PAUP 3 1 USER S MANUAL 225 THE APPLE 6 MENU About PAUP
221. in the analysis anyway or 2 the character does not apply e g the character represents different modifications of a structure that is absent entirely in some taxa In either case a possible solution is to assign to the taxon the character state that would be most parsimonious given its placement on the tree but see Maddison In press Effectively then only those characters that have non missing values will affect the location of any taxon on the tree The example below illustrates this concept Remember that for the purpose of computing tree lengths each character can be treated independently in this example one character is considered in isolation from all of the others For tree A PAUP would assign state 1 to the taxon with missing data B This allows a reconstruction requiring only one step over the full tree For tree B however assignment of state 0 is required in order to minimize the length required by this character again one step Thus both trees are equally parsimonious with respect to this character if one 24 PAUP 3 1 USER S MANUAL tree is in fact better than the other evidence for this conclusion will have to come from other characters A B B DE F 1 gt 1 0 0 0 0 E a Y7 1 1 0 0 Mego 4B A C 1 0 ae 1 1 E 0 W7 0 0 1 NP 0 0 vee 0 0 0 0 G G A B Demonstration of PAUP s handling of missing data Reconstruction of one character on two different trees see text
222. in the data matrix for a particular character must be represented in any character state tree assigned to that character Additional states that were not observed may also be included in the character state tree as well The system used to describe character state trees is straightforward Think of the character state tree diagram as a set of roadways connecting the nodes of the tree see figure below The itinerary is to visit all of the nodes character states of the character state tree in a circuit beginning at the root node state D following two simple rules 1 when you come to 86 PAUP 3 1 USER S MANUAL an intersection or fork in the road internal node always bear to the left and 2 when you come to a dead end terminal node turn around The path indicated by the arrows in the figure shows the sequence in which the nodes would be visited in this example A character state tree and the circuit followed in writing its description To write the character state tree description perform the following Operations as you make the circuit e When you leave a node traveling away from the root toward the leftmost descendant write a left parenthesis e When you leave a node traveling away from the root toward any descendant other than the leftmost descendant write a comma e When you leave the rightmost descendant of an internal node traveling toward the root write a right parenthesis e When you visit a terminal node or
223. ing an input file when unrecognized commands OPTIONS command keywords or formats are encountered If NOERRORSTOP is specified a warning message is issued and processing is allowed to continue NOJWARNRESET Ordinarily PAUP issues a warning message when an input file containing a DATA block is executed and a DATA block has already been processed Specification of NOWARNRESET suppresses this warning TORDER STANDARD RIGHT LEFT ALPHABET Specifies the convention used to order the tree see Changing the Order of Taxa on Trees in Chapter 2 STEPMATRIX OBSONLY ALLSTATES THREEPLUS1 When a stepmatrix contains three or more character states it is possible that full minimization of the tree length may require assignments of character states that were not observed in any of the 216 PAUP 3 1 USER S MANUAL terminal taxa to internal nodes When only a few character states were observed for a stepmatrix character but the stepmatrix is defined for a large number of character states PAUP provides three options for dealing with this problem If STEPMATRIX OBSONLY only those character states observed in terminal taxa are candidates for assignment to internal nodes If STEPMATRIX ALLSTATES any state contained in the stepmatrix definition may be assigned to internal nodes regardless of whether it was observed in a terminal taxon If STEPMATRIX THREEPLUS1 PAUP provides a compromise between these two extremes Charac
224. inimum number of changes For example in a two state character with states 0 and 1 we would count first the number of changes required under the assumption that state O is ancestral and then count the number of changes required under the assumption that state 1 is ancestral Polarity is then assigned in accordance with the assignment requiring the fewest changes I cannot imagine a situation in which one would be willing to assume that a character could not evolve from an ancestral state to a derived state more than once but unwilling to postulate the character s polarity Consequently PAUP does not implement this method If you can imagine such a situation let me know The unpolarized Dollo approach is not appropriate for restriction site data since allowing site presence to be ancestral and site absence to be derived would under the Dollo model imply that sites could only be lost once and then regained as many times as necessary clearly an unreasonable assumption Polarity specification For the unrooted Dollo method PAUP allows either the lowest or highest observed state as defined by the SYMBOLS list to be the ancestral state polarity up vs down respectively If you do not explicitly specify up or down up is assumed When an ancestor is included in the analysis rooted Dollo any state in the SYMBOLS list may be designated as the ancestral state See Assigning Character Polarities under Specifying Character Types i
225. ions ntax 4 nchar 20 format datatype dna gap matrix One ATGCT ATCCG TCATG ACCTA Two ATCCT AGCCG T AG ACGGA Three CTGCT AGCCG TGGAG TCCTA Four CTGAA CG ACATA AGTCA 1 endblock If DATATYPE is set to DNA the SYMBOLS list is set to ACGT and the following equate macros corresponding to the IUPAC IUB ambiguity codes are predefined AG puRine CT pYrimidine AC aMino GT Keto CG Strong AT Weak ACT not G CGT not A ACG not T lt BISOARZAD PAUP 3 1 USER S MANUAL 73 D AGT not C N ACGT unkNown Also the symbol X is interpreted as missing data If DATATYPE RNA the SYMBOLS list is set to ACGU and the same set of equate macros are defined except that U is substituted for T If DATATYPE PROTEIN the symbols list is set to ACDEFGHIKLMNPQRSTVWY corresponding to the standard IUB single letter amino acid codes A ala alanine C Cys cysteine D asp aspartic acid E glu glutamic acid F phe phenylalanine G gly glycine H his histidine I ileu isoleucine K lys lysine L leu leucine M met methionine N asn asparagine P pro proline Q gin glutamine R ag arginine S ser serine T th threonine V val valine W trp tryptophan Y tyr tyrosine ES nonsense chain termination The symbol X for unk
226. is critical to the run times the branch and bound algorithm can be applied to larger data sets on a Cray X MP than on a Macintosh Plus The above presentation of the branch and bound method while correct is an oversimplification of the algorithms actually used in PAUP which implements algorithmic refinements that greatly speed the computations These refinements designed to promote earlier cut offs in the traversal of the search tree include 1 using heuristic methods see below to obtain a near optimal tree whose length is used as the initial upper bound 2 designing the search tree so that divergent taxa are added early thereby increasing the length of the initial trees in the search path and 3 using pairwise incompatibility to improve the lower bound on the length that will ultimately be required by trees descending from a tree at a given node of the search tree PAUP 3 1 USER S MANUAL 35 Since the branch and bound method requires evaluation of all trees as its worst possible case why would we ever want to perform an exhaustive search In fact if we were interested only in the optimal trees the branch and bound algorithm would indeed be the method of choice However exhaustive searches can be used to generate the frequency distribution of tree lengths We may find it useful to know for example whether there are few or many near optimal trees or where some tree of prior interest lies in the distribution of tree lengths
227. is is the default The reason that taxon A and not taxon B is the basal most taxon when the tree is rooted such that the outgroup is paraphyletic is that taxon A is the primary outgroup the first outgroup taxon listed by the OUTGROUP command or chosen using the Define Outgroup menu command The primary outgroup will always be the basal most taxon when outgroup rooting and a paraphyletic outgroup is chosen 30 PAUP 3 1 USER S MANUAL AB CD E ABCDE o OD Outgroup rooting options a unrooted tree for five taxa outgroup is A B b tree rooted with monophyletic outgroup c tree rooted with paraphyletic outgroup d tree rooted with basal polytomy You have three options when using outgroup rooting and more than one outgroup is present in the analysis These are rooting at an internal node with a basal polytomy rooting such that the ingroup is monophyletic and the outgroup is paraphyletic with respect to the ingroup or rooting such that the ingroup is monophyletic with the outgroup a monophyletic sister group These options allow you to at least partially constrain the structure of the outgroup topology when the tree is rooted providing the specified ingroup is monophyletic See the section Rooting trees for output and character reconstruction If however the tree cannot be rooted such that the ingroup is monophyletic PAUP will include as many outgroup taxa as necessary to make the ingroup plus those outgroup taxa monophyletic
228. is the first DESCRIBE command and you do not specify a tree list only the first tree is described The following options are available PLOT CLADOGRAM PHYLOGRAM BOTH If PLOT CLADOGRAM branch lengths on the tree have no meaning and taxa are aligned at the right edge of the diagram If PLOT PHYLOGRAM branch lengths are drawn proportionally to the number of changes assigned to each branch PLOT BOTH requests output of the tree diagram in both CLADOGRAM and PHYLOGRAM formats NO LINKS Requests output of a table of assigned minimum possible and maximum possible branch lengths see Table of Branch Lengths and Linkages in Chapter 2 NO CHGLIST Requests output of a list of changes in each character see Change and Apomorphy Lists in Chapter 2 194 PAUP 3 1 USER S MANUAL NO APOLIST List of changes along each branch see Change and Apomorphy Lists in Chapter 2 NO DIAG Requests output of the minimum possible assigned and maximum possible length of each character and goodness of fit measures based on these quantities see Consistency Indices and Goodness of Fit Statistics in Chapter 2 NO PATRISTIC Requests output of the patristic distance matrix see Pairwise Homoplasy and Patristic Distance Matrices in Chapter 2 NOJHOMOPLASY Requests output of the pairwise homoplasy matrix see Pairwise Homoplasy and Patristic Distance Matrices in Chapter 2 NO LABELNODE Ordinarily internal n
229. ise treated as ordered characters Formally a character state tree imposes a partial order on the character states In most earlier programs users were required to recode character state trees using additive coding binary or otherwise see below PAUP now provides this capability automatically PAUP 3 1 USER S MANUAL 15 As an example suppose that you wish to assume that states C and D were derived independently from state B which itself was derived from state A The following character state tree would be used to represent this relationship E D B Either of the following additive codings breaking the character into a suite of independent characters could be used in lieu of a character state tree A 00 A 000 B 10 B 100 C 20 C 110 D 11 D 101 However such a coding makes interpretation of the output more difficult because you would have to decode the new characters back to the original states of the character state tree In this case the character state tree specification C D B A provides an equivalent definition of the character requires essentially identical computation time and eliminates the coding decoding steps Characters defined as character state trees are fully reversible however they may evolve only along the paths specified by that definition For instance in the above example a change from state C to state D would imply passage through state B and would require two steps Finally r
230. isplayed when you first start the program A copy of your data file ANONYMOUS FTP SERVER FOR PAUP SUPPORT We have set up an FTP server to support PAUP We will periodically post updated programs test versions documentation files and other announcements there To use the FTP server log in to onyx si edu vi PAUP 3 1 User s Manual 160 111 64 54 as anonymous and enter your e mail address when prompted for a password The overall structure of the ftp directories is described in a README file in the root directory When minor bug fix releases are issued we will post an updater program that will convert any version of PAUP Mac 3 1 to the new version The updater program will only work if you already have a copy of PAUP 3 1 PAUP 3 1 USER S MANUAL Vii PAUP 3 1 USER S MANUAL 1 Chapter BACKGROUND This chapter provides general background on the concepts underlying the methods used in PAUP Specific information on how to use the program follows in later chapters GENERAL CONCEPTS PAUP is a program for inferring phylogenies from discrete character data under the principle of maximum parsimony Parsimony methods search for minimum length trees trees that minimize the amount of evolutionary change needed to explain the available data under a prespecified set of constraints upon permissible character changes The best known discrete character parsimony method often called Wagner parsimony Kluge and Farris 1969 Farris
231. ither does not recognize or does not wish to interpret A program can easily skip to the end of the command by looking for its terminating semicolon even if it is otherwise unable to interpret the command The precise behavior of the program upon encountering such commands is not specified by the standard PAUP provides the options of allowing execution to continue with a one line warning that an unrecognized command was encountered or terminating the processing of a file with an error message PAUP 3 1 USER S MANUAL 67 NEXUS file identification NEXUS conforming files are identified by a NEXUS directive at the very beginning of the file line 1 column 1 General format of NEXUS files NEXUS files are entirely free format Blanks tabs and newlines may be placed anywhere in the file Unless RESPECTCASE is requested in the FORMAT command commands and data may be entered in upper case lower case or a mixture of upper and lower case If RESPECTCASE is requested case is considered significant in character state symbols and in names for assumption sets ANCSTATES specifications etc This is only true for standard roman alphabetic characters For example MULLERI and mtilleri are not considered identical if RESPECTCASE is off Comments may be included in the input file by enclosing them in square brackets For example This is a comment Unless the first character following the is an explanation point the comment is fo
232. itial 30 step trees saving all of the new shortest trees and passing them to the next round It s called steepest descent because instead of discarding current trees the instant a shorter one is found it continues looking for even shorter ones and uses the tree s that give the most improvement for the next round Note that if STEEPEST DESCENT is off and swapping on the first 30 step tree led to a 29 step tree PAUP would not swap on the other starting trees Thus if you want to force swapping on all minimal starting trees you must enable STEEPEST DESCENT Steepest descent is sometimes effective in finding other islands because 1t may allow the search to take a different path than it would otherwise One way to hop islands is to swap on nonminimal as well as minimal trees steepest descent is similar in that more nonminimal trees are tried as the search continues to find shorter and shorter trees For many data sets my experience is that steepest descent causes the program to get bogged down finding thousands of trees that will eventually be discarded anyway and a better strategy 1s to perform more replicates of the random addition sequence without steepest descent Searching under Topological Constraints Topological constraints are a new feature in version 3 of PAUP They are of two types monophyly constraint trees which contain all of the taxa in the data matrix and backbone trees which contain a subset of taxa To search under co
233. ition procedure Alternatively you can request that trees already in memory be used as the starting point by specifying NOSTEPWISE FROMTREE and or TOTREE can be used to select a subset of the available trees as the starting point for branch swapping ADDSEQ SIMPLE CLOSEST ASIS RANDOM Use ADDSEQ to specify the addition sequence to be used in the stepwise addition procedure These addition sequences are described in the section Stepwise Addition Options for ADDSEQ SIMPLE REFTAX reference taxon number By default the first taxon in the data file is used as the reference taxon Use REFTAX to specify an alternate reference taxon This option is relevant only for unrooted tree searches For rooted tree searches the hypothetical ancestor see ANCSTATES and ASSUME commands is used as the reference taxon Options for ADDSEQ RANDOM PAUP 3 1 USER S MANUAL 201 NREPS number of replications Use NREPS to specify the number of random addition sequence replications to be performed The default is 10 RSEED seed value Use RSEED to initialize the seed used to generate pseudorandom numbers used in obtaining random addition sequences The seed value must be between 1 and 2147483646 inclusive The seed is set to 1 when PAUP first starts it is updated every time a new random number is generated NO RSTATUS A status report showing the results of each random addition sequence replication is normally output you can
234. ituations occur in the first place The heuristic search uses the specified addition sequence finding the shortest tree until the last step when it must place the last taxon such that the constraints are violated This will then be the starting point for branch swapping Because you have asked the algorithm to deliberately narrow the solution space before branch swapping even begins it is possible that all optimal converse trees will not be found by branch swapping You have simply asked the algorithm to begin swapping on a tree from which it cannot reach the shortest trees that are incompatible with the constraint tree In some instances trying different swapping and addition options may find more trees but this is not always true Again if your data set is small enough to allow branch and bound searching this will not be a problem What are the solutions if an exact search is not possible In that case you must be as diligent as possible in trying different addition sequences especially random addition and different swapping algorithms Keeping Near Minimal Trees You can keep trees less than or equal to a certain length by setting the Keep all trees lt length ___ option in the search procedure dialog boxes or specifying KEEP ength in the HSEARCH BANDB and ALLTREES commands Selecting steepest descent forces the examination of suboptimal trees even when a shorter one has been found This examination of suboptimal trees can be i
235. k Of course this feature makes it extremely important to spell block names correctly otherwise essential blocks might be unintentionally skipped over during processing Every block starts with begin block type directive and ends with an endblock directive Each block is composed of one or more commands each terminated by a semicolon If a command name within a block is unrecognized a warning message is issued and the rest of the command ignored by skipping forward to its terminating semicolon Optionally PAUP will abort processing of a file if an unrecognized command is encountered Blocks can also be given a name immediately after the block type but before the semicolon Currently PAUP simply ignores these names Within a block the only restriction on the ordering of commands is the following one any command which affects the operation of a second command must precede the second command in the file Le only one pass is made through the file and all commands are executed immediately when they are encountered there is no lookahead capability Process is defined loosely Strictly speaking NEXUS conforming programs are free to ignore commands within these blocks For instance some commands may provide information or instructions that are irrelevant to a particular program The only requirement is that the program should at least provide the option to continue processing the file after encountering a command that it e
236. k N Y Hillis D M and J P Huelsenbeck 1992 Signal noise and reliability in molecular phylogenetic analyses Journal of Heredity 83 189 195 Holmquist R M M Miyamoto and M Goodman 1988 Analysis of higher primate phylogeny from transversion differences in nuclear and mitochondrial DNA by Lake s methods of evolutionary parsimony and operator metrics Molecular Biology and Evolution 5 217 236 Huelsenbeck J P 1991 Tree length distribution skewness an indicator of phylogenetic information Systematic Zoology 40 257 270 K llersj M J S Farris A G Kluge and C Bult 1992 Skewness and permutation Cladistics 8 275 287 Kluge A G and J S Farris 1969 Quantitative phyletics and the evolution of anurans Systematic Zoology 18 1 32 Lake J A 1987a Determining evolutionary distances from highly diverged nucleic acid sequences operator metrics Journal of Molecular Evolution 26 59 73 278 PAUP 3 1 USER S MANUAL Lake J A 1987b A rate independent technique for analysis of nucleic acid sequences evolutionary parsimony Journal of Molecular Evolution 4 167 191 Lundberg J G 1972 Wagner networks and ancestors Systematic Zoology 21 398 413 Maddison D R 1991 The discovery and importance of multiple islands of most parsimonious trees Systematic Zoology 40 315 328 Maddison W P 1989 Reconstructing character evolution on polytomous cladograms Cladistics 5 365 377 Maddison W
237. lable as an option Save Trees to File Use this command to save some or all trees currently in memory to a file in NEXUS PHYLIP FREQPARS or Hennig86 formats S Sample data files y amp Hard Disk O S Tasupa test Hnac yehus R fnglesperms ae Desktop Hnglasperms trees Herheridac ear B Carleen Characters Eject Save treefile as Cancel 5 Taxon test trees ormat EJ Include branch lengths EJ Root trees before saving Save All trees Otrees through The Save Trees to File dialog box The file format is chosen using the pull down menu in the Format option NEHUS no translation table FREQPARS 1 0 PHYLIP 3 4 Hennig86 1 5 The Format pull down menu You can choose to root the trees before saving by clicking on the Root trees before saving checkbox not available for FREQPARS format If PAUP 3 1 USER S MANUAL 273 you do select the Root trees before saving checkbox the Rooting Options button becomes active and selecting it will bring up the standard Rooting dialog box You can include branch length information by clicking on the Include branch lengths checkbox but only 1f the file format is NEXUS or PHYLIP NEXUS files can also be saved with or without a translation table If a translation table is selected PAUP will include a table that maps tokens in the tree specification to valid taxon names See the section The TREES block for a description of the use of
238. le data files included on your distribution disk The ASSUMPTIONS Block The ASSUMPTIONS block is used to declare character types and weights assumption sets and ancestral states polarities Its syntax is BEGIN ASSUMPTIONS block name OPTIONS 78 PAUP 3 1 USER S MANUAL DEFTYPE default character type POLYTCOUNT MINSTEPS MAXSTEPS USERTYPE name STEPMATRIX CSTREE description CHARSET character set name character list TYPESET name character type character list character type character list WTSET weight set name character weight character list character weight character list EXSET exclusion set name character list ANCSTATES ancestor name character state character list character state character list ENDBLOCK An example ASSUMPTIONS block follows with examples of each of the available commands begin assumptions options deftype ord polymcount addsteps usertype myOrd 4 0123 Sol 1 21 37 2 usertype myTr cstr 0 1 a 2 3 b c typeset mixed irrev 1 3 10 unord 5 7 charset odd 1 2 charset even 2 2 wtset one 2 1 3 6 11 15 3 7 8 wtset two 2 4 9 3 1 3 5 e 9 a a ee 3 2 1 xset nolarval 1 9 nestates allzero 0 ALL ncstates allone 1 ALL ancstates mixed 0 1 3 5 8 11 1 2 4 9 15 endblock Any number of USERTYPE TYPESET
239. le the command bandb upbound 30 will set the upper bound at length 30 or you can let PAUP find it by stepwise addition Ordinarily you will not need to specify an upper bound However the better the initial upper bound the faster the branch and bound search will proceed For large and or messy data sets you can sometimes reduce run times by performing a more extensive heuristic search and using the resulting tree length as the starting upper bound The options for stepwise addition are slightly different than those available in a heuristic search and include three choices furthest asis or simple PAUP can also output a frequency distribution of tree lengths using the FD option or the Tree length frequency distribution item in the Branch and Bound dialog box The maximum tree length to be shown must be specified by the KEEP length This option provides a way to examine the left tail of the distribution of tree lengths if there are too many taxa to obtain the full distribution via an exhaustive search see below To accomplish this you PAUP 3 1 USER S MANUAL 113 can perform an initial branch and bound search to determine the minimum tree length and then perform a second search specifying FD and a KEEP length some number of steps larger than the minimum length In this way you can determine the number of trees that are one step longer than the shortest two steps longer and so on The output for a branch and bound search is nearly i
240. mand lenfit 1 5 total single all ci will for trees one through five output tree lengths and consistency indices for all characters single all overall tree lengths and ensemble consistency indices total If you wish only to output the range of minimum and maximum tree lengths or best and worst fit measures for each character use the RANGE option For example the command lenfit single all total ci notl requests consistency indices and suppression of tree lengths for all characters as well as ensemble consistency indices Sum of min possible lengths 14 Sum of max possible lengths 36 Tree 1 2 3 4 5 6 Cl 0 609 0 609 0 609 0 609 0 609 0 609 Consistency indices for each character Tree 2 3 4 5 6 7 8 1 0 5 5 0 0 5 0 500 0 500 2 0 5 KD 0 E 0 0 333 0 500 3 0 5 333 ES 0 3 0 500 0 500 4 0 5 sD 0 iS 0 0 333 0 500 5 0 5 5 0 2 5 0 500 1 000 6 0 ab 05 0 lt 5 0 0 333 0 500 Best 0 E En 0 0 0 0 500 1 000 Worst TO 9 333 5 TO 9 0 333 0 500 Tree 9 4 5 6 1 0 0 vo ob 5 5 2 0 0 0 03 5 5 3 0 0 a5 0 0 0 4 0 0 0 lt 5 5 5 5 0 0 D 25 5 5 6 0 0 0 5 5 5 Best 1 0 0 0 0 0 0 Worst 0 0 wd 23 5 3 If there are many characters in the data matrix the most efficient use of this option is probably to limit output only to characters which vary over trees in memory i e u
241. mat to be imported You can select from text PHYLIP Hennig86 GCG MSF or NBRF PIR by using the pull down menu that appears when Import File is selected Examples of GCG MSF and NBRF PIR file formats are included in the sample data files included with the PAUP program Page Setup Print File Echo to Printer Print Selection Log Output to Disk Execute manual Export File Import File a PHYLIP 3 4 An Hennig86 1 5 0 GCG MSF NBRF PIR Text The Import File menu 232 Quit PAUP 3 1 USER S MANUAL Once that is chosen the standard file opening dialog box will appear unless PHYLIP is chosen in which case you must select the PHYLIP file options as well as the file Sample data files v D 5 Taxon test D Anacyclus D Angiosperms D Angiosperms trees D Berberidaceae D Cartoon Characters D cartoon phylip D command file Type of PHYLIP data file Discrete 0 1 Restriction site O Interleaved DNA Protein The PHYLIP file importing dialog box These PHYLIP file formats are described under Importing Data and Trees from Other Programs Remember that in the current version of PAUP PHYLIP options that require extra lines of input e g A T W etc are not supported and will probably cause the conversion to fail PAUP will successfully process PHYLIP user defined trees Although PAUP can handle the format of tree files output by Hennig
242. mber of trees you which displayed on each page If you print multiple trees per page you will probably need to reduce the font size and line width accordingly This menu also allows you to choose whether the trees are oriented horizontally or vertically The main Print Trees menu also allows you to choose font type and size and line size for the plots These may take some adjusting if your are printing multiple trees per page and some experimentation is usually necessary to get the best results The More options button allows you to select taxon label orientation details of circleTree display and whether or not the output will be in PostScript format 268 PAUP 3 1 USER S MANUAL XX Use PostScript Taxon label orientation Leave horizontal Rotate to match branch angle Rotate degrees g Taxon labels Horizontal O Radial Origin degrees clockwise from 3 o clock Extent Full O C degrees i Circle tree options The More Options dialog box Due to the limitations of Macintosh graphics at 72 dpi CircleTrees do not print very well unless you print directly to a PostScript printer so use a LaserWriter for best results The figure below illustrates the types of printed trees that PAUP can generate ABCODGHE F ABCODGHE F ES CS H G E G D D H E E A B c C 7 c d B F A Tree printing options a rectangular cladogram b slanted cladogram c phylogram d CircleTree
243. mines the action taken by PAUP if the limit on the number of trees that can be stored MAXTREES see above is reached during a search or a tree file operation If INCREASE PROMPT you will be given the opportunity to increase MAXTREES If INCREASE AUTO MAXTREES will automatically be increased by a number of trees equal to the current AUTOINC setting see below If INCREASE NO MAXTREES will not be increased and no prompt will be issued In this case a tree buffer overflow occurs which can affect the effectiveness of the search in progress The tree buffer overflow condition will be documented in the output AUTOINC n The AUTOINC value species the number of trees by which MAXTREES is increased when the number of trees saved PAUP 3 1 USER S MANUAL 215 reaches MAXTREES and the INCREASE AUTO option is in effect AUTOINC is initially set to 100 The following three options specify whether PAUP sounds a beep when various kinds of errors occur NOJERRORBEEP Ordinarily PAUP beeps to alert you that an error message has been issued Specify NOERRORBEEP to suppress these beeps NO QUERYBEEP Ordinarily PAUP beeps to alert you when it stops for your input before it can continue a process Specify NOQUER Y BEFP to suppress these beeps NO KEYBEEP Ordinarily PAUP beeps when you type a key that is invalid in the current context Specify NOKEYBEEP to suppress these beeps NOJERRORSTOP Ordinarily PAUP stops process
244. mmand or the Tree Order menu command Because systematists tend to think of the root as the lowest point in the tree diagram assume that the output has been rotated 90 clockwise so that left is actually up right is down etc If TORDER STANDARD trees are ordered so that taxon names appear from left to right in as nearly as possible the same order as the taxa were presented in the data matrix If TORDER RIGHT the tree is ladderized a term used in MacClade to the right Technically this means that rotations are performed around each internal node so that the descendant node with the greatest number of ultimate descendants is placed on the right and the descendant node with the fewest descendants is PAUP 3 1 USER S MANUAL 151 placed on the left TORDER LEFT is just the opposite of TORDER RIGHT Finally if TORDER ALPHABET the tree is ordered so that the taxon names appear as nearly as possible in alphabetical order regardless of the order in which the taxa were presented in the data matrix Examples of the same tree in different orders are TORDER STANDARD A EEE EEE EEE EEE EEE LEE EEE EEE EEE LEE KEE EEE taxonA gt AEE EEE EEE EEE EEE EEL EEE EEE EEE taxonB fle cccccccd gt Marron anar taxonc gt ficccccccccg fle cccccccd flaccccccce taxonF gt gt fete oF fice ccccd eorccrcec taxonG gt ficcccccccd Torre taxonD gt Fi EEE EEE EEE EEE EEE EEE EEK taxonE Fi EEE EEE EEE EEE EE EEE EEE EEE EEE EEE EEE EEL EE
245. models of character transformation that cannot be expressed under any other available coding method For instance morphologists can define partially unordered characters in which some transformations are required to follow a specified order but others may occur freely In the example below a character state tree for several alternative present states is specified but a transformation to the absent condition is permitted to occur with equal cost from any state OABCOD Oo 0 23 22 A 10121 absent B 0 12 Cc 2 0 3 D L 23 0 Character state graph and associated stepmatrix for a partially unordered character Transformations between states A B C and D follow a character state tree but immediate losses may occur from any state PAUP 3 1 USER S MANUAL Partial irreversibility where reversals are permitted for some transformations but not for others can also be implemented using stepmatrices 1 0 123 1 1 O 0123 ie DOTA BD DACA 3 iii0 Character state graph and associated stepmatrix for a partially irreversible character State 3 cannot reverse to any other state hence the entries of infinity i in the last row of the stepmatrix Most of the other standard types may be defined equivalently as stepmatrices For example the stepmatrices OF 12 0 01 2 LLa 1 0 1 and i0 1 2 10 1 1 0 110 correspond to the ordered unordered and irreversible types respectivel
246. mp O OHNY C amp O Or amp N amp N PY VVvVVVVV VV VV VV T O T a T nn T 4 4 4 o o 4 4 4 You can see the arrows are not all the same type This is because PAUP uses different arrows to indicate different types of change A double lined arrow means that the change occurs in all possible reconstructions 1 e is unambiguous A single lined arrow indicates that change occurs under some reconstructions but not others A double headed arrow indicates that the direction of change is undetermined that is the change occurs along the branch connecting the outgroup to the ingroup A single headed arrow indicates that the direction of change is unambiguously within the ingroup In the example above all four types of arrow are present Remember that these arrows are a reflection of the tree and optimization criterion Character changes with double lined arrows will not be affected by a different optimization scheme while those with single lined arrows will This is also true of the ouput listing the apomorphies for each branch as below Note that multistate characters have a CI of 1 000 when each state is derived only once To get the list of apomorphies for each branch type describe 1 apolist This gives the following output PAUP 3 1 USER S MANUAL 129 Apomorphy lists Branch Character Steps C Change taxonl lt gt node_12 2 uo lt gt 7 1393 lt gt 9 y lt gt 12 i lt gt node_12
247. n Chapter 2 for information on how to designate character polarities Irreversible Characters Irreversible characters are equivalent to ordered characters with the additional constraint of irreversibility being imposed Camin and Sokal 1965 As for ordered Wagner characters character states are linearly ordered according to their position in the SYMBOLS list specified in the FORMAT command of the DATA block However we now prohibit transformations from a more derived state to a less derived state For example if SYMBOLS ABCDE and state A is defined as the ancestral state the character state graph is all zero ancestor one step would be added to the tree length for every character in which the site was present in both the ingroup and outgroup corresponding to the site gain along the basal branch With an all missing unknown ancestor this additional step would not be required because the change would be from the missing state to present see MISSING DATA Thus the only difference between the two methods is that a constant is added to or subtracted from the tree length 14 PAUP 3 1 USER S MANUAL 0000 Polarity specification Although character transformations must always proceed in the direction of a more derived state they need not proceed in a single direction with respect to the SYMBOLS list Any state in the SYMBOLS list may be designated as the ancestral state with states preceding and or following this state
248. n is added to each of the three branches of the three taxon tree thereby generating all three possible unrooted trees for the first four taxa row 2 The process continues in a similar fashion adding the ith taxon to each branch of every tree containing 1 taxa generated during a previous step Thus for example row 3 contains all 15 possible trees for the first five taxa obtained by adding the fifth taxon to each of the five possible branches for the three trees obtained at the four taxon stage This process demonstrates the rationale for equation 1 for counting the number of possible unrooted bifurcating trees for T taxa for each of the possible trees for 1 taxa there are 2 i 1 3 21 5 32 PAUP 3 1 USER S MANUAL branches to which the th taxon can be connected Note that the order of addition is immaterial one could just as well have chosen a taxon at random to connect to the tree at each step B C 1 a A B D B B C D 2A 2B 2C A A B E E D B B D C B D U W 3A 3B 3 3D 3E A A j A B ED C D E E c D Cc D Cc B B N 3F 3G A A A c B Cc M D C E N A D E c A E D A B E E G B CE DB E B LE 7 pl E 3K 3L 3 3N 30 A A A A Generation of all possible binary tree topologies for five terminal taxa Evaluation of Equation 1 see Table 2 quickly reveals why exhaustive search procedures are useful only for small numbers of taxa There are over 2 million trees for 10 taxa and 34 million trees for 11 ta
249. n logging to a different file NewOutput replacing it if it already exists MEMAVAIL Use the MEMAVAIL command to find out how much memory is still available from the operating system DOS portable version only There are no options 208 PAUP 3 1 USER S MANUAL OUTGROUP Use the OUTGROUP command to assign one or more taxa to the outgroup The syntax is OUTGROUP taxon list ONLY Unless ONLY is specified taxa specified in the taxon list are simply added to the current outgroup If taxa have already been assigned to the outgroup and you want only those taxa specified in taxon list to remain in the outgroup specify ONLY any taxon that is not explicitly specified in the list will be transferred to the ingroup POSSPLOT Use the POSSPLOT command to request output of possible character state assignments MPR sets for one or more characters on one or more trees by superimposing the possible character states for each node on a diagram of the tree The syntax is POSSPLOT character list options The character list specifies the character s for which possible character state assignments are shown and consists of one or more character numbers character names or character set names see Character lists earlier in this chapter for details The following options are available TREES tree list The tree list specifies the tree numbers for which possible character state assignments are to be shown If thi
250. nce package 2 Hennig 86 J S Farris s IBM PC program for phylogenetic analysis and 3 tab delimited text a general format readable by Excel and other programs Hard Disk R gef exeeptian matin fleck HAMS TREES AN Backbone 2 Desktop hackbone 3 A Backbone example ii backbone test Export file as p Hennig86 1 5 Tab delimited text MERA Simple text Export X Data matrix tes descriptions O Interleave Max characters line The Export File dialog box A data matrix must have been successfully input to PAUP via a DATA block for this command to be available If there are trees in memory you may export these as well however tree export is supported only for PHYLIP and Hennig 86 formats To export the data and or trees name the exported file in the field provided by the standard file dialog box choose the desired format pop up menu and export options check boxes and click on Export PHYLIP s discrete 0 1 programs are restricted to binary characters Consequently if the format of the current PAUP data file is standard rather than DNA RNA or Protein and there are multistate characters present you will be presented with the following options Phylip programs for discrete 0 1 characters cannot handle characters with more than two states For UNORDERED multistate including stepmatrix characters Exclude from exported file Export unmodified For ORDERED multistate incl
251. ncreased by selecting Keep all trees lt length ___ This may allow the hopping of islands so that the global optimum is reached but can also substantially increase computation time One of the values of setting the Keep all trees lt length ___ option is that it allows one to examine the support for particular groupings in trees that are slightly longer than the optimal tree s We might for example keep all trees 3 steps longer than the optimal trees If a particular group is present in the optimal tree s and all the suboptimal trees which are kept then we might place more confidence in the support for that grouping If however a particular clade is found only in the shortest trees and becomes nonmonophyletic in trees one or two steps longer we might have less confidence in the support for that group 50 PAUP 3 1 USER S MANUAL The relationship between clade monophyly and tree length can be quickly evaluated by the use of topological constraints see Searching under topological constraints above but setting the keep option allows evaluation of the whole topology not just those aspects set in the constraint tree although as an alternative multiple constraint trees can be examined This strategy becomes more interesting when the frequency distribution of all possible tree lengths is examined produced as a by product of an exhaustive search if that search type is possible for the data set see the section Exhaustive searche
252. ng the data matrix Choose Assumption Sets Use this command to invoke previously defined assumption sets 252 PAUP 3 1 USER S MANUAL Choose Assumption Sets TypeSets y _wtsets y ExSets MA AncStates v Y Al Included larval Cen CD The Choose Assumption Sets dialog box Four kinds of assumption sets are available TYPESETs and WTSETs specify character types and weights respectively for sets of characters See the sample data file 5 taxon test data set for several examples EXSETs define sets of characters that are to be excluded from consideration during searches ANCSTATES specify ancestral states for each character Assumption sets are particularly useful when complex combinations of assumptions are being used By declaring assumption sets complicated sequences used to assign character types and weights etc can be avoided THE SEARCH MENU A ___ Items on the Search menu allow you to search for Show Constraints trees using heuristic or exact algorithms to perform Heuristic l bootstrap analyses to estimate the reproducibility ea of groupings and to perform Lake s method based i on linear invariants for DNA RNA sequence data Lake s Invariants Bootstrap Also commands dealing with input and selection of Random Trees constraints are included on the Search menu The Search menu Load Constraints Ordinarily topological constraints are defined via
253. nodes see above The unresolved part of the backbone tree is even less restrictive as even the relative topology within that part is not specified In the end whether you use a backbone or monophyletic constraint tree depends on your immediate goals if you are interested in the lengths of trees which are compatible with a particular hypothesis of monophyly use a monophyletic constraint tree however if you are interested in which trees are compatible with a less restrictive topology one that includes fewer taxa use backbone constraints In either case you may still be left with trees compatible with the constraint and those which are the shortest length if they are different from the former group Heuristic Searches and converse constraints When converse constraints i e trees are kept only if they are incompatible with the constraint tree are used during a heuristic search PAUP 3 1 USER S MANUAL 49 special problems may arise This strategy will cause trouble if a heuristic search does not find the shortest trees that do not sashays the constraint or if a heuristic search finds some but not all of those shortest trees The latter is probably more common but the former is more serious If for whatever reason the data set is one that does not lend itself to a branch and bound analysis 1t may be impossible to check whether either of these events has occurred so some degree of skepticism might be warranted Why might these s
254. nown is interpreted as missing data In addition two equate macros are predefined B DN asx asp or asn Z EQ glx glu or gln If you want to add additional character state symbols to the SYMBOLS list implied by the DATATYPE in effect you can still use the SYMBOLS option in the format command These additional symbols will be inserted at the beginning of the SYMBOLS list For example if you wanted to mix nucleotide sequence and restriction site characters in the same data matrix you could use the following FORMAT command 74 PAUP 3 1 USER S MANUAL format datatype dna symbols 01 The symbols list would then be set to 01 ACGT See Alignment Gaps below for another example Matching the states in the first taxon You may define character states relative to the states of the first taxon by using the MATCHCHAR option For example the following MATRIX command is equivalent to the one above format datatype dna gap matchchar matrix One ATGCT ATCCG TCATG ACCTA Two es Es A O Three Chia Geo GGA Tita Four C AA A A GTC d In many cases particularly with sequence data it is convenient to separate the data into blocks of characters This is accomplished using the INTERLEAVE option For the sequence data example above we could first supply the data for positions characters 1 10 and then for positions 11 20 as follows begin data dimensions ntax 4 nchar 20 format d
255. nstraints you must define and select a constraint tree this tree is used to restrict the set of trees retained by heuristic or exact searches Monophyly constraint trees Monophyly constraint trees are usually incompletely resolved 1 e contain one or more polytomies These polytomies on the constraint tree indicate uncertainty with respect to relationships rather than simultaneous divergence into more than two descendant lineages A tree being evaluated the trial tree is said to be compatible with the constraint tree if and only if it is either identical to the constraint tree or it can be transformed into the constraint tree by deleting collapsing one or more of its branches In more biological terms constraint tree compatibility means that any statement of relationship among taxa implied by the constraint tree must also be true for the trial tree Thus trial trees compatible with the constraint tree are usually more highly resolved versions or refinements of the constraint tree For example in the PAUP 3 1 USER S MANUAL 45 figure below trial trees b and c satisfy the constraints imposed by tree a but trial trees d and e do not ABCDEFGH a ABCDEFGH ABCDEFGH ABCDEFGH ABCEDFGH Y e Example for constraint tree compatibility a The constraint tree b c Two trees that are compatible with the constraint tree d e Two trees that are incompatible with the constraint tree lt If a group of particula
256. numbers in the Heuristic dialog box or they can be obtained by stepwise addition using STEPWISE or by selecting this in the Heuristic dialog box Using trees in memory allows you to use the results of a previous analysis as the starting point for a new search Any or all trees in memory may be used so swapping can start on both optimal and suboptimal trees NOTE Selecting the starting tree range cannot be done if trees in memory have been filtered You must either start with tree s obtained by stepwise addition or swap on all trees in memory If trees in memory are not used you must specify which addition sequence to use to obtain starting trees You must also specify how many equal length trees PAUP will keep during each step of stepwise addition using the HOLD option or by entering this value in the Heuristic dialog box Stepwise addition is selected using the ADDSEQ option or by choosing it in the Heuristic dialog box You may opt to add taxa in the sequence they are in ASIS to add the closest taxon at each step CLOSEST by following the simple algorithm SIMPLE of Farris 1970 or in random order RANDOM The SIMPLE option requires that you choose a reference taxon default is the first taxon in the matrix see the section Stepwise Addition NOTE The reference taxon for simple addition sequence is ignored if the tree is rooted in that case the ancestor serves that role Random addition requires a few more options
257. o other via by equating one symbol to the other If you say issued the command format missing equate ee then both hyphens and question marks in the data matrix would be treated as missing data Equate macros are also useful when more than one character state is being assigned to some taxa see Multistate Taxa For example if you specify format equate R AG Y CT then R s and Y s ambiguity codes for purine A or G and pyrimidine C or T respectively in the data matrix are converted to the corresponding character state set In fact use of DATATYPE DNA below causes this equate macro as well as others to be predefined Using a subset of the characters By default PAUP uses all of the characters in the data matrix The IGNORE and ZAP options can be used to restrict the characters actually used by the program If you request IGNORE INVAR invariant or constant characters are ignored Alternatively you can specify IGNORE UNINFORM to request that uninformative characters be ignored To ignore specific characters you can use the ZAP option For example the command options ignore uninf zap 1 10 causes the first ten characters plus all other uninformative characters to be ignored Ignoring a character is effectively the same as physically removing it from the data matrix the only meaningful difference is that physical removal would affect the numbering of the downstream characters whereas ignorin
258. o penalize characters which fit the tree s poorly are homoplastic and reward characters which fit the tree well By successively evaluating and reweighting the goal is to arrive at some stable topology or topologies In practice this amounts to running a search choosing a weighting function reweighting and running another search Reweighting can be based on the consistency index CI retention index RI or rescaled consistency index RC Since multiple most parsimonious trees will generate different weight sets PAUP has the ability to weight by the best worst or mean fit to multiple trees It is up to the user to justify this approach especially given the range of options under which it may be run For example the primary variable is the index chosen as the basis for reweighting but it is not immediately 62 PAUP 3 1 USER S MANUAL clear which index should be preferred You must also choose whether to reweight based on best worst or mean fit to the trees The default settings of best fit according to rescaled consistency correspond to the xsteps w command of the Hennig86 program Farris 1988 Since different index and fit choices can potentially lead to very different trees you must be very careful about conclusions based on a search strategy with many arbitrary components built in If you do pursue this method it is probably a good idea to experiment with different reweighting schemes in order to get some crude idea of
259. obtained by using the observed states that is not one of the observed states The length of tree d is 7 so state D fails the test and cannot be a candidate for assignment to the interior node 3 3 o C 2 2_K o0 Cc K Cc K a b 3 3 AS AN C K Cc K c d Example of the 3 1 rule a and b three taxon trees with observed states at node length 5 c non observed state that satisfies the rule with length 5 d state which does not pass the rule length 7 136 PAUP 3 1 USER S MANUAL For real data sets I have never discovered a case for which STEPMATRIX THREEPLUS 1 fails to obtain an optimal reconstruction as long as all triplets of states satisfy the triangle inequality However hypothetical counterexamples have been constructed Wayne Maddison personal communication For example the 3 1 rule excludes a state required to obtain an optimal reconstruction for the following stepmatrix a b cde a 313 L389 b Lota Ae wl ss 29 e 13 13 1359 d 13 13 13 gt 9 e O Os Der 9 If the only observed states are a b c and d the 3 1 rule will not allow state e to be reconstructed at internal nodes The cost of assigning any of the observed states will always be 26 while the cost of assigning e will always be 27 so the 3 1 rule excludes it as a candidate for internal nodes However for the following four taxon tree state e is actually the best state for internal nodes even though i
260. od that you have found all of the islands and all of the most parsimonious trees contained in those islands If on the other hand you are still finding new islands after 100 replications a high probability exists that more trees remain to be found SThis statement is strictly true only when the collapse zero length branches option is declined Rearrangements are defined only in terms of binary trees see Branch Swapping Before a tree in which zero length branches have been collapsed is input to the swapping procedure polytomous trees are first converted to binary trees by arbitrarily resolving polytomies into dichotomies Although precautions are taken in the algorithm to avoid consistently resolving the polytomies in exactly the same way this arbitrary resolution process can nonetheless prevent some trees in an island of binary trees from being found Maddison 1991 PAUP 3 1 USER S MANUAL 39 The biggest drawback of stepwise addition algorithms is that they are too greedy Like the nearsighted pilot who is unable to scan the horizon and instead must simply proceed up the nearest hill these methods strive for optimality given the current situation rather than attempting to look more broadly into the future Thus one placement of a taxon may be best given the taxa currently on the tree but that placement may become suboptimal upon the addition of subsequent taxa Once a decision has been made to connect a taxon to a certain point h
261. odes of a 3 branch tree Then every character state contained in the stepmatrix definition is assigned in turn to the single internal node and the length required for the character given these state assignments is computed If assignment of an unobserved character state to the internal node yields a length less than or equal to that obtained from one of the three members of the triplet this unobserved state is added to the set of candidate states This would be achieved by the command set stepmatrix threeplusl How does the 3 1 rule work in practice Take the following stepmatrix for amino acids as an example this is taken directly from the PROTPARS example included with the PAUP program only the stepmatrix is reproduced here Each of the symbols designates either one or more amino acids a stop codon or a deletion corresponding to the standard IUB single letter amino acid codes see the section Predefined formats for molecular sequence data 134 PAUP 3 1 USER S MANUAL ACDEFGHIKLMNPQR12TVWY AI 0021 112 1 2 2 2 22 371 2212 01 Did ds 3 1 2022 1272221222 1 220011 3 D 12012112222122222212123 E 1 2100 22 12241 222 22 04d DoDD 221 3 F 2252 22 2 2D A TD 2 21 23 61 11112022222222 2 De AS Bcd 3 H 221222 22 21 202 22 2 123 2222 2 2 1 1122 2 22 23 KI 222 12 22 021 1722 2 2 222 3 L 2222 2 2 12 22 12 3 M 2222222 1 0 2 2 22 2 2 X 323 N 221222 12202222 2
262. odes on the tree diagram are labeled with a node number that is referenced by other output information change lists apomorphy lists etc NOLABELNODE can be used to suppress the labeling of internal nodes XOUT NONE TERMINAL INTERNAL BOTH Requests output of a table of character state assignments for each tree XOUT INTERNAL requests output of the character states assigned to internal nodes for each tree XOUT TERMINAL requests a listing of the original data matrix This option is not particularly useful as you would ordinarily use the SHOWMATRIX command to list the data matrix XOUT BOTH requests a listing of the original data matrix plus the states assigned to internal nodes NO CSPOSS Requests a listing of the possible character state assignments MPR sets for each tree See Character State Reconstructions in Chapter 2 for more information PAUP 3 1 USER S MANUAL 195 DOS EDIT NOJFVALUE Requests output of the F value and F ratio see F Value and F Ratio under Goodness of Fit Statistics in Chapter 1 ROOT OUTGROUP LUNDBERG MIDPOINT OUTROOT POLYTOMY PARAPHYL MONOPHYL NO TCOMPRESS OPT ACCTRAN DELTRAN MINF ANCSTATES ancestral states name NO SHOWIGNORE NO CMLABELS NO CMCSTATUS CMDCOLWID column width See Options Affecting Multiple Commands earlier in this chapter Use the DOS command to execute a DOS command and return immediately to PAUP or to enter
263. oint of the longest path connecting any pair of taxa The following command will output a tree with midpoint rooting describe 1 root midpoint Essentially this is a way to output a tree without explicitly rooting it and should only be used as a last resort Use of this option forces the use of MINF optimization which is the only optimization method that does not depend on the position of the root Saving Trees to Files If the search required to find a given set of trees uses a significant amount of computing time saving the descriptions of these trees to a tree file may be desirable for several reasons First you may need the trees at a subsequent time e g to obtain additional output information about the trees to plot the trees in a different way or to use the trees as input for a subsequent search after you have terminated the PAUP session If you have saved the trees to a file you can then recover the trees quickly without having to repeat the lengthy search Second the tree file can be input to other programs that read user defined trees Since the trees are written to the file as a NEXUS TREES block any NEXUS conforming program that accepts user defined trees e g MacClade will be able to read the tree file Other programs that follow the standard format for tree description described in the section Describing Trees e g PHYLIP PLOTGRAM will be able to interpret the tree file after minor editing 1 Note that PAUP s e
264. one using the POSSPLOT option of DESCRIBE or Possible state assignments item in the Describe Trees dialog box A typical command would be describe 1 possplot with the following output Possible state assignments for character 1 on tree 1 AK EEE EEE EE EEE EE EEE EE EEE EEE EEE EEE EEE EEE LEE EEE LEE EEE EEE 0 taxonl gt A EEE EE EEE EEE EEE EEE EEE EEE EEE ELE EEE ELL EE EEK 0 taxon fic cc cccccccd AEE EEE EEE EEE EEE EEE EEE EEE OEE 0 taxon3 ficccccccccdd AEE EEE EEE EE EEK 1 taxons fic cc ccccccae fle cc cccecccecccccccccce 0 taxons NAO OCECCCEE fleccccccccce 1 taxon ficccccceccc ccccccccccee 1 taxon The two question marks indicate that the assignment of character 1 at those nodes is ambiguous they do not indicate that the state is missing default meaning of in a data matrix Reconstruction of multistate characters may be ambiguous or two or more states may be possible at a node as below 132 PAUP 3 1 USER S MANUAL Possible state assignments for character 3 on tree 2 taxonl taxon2 taxon3 taxon4 taxon5 taxon6 taxon7 NEELIE RE RRE REE RRE REE RRE RE RRE RRE REE ERRE ERRATA RRE ERRE RE gt PECE EEEE EEEE EEEE EEEE EEEE forced ERRE RRRERE RRE ERRE TEE RR RRE ERRE EEE RARE EEK ficccccccccQ2 ERAN ERRE er ficccccceccQ2 RAEE RENE RENE EN ficcccccccce fle cc ccccccce PrPONNOO Fic OOD EEE EEE EK You can also see the assigned and possible reconstructions by choosing states for interior node
265. ons such as and issued the command exclude 4 5 then all but the first and last characters would be excluded i e name 4 through number 5 Because of the potential confusion it is suggested that you always include at least one nondigit in alphanumeric character names The name ALL is reserved for a character list containing all of the characters in the data matrix see below You may not name a character ALL Other names Identifiers are also used for TYPESET WTSET EXSET CHARSET TAXSET and ANCSTATES definitions These names follow the same rules as taxon and character names and have a maximum length of 10 characters Common Command Elements The following elements are used in more than one command and are defined here to minimize redundancy Taxon lists A taxon list is a sequence of one or more taxon identifiers names and or numbers If two taxon identifiers are separated by a hyphen this indicates that the range of taxa between the first taxon and the second taxon inclusive are to be included in the list The second taxon must have a higher number than the first taxon PAUP 3 1 USER S MANUAL 169 A taxon list composed of the reserved name ALL specifies all of the taxa in the data matrix Examples of valid taxon lists assuming that corresponding taxon identifiers have been defined in the DATA block are as follows heteroclitus M23 045 S1 T5 1 3 5 8 all Character lists A character list is a s
266. onverse CONSTANTS Zch eG a a soe v esos S E aea 44 Keeping Near Minimal Trees ooooocnnncccnnocccnnnccconanccinnnoss 44 Zero Length Branches and PolytomMieS ooooonnncnnonononccnocnnncncnnncnnns 45 Tree to Tr ee DISTANCES corssier iia 48 Consensus e Sci 50 Strict CONSENSUS miccional a a 51 Semistrict CONS ASUS ai 52 Majority rule COMSCNSUS cceeeceeeeeceeeeeeeeeteeeetteeeeaees 52 Adams CONSENSUS m A 52 Consensus inNdic Ss sccsnieissn nnti nsin 53 Go dness f Fit Statisti s enice is 54 A Posteriori Character WelghtiN8 ooonooccnnoncccnoncnononcnnnnnccnnnnnnnnn 55 The SOUS ADD iaa os 56 Lake s Method of InvarlantS oococnnncccnoncccnonccnnoncnononcnnnnnccnnnncnnnnnos 56 Pseudorandom Number GeneratiON oooooononccconnccnnoncnononaninnnccnnnnnos 57 CHAPTER 2 LO LS a D o EE 59 Input files ai ico 59 The NEXUS Formati a2 Jase is 59 O tue Erne aa 60 NEXUS file identificati0N ooooooconnnnccconcccnnnnnnos 60 General format of NEXUS files oooo 61 Hes A ABC iresten yates e acta tut 61 Entering the matrix in transposed format 63 Placing taxon and character names ooooonnnnnonncnn 63 Character state symbols oocoonocccnnoncccnonccononaninnnnos 64 Using alphanumeric character names 65 Predefined formats for molecular sequence Matching the states in the first taxon 67 ALAMEDA 68 EQUATE Macro ccococccocnonnnnnnononoononannnnononorons nanos 69
267. oot At each internal node the algorithm determines which character states are potentially allowed and calculates the minimum possible lengths of the clade above that node for each allowed state The second stage proceeds from the root to the tips and determines a set of optimally reconstructed internal states by using the information from the first pass and information from the second pass below that node The optimal state minimizes the sum of two quantities the length of the clade above that node given each allowed state and the cost of transforming to each allowed state given the state assigned to the ancestor of that node 22 PAUP 3 1 USER S MANUAL In most cases there will only be one optimal assignment at a node but it is possible that more than one optimum may exist In that case a particular reconstruction may be favored on the basis of additional criteria The most common ancillary criteria are accelerated transformation ACCTRAN and delayed transformation DELTRAN Swofford and Maddison 1987 Maddison and Maddison 1992 Swofford and Maddison 1992 Basically these assign states at internal nodes in order to delay DELTRAN or accelerate ACCTRAN character transformation within the tree By pushing transformations up or down the tree these optimization methods lead to very different hypotheses of character change Delaying change will lead to a preference for two origins of a character state parallelisms while acceler
268. ooted binary trees for 8 taxa lis compatible with this constraint tree Of the 10 395 distinct unrooted binary trees for 8 taxa 1 is compatible with this constraint tree You may also choose to retain only those trees which are not compatible with the constraint tree In the first example above this will retain all trees in which the human chimp group does not appear This is achieved by using the CONVERSE and ENFORCE options during a search or by selecting this in the search dialog box For example hsearch enforce converse constraints testl will keep trees not compatible with the constraint tree test1 Either type of constraint tree may also be used to filter trees after a search though this is not the same as constraining during a search In that case the constraint trees are defined in the same way and are invoked either by using the FILTER command or by the Filter Trees menu command see the section Filtering Trees Complications can arise when using converse constraints with a heuristic search See the section Heuristic Searches and negative constraints IMPORTING AND EXPORTING TREES AND DATA PAUP also has the capability of exporting trees and data in the format of other programs saving you the effort of editing NEXUS files yourself This gives you the option of importing your trees and data matrix into another program for comparative analysis and vice versa This is achieved using the Import File and Export File
269. op of the highest hill That is we do not know whether we have arrived at a global optimum or merely a local optimum Fortunately the heuristic methods used in PAUP are so far as we can tell very effective Two basic strategies are used An initial tree or set of trees is obtained by stepwise addition Then the tree is subjected to trial rearrangements that attempt to find shorter trees This second process is called somewhat loosely branch swapping Stepwise Addition Stepwise addition operates by connecting taxa one at a time to a developing tree until all taxa have been placed First three taxa are 36 PAUP 3 1 USER S MANUAL chosen for the initial tree Next one of the remaining unplaced taxa is selected for next addition Each of the three trees that would result from joining the unplaced taxon to the tree along one of its three branches is evaluated and the one in this example but more can be saved at each step with the HOLD option whose length is optimal is saved for the next round In this next round yet another unplaced taxon is connected to the tree this time to one of the five possible branches on the tree saved from the previous round Again one of the resulting trees is saved for the next round The process terminates when all taxa have been joined to the tree This process is illustrated in the figure below where five taxa are added to a tree First three taxa are joined A B and C Next taxon D is attac
270. or Load the saved trees as backbone constraints Now perform a search while enforcing each constraint tree in turn and save all trees compatible with the backbone Output them using Lundberg rooting since the ANCSTATES and ancestor are identical you will be able to see each of the places the tree can be rooted If there is only one possible rooting a single tree will result from this process Below are the two trees that are recovered for the example above The first tree is identical to the tree displayed above PAUP 3 1 USER S MANUAL 141 Tree number 1 fleccccccce taxonA Mercer rlorereccere taxonB fleece ce 15660 cre eee taxonC flececcecceI4ececececececeececececececeee taxonF A TBE EEE EEE EEE EEE EEE KEE taxonE ADDL EE taxonD cececec1l fleccccccce taxonG gt FEE EEE EEE EEE EEE EEE EELN OCC COCK taxonH KURU RR a ancestor ree number 2 A EEE EEE EEE EEE EEE EEE EEL EEE LEE EEE ELE LEEK taxonA gt AEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE taxonB 6446616 gt A EEE EEE EEE EEE EEE EEE EEE EE ELLE EE taxonC ficc cece gt flecceccccccececcce taxonD ficccccccld flecccccl l flecccccce taxonG gt EEE ficcccccclOccccccce taxonH ficcccce 3 Fie EEE taxonE gt gt gt gt gt FEE EERE EEE EEE EEE EK taxonF FEE EEE EEE EEE EEE EEE EEE EEL EEE LEE EEE ancestor These trees can then be saved or analyzed further The last rooting option Midpoint rooting is straightforward the tree is rooted at the midp
271. ore previously deleted taxa A dialog box with two scrollable lists will appear 250 PAUP 3 1 USER S MANUAL Nondeleted taxa Deleted taxa taxonD Delete ancestor lt Restore lt Cancel ok The Delete Restore Taxa dialog box The list on the left contains nondeleted taxa and the list on the right contains taxa that have previously been deleted Select taxa in either list and click on the appropriate button to change the deletion status of the taxa If you have defined taxon sets taxsets you can select all taxa in the set in either list by pulling down the appropriate Taxsets menu Define Outgroup Use this command to define the outgroup A dialog box with two scrollable lists will appear Ingroup taxa Outgroup taxa 3 gt Fo Bulgrosp gt ancestor To Ingroup Cancel The Define Outgroup dialog box The list on the left contains taxa currently assigned to the ingroup and the list on the right contains outgroup taxa Select taxa in either list and click on the appropriate button to move taxa between the ingroup and the outgroup If you have defined taxon sets taxsets you can select all taxa in the set in either list by pulling down the appropriate Taxsets menu PAUP 3 1 USER S MANUAL 251 Show Character Status This command produces a listing of the current status of all characters Output includes the character type states observed exclusion
272. ormed for each resampling of the characters If METHOD BANDB the search is performed using the branch and bound algorithm CONLEVEL confidence value Use CONLEVEL to specify the minimum frequency of the bootstrap replicates expressed as a percentage in which a group is supported in order to be included in the bootstrap consensus tree loosely the width of the confidence interval For example to obtain a bootstrap consensus tree that shows only those groups which occurred on more than 80 of the trees you would specify CONLEVEL 80 CONLEVEL must be at least 50 which is the default NO KEEPALL If you request KEEPALL groups occurring at frequencies less than CONLEVEL will also be retained in the bootstrap consensus as long as they are compatible with all groups that are already included in the consensus Effectively this forces CONLEVEL 5S0 because any group occurring in 50 or more of the replicates will automatically be compatible with all more frequently occurring groups The search options are the same as for the HSEARCH command if METHOD HEURISTIC or the BANDB command if METHOD BANDB PAUP 3 1 USER S MANUAL 187 CHARSET The CHARSET command used to define character sets is ordinarily issued from within the ASSUMPTIONS block However you may also issue it from the command line or from within a PAUP block See Commands used in the ASSUMPTIONS Block for the description of this command CHGPLOT Use the CHGPL
273. orms 90 PAUP 3 1 USER S MANUAL Character state tree one e a Stepmatrix 10_1 TO a c g t a 10 1 10 c 10 10 1 g 1 10 10 t 10 1 10 FROM Because of the limitations of lineprinter style graphics this ouput for character state trees can be confusing For example usertype a cstree a b c showuser generates the following ouput a ko l b l At first glance this implies a bifurcating cstree which is not what was specified in the original usertype command The trick is to ignore all apparent nodes that do not have either a state symbol or an asterisk Thus a b and c all connect directly to because the apparent node joining the edge leading to b and c does not really exist Assigning Character Types There are four ways to specify character types e By using the DEFTYPE option in an ASSUMPTIONS block For example the following ASSUMPTIONS block sets the character type for all characters to ordered begin assumptions options deftype ord endblock Note that the default character type must be one of the predefined character types it can not be a user defined type PAUP 3 1 USER S MANUAL 91 e By defining a type set using one or more TYPESET commands in the ASSUMPTIONS block For example to make characters 3 and 7 unordered characters 9 through 13 Dollo and the remaining characters ordered you could use the following ASSUMPTIONS block begin assump
274. ossible Optimization This command allows you to specify the method used to resolve ambiguity when optimizing ordered Wagner and unordered Fitch characters Choose method for optimizing ordered Wagner and unordered Fitch characters Accelerated transformation ACCTRAN Delayed transformation DELTRAN Minimum F value MINF The Optimization dialog box Three choices are available 1 delayed transformation which can be thought of as preferring parallelisms over reversals 2 accelerated transformation which maximizes prefers reversals to parallelisms and 3 minimum F value which moves length from interior branches toward peripheral branches wherever possible See the section on Character Optimization for a discussion of the differences between ACCTRAN DELTRAN and MINF optimization Set Maxtrees Use Set Maxtrees to specify the maximum number of trees MAXTREES that can be held in memory by PAUP PAUP 3 1 USER S MANUAL 239 Enter a value for the maximum number of trees to be saved Prompt for a new value O Automatically increase by The Set Maxtrees dialog box The trees are stored in a tree buffer which can be resized on demand if it becomes full Setting the MAXTREES parameter to a large value will reduce the likelihood that the tree buffer will become full during a search at the expense of a larger chunk of memory being tied up and
275. otting Trees oooonnoccnnoncccnocccnnoncninnncnnnnnnos 141 Changing the Order of Taxa on Trees cecceeeeeeeeeeees 142 User Defined Teens riada 143 Defining and Using Topological ConstralntS oooocnnnnnnninnmmm 146 Importing and Exporting Trees and Data ee eeeee cree 149 Examining Current Status ui eee 152 Data Mato sd de del Noah eS 152 CHALETS A A EA Se 152 User Defined Character TYPES oooonooccnnoccconocccononacinancnonnos 153 Ancestral States ANCSTATES cccccccccesssceeessseeees 153 WSS OVALS ineno a A E 154 CHAPTER 3 COMMAND REFERENCE ocococococccoccnocnnoonnocnoconononnnconncconocancnnancconnc ns 155 Command Ie A N 155 POSIT CTS ai E RR 156 Taxon identidad 157 Character idoneidad 157 Other Hames ie oere na a R SS 158 Common Command Elements ooooconnnoccconcccnonnnnnnnnccnnnnno 158 Taxon ASS A Ns oa 158 Characteristics 158 CHATACIED E A 159 Tres lt ninia 159 Commands used in the Data Block o ooooooccccnonoccccnononcnnnonnns 159 CHAARLABE ES io is 159 DIMENSIONS oa uan 160 FORMA Tiana 160 MATRIX 2555 id 162 OPTION Ss id 162 STATE A A out caetracies cna RN 163 TAXLABELS traia dani 164 Commands used in the ASSUMPTIONS Block 164 ANESTA TES tias 164 CHARS E dni 165 EXSE E icaiaa ara a a T RTE A ee 165 OPTION Sansa 165 Md A igure a cna a T a i 166 RUS S AT A e E E o 166 WISE ina n a a a 166 Commands used in the TREES Block oooonocccnnocccinn
276. ou want to use any other symbols to designate character states you must explicitly define an alternate SYMBOLS list The standard symbols for molecular sequence types DNA RNA PROTEIN are given in the section Predefined formats for molecular sequence data Chapter 2 The format of a symbols list is a sequence of single character symbols the entire list is then enclosed within double quotes 172 PAUP 3 1 USER S MANUAL INTERLEAVE If INTERLEAVE is not specified the character state data are entered sequentially by taxon i e all character state data for the first taxon are entered before beginning the second taxon and so on Specification of INTERLEAVE allows the data to be entered in blocks of characters This format is often used for nucleotide and amino acid sequence alignments but may be useful in other contexts as well MATCHCHAR match symbol If a match symbol is specified any occurrence of that symbol in the data matrix is translated to the state or state set occurring in the first row of the matrix EQUATE symbol expansion The EQUATE option provides a simple macro facility for translating character state specifications in the data matrix to alternate character state specifications The symbol component must be a single character Expansion is either a valid character state or a character state set Any number of EQUATE macros may be specified following the equal sign but only one pair of double quotes
277. owever we must usually accept the consequences of that decision for the remainder of the stepwise addition process perhaps ending up in a local optimum as a result PAUP provides one option that attempts to deal with excessive greediness If the number of trees held at each step HOLD is set to n the n shortest trees from each step are considered in turn during the next step Note that n trees are retained even if some are longer than others For example we might find in a given step trees of lengths 25 25 25 26 27 27 and 28 The 26 step tree may ultimately be the best choice for retention but if we saved only one of the 25 step trees then it would be discarded prematurely If HOLD were 4 or greater however the 26 step tree would be retained and would eventually rise to the top Setting HOLD gt 1 is also useful for minimizing the effect of ties early in the stepwise addition process since ties are to some extent followed out rather than being broken arbitrarily In order to determine the length of the tree that would result if a taxon were to connect to the tree at a particular point PAUP evaluates the minimal length of the full tree given that placement Some algorithmic tricks are used to minimize redundant calculations however Note that this approach differs considerably from that of Farris 1970 In his method character state assignments made during each step are retained for all subsequent steps These state assignm
278. parts of islands Steepest descent The first round of swapping begins with the trees s in memory These can come from several sources user defined trees an external treefile a TREES block in a data file or from stepwise addition In the simplest case PAUP will swap only on minimal trees although this can be changed In that case the first round begins by discarding all nonminimal trees and swapping on the first of the minimal trees For example if we start with five trees in memory of length 30 30 30 31 and 32 PAUP will begin by discarding the two trees longer than 30 steps It will then begin swapping on one of the 30 step trees If no trees shorter than 30 steps are found it will go on to swap on another 30 step tree If STEEPEST DESCENT is off when a tree of less than 30 steps say 29 steps is found at any point all 30 step trees from that round are discarded That 29 step tree is now the input for the next round of swapping Subsequent rounds begin each time a shorter tree is found The process terminates when swapping on all starting trees from a previous round does not find any shorter trees PAUP 3 1 USER S MANUAL However if the STEEPEST DESCENT option is on the round is not abandoned when a shorter tree is found but continues until all trees from the previous round have been examined In the example above this means that the round would not end as soon as a 29 step tree was found PAUP would swap on all of the in
279. pecifies a class interval for the frequency distribution of tree lengths By default HINTERVAL 1 so that the number of trees at each length is output If HINTERVAL gt 1 then adjacent tree length values are pooled into tree length classes of width interval value SAVEFD If SAVEFD is specified the data for the frequency distribution of tree lengths are saved to a text file for input to other programs The SAVEFD option is not persistent 184 PAUP 3 1 USER S MANUAL FILE file specification File specification specifies the name of the file to receive the frequency distribution ignored unless SAVEFD is specified If SAVEEFD is specified and FILE is not specified a default file name is used If file specification contains any of the characters equal sign semicolon colon or blank it must be enclosed within single quotes Explicit specification of FILE also implies SAVEFD NO REPLACE Ordinarily if SAVEFD is requested and the specified file see FILE option above already exists you will be prompted for confirmation that the existing file should be replaced REPLACE suppresses this prompt the existing file will be quietly overwritten by the new data ANCSTATES The ANCSTATES command used to define ancestral states is ordinarily issued from within the ASSUMPTIONS block You may also issue it from the command line or from within a PAUP block See Commands used in the ASSUMPTIONS Block for the description of this
280. plicitly The drawback to this approach is that unless the default initial seed is explicitly overridden exactly the same analysis will be performed in any two runs of the program with identical command sets For example conducting a second random addition sequence search on a second set of 64 PAUP 3 1 USER S MANUAL 100 replicates will produce exactly the same set of trees as an initial set of 100 replicates if the program is restarted between the two searches unless you override the default initial seed Keep this in mind PAUP 3 1 USER S MANUAL 65 Chapter USING PAUP This chapter provides general information on the use and features of PAUP For details refer to the COMMAND REFERENCE and INTERFACE chapters This chapter is intended to be implementation independent e it pertains to all existing implementations of PAUP Version 3 including the Macintosh IBM PC and mainframe minicomputer versions Features and capabilities available only under specific implentations are described in the INTERFACE chapter INPUT FILES Input files for PAUP are standard text files containing commands and or data These files will usually adhere to the NEXUS format described below Files must be standard text files For example on the Macintosh input files can be created using any editor that can output files of type TEXT 1 e word processors spreadsheet programs and text editors MacClade 3 0 and later users can use MacClade s integ
281. ply groups of characters that can be referred to by a single name in other commands The syntax is CHARSET character set name character list 176 PAUP 3 1 USER S MANUAL The character set name must not be identical to any of the original character names m See Simplifying Input with Sets character sets in Chapter 2 for examples on the use of EXSET Use the EXSET command to define an exclusion set The syntax is EXSET exclusion set name character list If you precede the exclusion set name with an asterisk any previously excluded characters are re included and the characters specified by character list are excluded Otherwise the exclusion set is simply defined for later use by the ASSUME command or the Include Exclude Characters and Choose Assumption Sets menu command m See Simplifying Input with Sets use exclusion sets in Chapter 2 for information on how to OPTIONS Use the OPTIONS command of the ASSUMPTIONS block to specify certain assumptions related options recognized by PAUP and by other NEXUS conforming programs The syntax is OPTIONS option specification option specification The only option used by PAUP is the following DEFTYPE character type The character type specifies the character type that applies to characters not explicitly assigned a type by CTYPE and or CHARSET commands It must be one of the standard types ORD UNORD DOLLO DOLLO UP DOLLO DN IRREV IRREV UP
282. point for the already chosen taxon In the closest addition sequence however every unplaced taxon must be connected to the tree at every possible branch In addition each of the C T 3 possible triplets of taxa must be evaluated at the Start 3 Simple This option corresponds to the order in which taxa are connected in the simple algorithm of Farris 1970 As for as is an addition sequence is determined prior to beginning the stepwise addition process however a more elaborate criterion for determining the sequence is used First the distance between each taxon and a reference taxon is calculated Farris called this distance an advancement index The taxa are then added in order of increasing advancement That is the reference taxon and the two taxa closest to it form the initial three taxon tree and the remaining taxa are added in the order given by their rank in the array of advancement indices In Farris 1970 the reference taxon represented a hypothetical ancestor possessing the assumed ancestral state for each character However the algorithm can be used with any taxon chosen as the reference 4 Random A pseudorandom number generator is used to obtain a permutation of the taxa to be used as the addition sequence When ties occur under the simple and closest addition sequences they are broken arbitrarily Unfortunately no one strategy seems to work best for all data sets the best approach is to try as many altern
283. ppending a suffix to the IRREV keyword If no suffix is provided IRREV then up is assumed TRREV UP and IRREV DN can be used to explicitly request the up and down options respectively In versions of PAUP that provide a menu mode pop up menus in the Set Character Types dialog box may be used to choose one of the options For further information on assigning polarities from the menus in the Macintosh and IBM PC versions see Chapter 4 For further information on defining and choosing ancestors see Defining Ancestral States elsewhere in this chapter Defining Your Own Character Types In addition to providing the standard character types mentioned above PAUP allows you to define your own character types via the USERTYPE command Two kinds of user defined character types are available Character state trees allow you to define a character state graph that specifies a linear or branching relationship among the character states Stepmatrices assign a cost for the transformation from every state to every other state See Character Types in Chapter 1 for more discussion on potential uses of user defined character types Character state trees Character state trees are described using a parenthetical notation that defines the shape of the tree The nodes of the character state tree are labeled by symbols corresponding to the character state symbols used in the data matrix Obviously all of the states observed
284. ppresses this status output PAUP 3 1 USER S MANUAL 181 Tree rooting options ROOT OUTGROUP LUNDBERG MIDPOINT The ROOT option is used to specify how unrooted trees are to be rooted prior to output You can choose OUTGROUP rooting using whichever outgroup you have selected MIDPOINT rooting which roots the tree at its midpoint or LUNDBERG rooting which requires that a previous ANCSTATES command has been issued By default OUTGROUP rooting is in effect OUTROOT POLYTOMY PARAPHYL MONOPHYL If OUTGROUP rooting is selected there are three options for output The outgroup can make up a polytomy next to the ingroup POLYTOMY the default or it can be made to be paraphyletic relative to the ingroup PARAPHYL or the monophyletic sister group to the ingroup MONOPHYL Tree output options NO TCOMPRESS Specify COMPRESS to output tree diagrams in a vertically compressed format The resulting diagram is not as aesthetically appealing but it allows more of a large tree to be seen on one screen or in one window and it takes less paper to print Options for character matrix listings NO SHOWIGNORE Unless SHOWIGNORE is specified ignored characters are not shown in character matrix listings For information on ignoring characters see Using a subset of the characters in Chapter 2 NO CMLABELS By default character names are used to label the columns of character matrix listings If you want to use numbe
285. provides a facility for excluding characters from consideration when computing tree lengths Although the excluded characters will not be allowed to influence the selection of optimal trees they will still be optimized to these trees For example you can construct trees using only a subset of the characters but still see how the remaining characters would be mapped onto the resulting tree s A dialog box with two scrollable lists will appear PAUP 3 1 USER S MANUAL 247 Included characters Excluded characters 5 a o Cancel ox 1 2 3 4 6 7 8 9 1 The Include Exclude Characters dialog box The left list contains the included characters and the right list contains the characters that are to be excluded Select the characters you want to include or exclude and press the appropriate arrow button to move characters from one list to the other You can also double click on a taxon to move it from one list to the other Click on OK when you are finished deleting and restoring taxa If you have defined character sets you can quickly select all of the characters contained in the set by pulling down the appropriate Charsets menu If you have defined exclusion sets you can immediately set the exclusion status of all characters by choosing an item from the Exsets menu Set Character Types Use this command to change the type of one or more characters Character Type Selection Characters
286. psing is accomplished The syntax is CONDENSE options Two options are available NO COLLAPSE Unless NOCOLLAPSE is specified branches whose maximum possible length is zero are collapsed to yield polytomies COLLAPSE reverses the effect of a previous NOCOLLAPSE DELDUPES Unless DELDUPES is specified duplicate trees will be eliminated CONSTRAINTS Use the CONSTRAINTS command to define a constraint tree The syntax is CONSTRAINTS constraint tree name tree specification The tree specification must follow the format described under User Defined Trees in Chapter 2 CONTREE Use the CONTREE command to request computation of strict semistrict combinable component Adams and or majority rule consensus trees see Consensus Trees in Chapter 1 The syntax is CONTREE tree list options The tree list specifies which trees to include in the consensus the default is ALL The following options are available NO STRICT By default a strict consensus tree is computed Use NOSTRICT to suppress this computation PAUP 3 1 USER S MANUAL 189 NO SEMISTRICT Specify SEMISTRICT to request computation of a semistrict combinable component consensus tree NOSEMISTRICT reverses the effect of a previous SEMISTRICT specification NO MAJRULE Specify MAJRULE to request computation of a semistrict combinable component consensus tree NOMAJRULE reverses the effect of a previous MAJRULE specification
287. ption SetS ooooooccnnoncconocnnononaconnnnnnonccnnonanos 236 The Seareh Menu ini adi 237 Load Constraints oooooncncnnncnnononnnanconncnnononananicicncnnanononess 237 A A Dagar ES 238 PACUERS CE cis 239 Branelva d Bound ra 241 EXDAUS VE in 242 Lake aran o a 243 o A A TN 243 IANO TES cP E E E E EE 244 The Trees Men eiii in deals 245 Tree Oia lonettadatonsaes 245 Clear e Sana 245 Condense Tres rte osas 245 Root Trees Deroot TreesS ooocoonoocccnonccononcnonnnccnnnnccinnnnos 246 Tre to Tree DISCS ie 247 Lengths and Fit Measures cocoocccnonconcnoncnononaninnnaninnnccnnnnnos 247 IGE ATCES A IR E 248 Remove Filter usina isis 249 Reverse Al es ao le 249 Show Trees ditat Ge Goal este ancesansaban a a e at 249 Describe Trees areg o a o a e 249 Show ReconstructionS cooooocccnonccononccononacononanonnncnnnnnanonns 250 PUMAS do as 251 Computo COSER iS 253 PEA a a aa 254 Get Trees from Flint told 254 Save A eee ated nalsancen E 256 REFERENCES isisisicatistcinis ji tailt 259 PRELIMINARIES This manual is for version 3 1 of PAUP for the Apple Macintosh It also serves as preliminary documentation for the IBM PC and Portable mainframe and workstation releases of Version 3 which as of this writing have still not been released The manual is not as polished as I had hoped it would be at this point Hopefully the weaker sections will be enhanced by the time version 3 2 is released I apologize for not including
288. r terminal taxa an order may be imposed based on other criteria purely for output purposes Trees are also classified according to the way in which nodes are labeled In phylogenetic analysis trees are usually considered to be terminally labeled That is the terminal nodes correspond exactly to the biological taxa under study but the labeling of the internal nodes is arbitrary If all internal nodes are associated with actual objects e g fossil taxa and are not merely hypothetical constructs the trees are completely labeled Note that although some programs e g MacClade may permit actual taxa to occupy ancestral positions PAUP considers only terminally labeled trees if a taxon is assigned to an internal node e g in a user specified tree description it is popped out to a terminal position Examples of the types of trees discussed above are shown below PAUP 3 1 USER S MANUAL A BCD E C B A D E c A D i XY aC a b c A B C D E A B C D E NY 7 A D ed lt lt Ye C E l R d e 1 A B C D E A BG D E A B C D E G Pa R 9 h i Examples for tree terminology a Unrooted binary tree b Rooted binary tree rooted at an internal node of degree 2 c Another rooted binary tree Trees b and c are equivalent as unordered trees but distinct as ordered trees d Binary tree rooted at terminal taxon R e Completely labeled rooted binary tree f Unrooted nonbinary tree g Rooted nonbinary tree h Pa
289. r interest is not monophyletic on the minimal trees for your data set constraining a search so that only those trees consistent with the group s monophyly are retained makes it easy to determine how much longer are the shortest trees on which the group is monophyletic For example if the most parsimonious tree for your data set does not support the monophyly of a genus according to someone else s classification established on the basis of other data you might be less inclined to challenge that classification if the genus were monophyletic on a tree only one step longer than if 28 additional steps were required for the group s monophyly This notion can be extended to testing your data against an entire classification rather than only a single group Since established classifications are usually not fully resolved you can constrain the search so that only trees that are consistent with that classification are retained Thus where the classification is unambiguous the relationships it implies must be maintained before a trial tree is accepted However other aspects of the classification can be resolved in the way that is most parsimonious for the data at hand Before the topological constraints feature was implemented in PAUP users could force the monophyly of particular groups by including dummy synapomorphies in the data matrix and weighting them so heavily that any tree on which the group was not monophyletic was immediately rejected This
290. r internal documentation of the data file and is otherwise completely invisible during the processing of the file If the comment begins with as in This is a visible comment the comment is visible NEXUS programs are free to treat visible comments in any appropriate manner PAUP simply echoes them to the output destinations display window output file and or printer exactly as they appear in the file Although not a requirement it is recommended that every NEXUS file begin with a visible comment that at least briefly describes the contents of the file e g source of data date entered etc Note to MacClade Users If you use PAUP to manually create a NEXUS file for input to MacClade note that style comments are visible only if they are contained within a DATA TAXA or TREES block The following sections provide a brief description of the major elements of the NEXUS format A more formal description will be published elsewhere D Maddison Swofford and W Maddison in prep The DATA Block The DATA block contains the data matrix and other associated information Its syntax is 68 PAUP 3 1 USER S MANUAL BEGIN DATA block name DIMENSIONS NTA X number of taxa NCHAR number of characters FORMAT MISSING missing symbol LABELPOS LEFT RIGHT SYMBOLS symbols list INTERLEAVE MATCHCHAR match symbol EQUATE lt symbol expansion gt TRANSPOSE RESPECTCASE
291. r of random trees to be evaluated The default is NREPS 1000 HINTERVAL interval value SAVEFD FILE file specification NO REPLACE NO INCLUDEANC ANCSTATES ancestral states name These options have the same meaning as in the ALLTREES command see above RESTORE Use the RESTORE command to restore currently deleted taxa for subsequent analyses see Deleting and Restoring Taxa in Chapter 2 The syntax is 210 PAUP 3 1 USER S MANUAL RESTORE taxon list ONLY Unless ONLY is specified taxa specified in the taxon list are simply removed from the set of currently deleted taxa Specify ONLY to delete in addition any taxon not explicitly specified in the list REVFILTER Use the REVFILTER command to reverse the effect of the current filter All trees that are currently hidden by the filter will become visible and all trees that were previously visible will be hidden There are no options REWEIGHT Use the REWEIGHT command to assign weights to the characters based on their fit to the trees currently in memory see Successive weighting in Chapter 2 The syntax is REWEIGHT options The following options are available BASEWT base weight Base weight specifies the maximum possible weight that a character can be assigned corresponding to an index value see below of 1 Weights are scaled from 0 to this value The default is BASEWT 1000 INDEX RC ICIIRI The INDEX option is used to spe
292. r of times pair supported Number of times pair rejected 0 005 Number of times pair supported Number of times pair rejected 0 001 Number of times pair supported Number of times pair rejected These results show that the human chimp gorilla and human chimp groupings receives some support at P 0 01 but only from one quartet in each case The most useful approach for more than four taxa is to divide the taxa into four groups and evaluate all quartets consistent with this arrangement For example if we wanted to assume monophyly of chimp gorilla human but use both orang and gibbon as outgroups to resolve the trichotomy we could use the command lake mode fourgrps sumtabonly grpa 1 grpb 2 grpc 3 grpd 4 5 which generates the following output PAUP 3 1 USER S MANUAL All output other than summary table suppressed Evaluating quartets with one member from each of the following groups Group A human Group B chi mp Group C gorilla Group D orang gi bbon 2 quartets to be evaluated ey Es ahd G trees E Group A Group B Group C Group D F Group A Group C Group B Group D G Group A Group D Group B Group C Summary of results from Lake s method E F G Lake s method Times tree favored 25 Total counts favoring tree 6 Standard parsimony Times tree favored de Total counts favoring tree 35 2 28 Transversion parsimony Times tree favored Lo Total counts favoring
293. rated spreadsheet editor to create files for input to either PAUP or MacClade The NEXUS Format The NEXUS format was designed by David Maddison Wayne Maddison and David Swofford to facilitate the interchange of input files between programs used in phylogeny and classification Data files that conform strictly to the NEXUS guidelines can be input to any program that fully supports the NEXUS standards Currently these programs include PAUP MacClade Maddison and Maddison 1992 and COMPONENT Page 1993 Among other things this means that data files created using MacClade s spreadsheet editor can be input to PAUP and PAUP tree and data files can be input to MacClade for further analysis The book accompanying the MacClade program has detailed instructions on the simultaneous use of PAUP and MacClade 66 PAUP 3 1 USER S MANUAL Blocks The characteristic of NEXUS files that allows them to be so portable is the block concept A block is a well defined subsection of an input file that can either be read or ignored by any NEXUS conforming program The format of three blocks DATA ASSUMPTIONS and TREES is defined by the standard any conforming program must either ignore these blocks entirely or be able to process commands in the block as defined by the standard Other blocks are program specific If a program encounters a block that it either does not recognize or does not want to deal with it simply skips over the entire bloc
294. re approximation will be used to test the significance of deviations of invariant scores from 0 if the number of informative positions is less than n If MODE CHOOSE4 you must specify exactly four taxa using the following option TAXA taxon list The taxon list must contain exactly four taxa PAUP 3 1 USER S MANUAL 205 If MODE ALLQUART you may limit the number of included taxa by selecting at least four taxa using the following option TAXA taxon list If MODE FOURGRPS you must assign at least one taxon to each if the four groups using the following options GRPA taxon list for first group GRPB taxon list for second group GRPC taxon list for third group GRPD taxon list for fourth group LEAVE Use the LEAVE command to terminate processing of an input file Ordinarily PAUP continues processing until the end of an input file is reached If for whatever reasons you do not want some of the commands in the input file to be processed insert a LEAVE command into the file at the point where you want execution to stop PAUP then continues as if the end of the file were reached at that point LEAVE has no effect if issued interactively LENFIT Use the LENFIT command to request a listing of tree lengths and or fit measures for one or more trees The syntax is LENFIT tree list options If no tree list is specified ALL is assumed By default only tree lengths for single characters overall or bot
295. read the Mal tartas sd ads Rin v Technical SUPP OTI hae Gees eRe eee v Anonymous FTP Server for PAUP Suppott ooocccnocccinoccnononccnnnnnss vi CHAPTER 1 BACKGROUND utili 1 General Conce piStssnindi analista cicldades coco ali dit 1 Tree Terms a dia 4 Character TT DES A ea A AS 6 Ordered Wagner Characters ooonoccnocococcconcnnnancnnncconcnnnnss 7 Unordered Characters tic 8 Dollo Characters siii iii lar 8 Rooted vs unrooted Dollo models 8 Polarity SpecilicatiO Ma tosirnginai td 12 Irreversible Characters caco 12 Polarity SpecificatiON ooooccononcnoocnnonnconnnnnnncon anno 12 User Defined Character TyP8S ooooooccnnoccconocccononaninnncnnnnos 13 Character state tees 25 disii iios 13 SEPAIS eas oo ee ees 15 Tree Lengths and Character WelghtS oonooonnnninninoncnonnnnncncnnncnnos 18 Character State OptiMIZatlON oonooccnoccconnnonnnonnncnnncconccnnanona nono no ns 19 Missing Data tits 21 Outgroups Ancestors and RO0fS ooooconocccnoccnocncooncconncnoncnan ccoo nono 24 Searching for Optimal Trees ion lic 27 Exact methods ii at 28 EXA ra econ 28 Branch and bound algorithM ooonncnnnnnnnnnnnnncnm 29 Heuristic Methods sinana nana n aa 32 Stepwise Addis 32 Branch Wap giste siest iyisi iee acas 36 Searching under Topological Constraints 40 Monophyly constraint trees ooooonoccnnncnnocnnonncnnss 40 Backbone constraint trees ooooocnnnncccnonncionncnnnnos 43 Heuristic Searches and c
296. ree and G tree for four taxa and calculates which tree is statistically significant We will not describe the method here introductory presentations can be found in Holmquist et al 1988 and Swofford and Olsen 1990 Because the method deals with only four taxa a quartet at a time PAUP has three options for handling more than four taxa In that case you can choose to evaluate a specified subset of four taxa evaluate all possible quartets or partition the taxa into four groups where the quartet is made up of one taxon from each of the four groups all possible quartets of this type being evaluated PSEUDORANDOM NUMBER GENERATION A few of the capabilities provided by PAUP require generation of sequences of random numbers i e random addition sequence in stepwise addition sampling of characters in bootstrap analysis and evaluation of the lengths of random trees PAUP uses a linear congruential method starting with any number x between 1 and 2 147 483 646 inclusive the next number in the sequence is given by Xi 1 397 204 094 x mod 23 1 see Fishman and Moore 1982 The first number represents the seed PAUP always uses 1 as the initial seed rather than an arbitrary value such as one obtained from the system clock The main reason for this decision was so that runs from the same data file would always generate identical results on any kind of computer even if the user neglected to define the initial seed ex
297. ree description format is the same as for any other user defined tree See the section User defined trees for details of tree description An equivalent procedure is to use the Load Constraints menu command This is quite useful any tree in that has been previously saved may potentially be loaded as a constraint tree These may include trees produced under different search conditions using the present data set or trees from other studies of the same taxa MacClade s tree manipulation interface is particularly useful in constructing and defining different constraint trees In PAUP the command either on the command line or in PAUP block for defining a monophyly constraint tree called yourname would look like CONSTRAINTS yourname tree specification 156 PAUP 3 1 USER S MANUAL where the tree specification would follow the rules for user defined trees above Constraint trees are at least partially unresolved otherwise only one tree would match the constraint For example the command constraints he human chimp gorilla orangutan gibbon defines an unresolved tree on which the only clade other than the trivial all species clade is a group containing human chimp For monophyly constraints you do not necessarily have to enter all taxa into the tree definition This is because omitted taxa are joined to the root node of the subtree described by the tree specification for the included taxa For example to force the gro
298. rees that must actually be evaluated If we reach the end of a path on the search tree and obtain a tree whose length is less than or equal to the upper bound L then this tree is a candidate for optimality If however this length is less than L then this tree is the best one found so far and we have improved the upper bound on the length of the optimal tree s This bound improvement is important because it may enable other search paths to be terminated more quickly When the entire search tree has been traversed all optimal trees will have been identified Several factors influence the running time of the branch and bound algorithm The quality of the data is perhaps the most important factor large data sets with little homoplasy will run quickly because most paths of the search tree are terminated early The speed with which the length of each tree can be evaluated a function of the character types is also important In general undirected character types run faster than directed types because certain algorithmic tricks for rapidly computing tree lengths can be used when the tree length does not depend on the location of the root Ordered Wagner characters are much faster than unordered characters for similar reasons User defined stepmatrix characters on the other hand run very slowly due to the enormous amount of computation required to compute the length of even a single tree Finally of course the speed of the available computer
299. remain deleted when a new DELETE command is issued If you want only the taxa specified a single DELETE command to be deleted you can use the ONLY option delete 6 8 11 only This is equivalent to the pair of commands restore all delete 6 8 11 You can also use the Delete Restore Taxa menu command to delete and or restore taxa see Chapter 4 for details DISTANCE MATRICES A matrix of pairwise distances between taxa can be computed using the SHOWDIST command or the Show Distance Matrix menu command Two distances are computed for each pair of taxa i and j 1 The absolute distance 106 PAUP 3 1 USER S MANUAL d i j gt wi AUN XikoXjk gt keS where is the set of characters that are not ignored or excluded wk is the weight currently assigned to character k xik and xjk are the states of character k in taxa i and j respectively and diff xik xjk is the cost of a change from state xjx to state xjk The costs specified by diff y z are determined as follows If either y or z is missing or if both y and z are missing diffly z 0 Otherwise for unordered characters diff y z 1 if y z 0 otherwise for ordered Dollo and irreversible characters diffly z ly for user defined character state trees diff y z the number of branches lying on the path of the character state tree connecting states y and z for stepmatrix characters diffly z Syz the corresponding element of the applic
300. rent ancestor and its states For example PAUP 3 1 USER S MANUAL 163 Character states for current ancestor new tod Le Daal 1234567890123 45 6 new 72227272727 000000000 The status lines are omitted if SET NOCMSTATUS is used or this output is suppressed in the Character Matrix Format dialog box Tree Status To display the status of trees in memory use the TSTATUS command or the Tree Info menu command This will display the number of trees in memory and their source For example 3 unrooted trees in memory No tree filter in effect Source of trees Heuristic search If trees in a file are added to trees in memory PAUP will identify treefile as the source for all trees in memory PAUP 3 1 USER S MANUAL 165 Chapter COMMAND REFERENCE This chapter describes the PAUP command format Unless stated otherwise these commands may be included in the PAUP block of a NEXUS file or typed from the command line The EXECUTE and EDIT commands may only be issued from the command line All users will need to be familiar with the commands available in the DATA ASSUMPTIONS and TREES blocks which are normally placed in PAUP input files In versions of PAUP that provide a menu based interface only knowledge of these block commands is required as other instructions to PAUP can be accomplished through the menu system Some users however may prefer a command driven interface Consequently except for certain machine specific features
301. representing more derived states For instance state C could instead have been chosen as the ancestral state in the example above The character state graph would then be See Assigning Character Polarities under Specifying Character Types in Chapter 2 for information on how to designate character polarities NOTE Felsenstein 1984 has described a variation on irreversible character parsimony that he calls the unordered Camin Sokal method analogous to his unordered Dollo method discussed above Again unpolarized Camin Sokal would have been a more appropriate name As in the unordered Dollo method the unpolarized Camin Sokal method evaluates each character on a given tree under all possible polarity assignments and chooses that polarity which allows the minimum number of changes It seems exceedingly unlikely that one would be willing to assume that a character was irreversible but ignorant as to its polarity so I have not implemented unpolarized irreversibility in PAUP User Defined Character Types Two kinds of user defined character types are available Stepmatrices allow you to define arbitrary character types by specifying the distance from any character state to any other Character state trees provide an easy method for inputting branching character transformation series Character state trees Character state trees are used to specify a non linear branching relationship among character states which are then otherw
302. requency distribution plot i O Include ancestor The Random Trees dialog box This distribution can also be output to a file 1f you select Save tree length frequency distribution to file You can choose the random seed the PAUP presents or one of your own Usually the box Include ancestor will be unchecked unless you have defined an ancestor separate from the standard all missing ancestor If not all rootings will equally parsimonious and will be saved as distinct trees THE TREES MENU mii The Trees menu contains items for outputting trees Clear Trees and associated information and for performing Condense Trees os Root Trees other tree related activities Tree to Tree Distances Lengths and Fit Measures Filter Trees Remove Filter Reverse Filter Show Trees Describe Trees Show Reconstructions Print Trees Compute Consensus Print Consensus The Trees menu Get Trees from File Save Trees to File Tree Info The Tree Info command provides a quick summary of the number of trees currently in memory and how they were obtained Clear Trees This option will clear all trees in memory If it is selected the following warning will appear PAUP 3 1 USER S MANUAL 261 Are you sure you want to clear all B trees from memory No __ves ___ The Clear Trees warning box Condense Trees This command allows you to retroactively collapse
303. ription perform the following operations as you make the circuit When you leave a node traveling away from the root toward the leftmost descendant write a left parenthesis e When you leave a node traveling away from the root toward any descendant other than the leftmost descendant write a comma e When you leave the rightmost descendant of an internal node traveling toward the root write a right parenthesis 154 PAUP 3 1 USER S MANUAL e When you visit a terminal node or visit an interior node for the last time write the node information if any for the node NOTE Usually the node information will consist of a taxon identifier for terminal nodes and nothing for internal nodes You are only entering the relative topology of the tree therefore information about internal node labels and branch lengths is not included In any case branch lengths are dependent on a particular data matrix and are not something that are fixed on a topology When user defined trees are read into memory they can later be output with node labels and branch lengths although if they are output to a treefile only the branching topology is preserved See the sections on Diagnosing Trees and Saving Trees to Files for descriptions of how trees are described and manipulated Applying these rules to the example of above the tree description would develop as follows Number in Sequence 0 nb WN ke 10 11 12 13 14 15 16
304. ront See The PAUP Editor for more information on PAUP s editing capabilities THE OPTIONS MENU options The Options menu is used to control a number of Multistate Taxa Optimization option settings that affect PAUP in various ways Set Mantrees Character Matrix Format Searching Rooting Tree Order Stepmatrices Ignore Characters Semigraphics Editor Warnings Errors The Options menu NEHUS Format Startup Preferences Restore Option Settings Multistate Taxa This command allows you to specify how PAUP treats multistate taxa when computing tree lengths Multistate taxa are those for which more than one character state was assigned to taxa in the data matrix for some characters see the section Multistate Taxa in Chapter 2 When one or more taxa are coded as having multiple states interpret multiple states as uncertainty polymorphism The Multistate Taxa dialog box 238 PAUP 3 1 USER S MANUAL When multiple states are interpreted as uncertainty PAUP will choose a state from the set of available states that allows minimization of the tree length When multiple states are treated as polymorphism PAUP assumes that the terminal taxon is actually a heterogeneous group In this case all but one of the states in the polymorphic terminal taxon must be derived from a monomorphic ancestral taxon in the most parsimonious way p
305. root which means that it may be invoked only when ordered Wagner or unordered characters are in use Other character types Dollo irreversible stepmatrix cannot be used with MINF optimization hence are not allowed It is possible that a particular ancestor may have multiple equal length attachment sites which leads to the following output Note Ties found in Lundberg rooting The following root locations are equally parsimonious 9 lt gt 10 14 lt gt taxonA Tree length 23 Consistency index Cl 0 609 Homoplasy index HI 0 391 Retention index RI 0 591 Rescaled consistency index RC 0 360 fleccccccecce taxondA fleccccceecldcccccccccee taxonB Marcar 1er erre erereees taxonC Ae TIE eee EEE EE EEE EE cerca taxonF Ae TLC CEE EEE EEE EEE EEE EEE EEE ELE EE EEE EE EEE taxonE AK ETD EEE EEE EEE EEE EEE EEE EEE EEE EEE EE EEE LEE EE ELE EEE taxonD fleccccccecce taxonG FE EEE EEE EEE EE EEE EE EEE EEE EEE EEE EEE EE LEE EEOC EEE taxonH You can see that PAUP only prints one of the trees but you may wish to recover all equally parsimonious trees for further analysis To recover all the multiple roots for a optimal trees you must include an ancestor taxon in the matrix with the same states as were in the original ANCSTATES definition but do not change the ANCSTATES settings Then exclude the ancestor and calculate all optimal trees for the ingroup Next save the trees to a file Restore the ancest
306. roup The syntax is INGROUP taxon list ONLY Unless ONLY is specified taxa specified in the taxon list are simply removed from the current outgroup If you want only those taxa specified in taxon list to be included in the ingroup specify ONLY any taxon that is not explicitly specified in the list will be transferred to the outgroup 204 PAUP 3 1 USER S MANUAL LAKE Use the LAKE command to perform an analysis using Lake s 1987b method of linear invariants This command is available only when DATATYPE DNA or DATATYPE RNA The syntax is LAKE options The following options are available MODE CHOOSE4 ALLQUART FOURGRPS If there are more than 4 non deleted taxa in the data set this option specifies whether you want to choose 4 taxa for analysis MODE CHOOSE4 analyze all possible quartets MODE ALLQUART or divide the taxa into four groups and analyze all quartets containing one member from each group MODE FOURGRPS NO SPECTDIST SPECTDIST requests the output of the spectral distribution the number of positions falling into each of the 36 possible patterns considered informative by Lake s method NOJBRLEN BRLEN requests output of branch lengths calculated by Lake s operator metrics Lake 1987a NO SUMTABONLY SUMTABONLY limits the output to a summary table of results rather than outputting results for each quartet EXACTN n An exact binomial test rather than the chi squa
307. rs 2 FROMTREE and lt TOTREE are input to branch swapping FROMTREE defaults to 1 and TOTREE defaults to the number of trees in memory so that ordinarily all trees are swapped on These options are meaningful only if NOSTEPWISE is specified NOJUSENONMIN If NOSTEPWISE or HOLD gt is requested it is possible that the trees in memory at the time branch swapping begins are not all equal in length Ordinarily only the shortest available trees are input to the branch swapping procedure If you want to swap on nonminimal trees as well specify USENONMIN NOUSENONMIN reverses the effect of a previous USENONMIN specification NO STEEPEST Specify STEEPEST to request use of the steepest descent modification to the branch swapping procedure see the section on Branch Swapping NOSTEEPEST reverses the effect of a previous STEEPEST specification NCHUCK maximum number CHUCKLEN tree length for chucking If this pair of options is used no more than maximum number of trees of length greater than or equal to tree length for chucking will be retained in a search or in a random addition sequence replicate See Heuristic Methods in Chapter 2 for more information on these settings NOJABORTREP If ABORTREP is requested and NCHUCK and CHUCKLEN values have been specified the current random addition sequence replicate will be aborted if the chucking limits are hit PAUP 3 1 USER S MANUAL 203 RETAIN number of trees to retain
308. rs even when character names are available specify NOCMLABELS NO CMCSTATUS If CMCSTATUS is specified characters that are constant zapped uninformative or excluded are identified by asterisks at the top of each column of a character matrix listing PAUP 3 1 USER S MANUAL CMDCOLWID column width The value specified for column width determines the number of columns used for each character in the data matrix The default is CMDCOLWID so that one blank column appears between each column of character state data For sequence data you may want to use CMDCOLWID 1 in order to fit more characters onto each line of output NO CMSHOWEOQ Unless CMSHOWEQ has been specified if the possible state assignments to an interior node correspond to a multistate taxon code specified in an EQUATE macro the corresponding EQUATE character is shown rather than the equivalent set of character states Other options OPT ACCTRAN DELTRAN MINF The entry following OPT determines how the characters are optimized on the tree s in memory ACCTRAN default uses accelerated transformation DELTRAN uses delayed transformation while MINF optimizes so as to minimize the f value of Farris 1972 See the section on character state optimization for detailed discussion of these options ANCSTATES ancestral states name Change the ancestor currently in effect to the ancestral states name defined in an earlier ANCSTATES command or to S
309. rs of taxa and characters is 32767 but practical limitations on computing time and available memory will reduce the maximum numbers of taxa and characters that can be accommodated PAUP 3 1 USER S MANUAL 171 FORMAT The FORMAT command is used to specify information pertaining to the format of the data file The syntax is FORMAT option specification Any or all of the following option specifications may be given MISSING missing symbol The missing symbol specifies a character used to represent missing data Any alphabetic numeric or other character that may be used as a character state symbol may be used as the missing symbol If MISSING is not specified it defaults to TRANSPOSE If TRANSPOSE is specified rows of the data matrix correspond to characters and columns correspond to taxa Otherwise rows correspond to taxa and columns to characters LABELPOS LEFT RIGHT If LABELPOS LEFT taxon names or character names if the data matrix is transposed begin each row of the data matrix i e precede the character state data If LABELPOS RIGHT then these names end each row of the data matrix The default is LABELPOS LEFT SYMBOLS symbols list The SYMBOLS list defines a set of permissible symbols that may be used to designate character states The default SYMBOLS list is 01 for the STANDARD data type see below which means that the only non missing character state symbols permitted are 0 and 1 If y
310. rule are allowed PAUP 3 1 USER S MANUAL 135 Possible state assignments for character 3 on tree 1 C er r E Delta Lo F ilo L C Gamma Why are the additional states R arg and stop codon allowed where only C cys and K lys were allowed in the previous example The 3 1 rule takes all possible triplets of observed states and evaluates the length when every state in the stepmatrix is placed at the single internal node Thus there is a 3 component composed of triplets of observed states and a 1 component composed of each of the states in the stepmatrix The figure below shows the results of the 3 1 rule when various are assigned as the 1 component Since in this example there are only three observed character states there is only one triplet C K and to make up the 3 component To begin we measure the length when the members of the triplet the observed states are placed at the internal node This is shown in a and b where the overall length is 5 the length to any state from the or missing state is always 3 Thus for any other state to pass the 3 1 rule the length of the tree must be less than or equal to 5 which is the length obtained when only observed states are allowed If the calculated length from the stepmatrix is more than 5 the state in question fails the 3 1 rule and cannot be assigned to the interior node In c a state is determined to pass the rule length 5 equal to that
311. s for a description of this option With detailed information about the number and lengths of all trees we can better evaluate the strength of evidence for a particular group In the example above we kept all trees 3 steps longer than the optimal tree s However the frequency distribution might show that there are few trees 3 steps longer but many trees 4 steps longer which might also be profitable to examine In this way it can be used in concert with the keep option in exploring suboptimal trees As in the case of evaluating the length of trees which do not meet constraints of monophyly you are still faced with the decision as to how many steps longer a tree must be before the existence of a clade is considered to be insupportable Despite this difficulty it is satisfying to know how long a tree must be before a particular clade becomes nonmonophyletic ZERO LENGTH BRANCHES AND POLYTOMIES As discussed above PAUP s search algorithms find only binary trees When the data are insufficient to determine fully resolved trees trees containing polytomies may be preferable to binary trees containing one or more branches of zero length Dichotomies that collapse into polytomies when zero length interior branches are deleted are said to represent arbitrary resolutions because character support for some of the groupings is absent or ambiguous Usually but not always a smaller number of trees containing one or more polytomies is preferable to
312. s a name for the tree file If you do not explicitly specify a filename a default filename will be used FROM starting tree number If nonzero starting tree number specifies the number of the first tree to save TO ending tree number If nonzero ending tree number specifies the number of the last tree to save NOJROOT If the current trees in memory are unrooted and ROOT is specified trees are rooted using the rooting options currently in effect before they are saved Note that the process of rooting the trees slows down the saving operation considerably If you are saving the trees only with the intention of rereading them into PAUP there is no need to root the trees However if you are exporting the trees to another program rooting them may be desirable The ROOT option has no effect if the trees in memory are already rooted NOROOT can be used to reverse the effect of a previous roor Specification REPLACE APPEND Ordinarily if SAVE is requested and the specified file see FILE option above already exists you will be prompted for confirmation that the existing file should be replaced Explicit specification of REPLACE suppresses this prompt the existing file will be quietly overwritten by the new data Alternatively you may specify APPEND in which case a new TREES block will be concatenated to the end of an existing file The APPEND option is PAUP 3 1 USER S MANUAL 213 SET mainly useful only for arc
313. s an improved tree according to their length under the optimality criterion then we will eventually arrive at the global optimum However if the path to the optimal trees requires us to pass through intermediate trees that are inferior to the best one s yet obtained we will once again find ourselves trapped in a local optimum unless an option is provided for branch swapping on suboptimal trees the KEEP option in PAUP may be used for this purpose A related problem concerns plateaus on the optimality surface It may be the case for example that an optimal tree lies several rearrangements away from the current tree and that these rearrangements all correspond to trees having equal lengths under the optimality criterion If the intermediate trees are discarded because they are not better then the optimal tree will not be found The MULPARS option requests the saving of all of the equally most parsimonious trees Swapping on polytomous trees The rearrangement algorithms above are defined only for dichotomous trees However when the COLLAPSE zero length branches option is in effect trees recovered from memory for input to branch swapping may have one or more polytomous nodes In PAUP 3 1 USER S MANUAL 43 this case PAUP dichotomizes the tree arbitrarily in order to prepare the tree for branch swapping This arbitrary resolution can introduce additional complications Most importantly the resolution of the polytomy that PAUP
314. s and possible state assignments selection in the Describe Trees dialog box or by using the XOUT option for DESCRIBE The output describing the possible state assignments is a matrix of nodes by characters as below The possible states for each character is displayed in a row for each internal node on the tree The output for the actual assignments for a given tree and optimization is also in matrix format Requesting states for terminal taxa with the following command describe 1 xout both results in the printing of the input data matrix as well as the matrix of assigned states as below Data matrix and reconstructed states for internal nodes Node 23456789 Once again the question mark indicates that each available character state appears at that node in at least one reconstruction It does not mean that any character is considered missing at that point Stepmatrix Character Reconstruction Special Considerations When a stepmatrix contains three or more character states it is possible that full minimization of the tree length may require assignments of character states that were not observed in any of the terminal taxa to internal nodes When only a few character states were observed for a stepmatrix character but the stepmatrix is defined for a large number of character states This will often be the case for protein sequences analyzed in
315. s be the same base This is indicated by Transversion parsimony 1133 1313 1331 where 1133 1313 and 1331 correspond to Lake s category codes When there are many more than four taxa to be divided into four groups the use of taxon sets can greatly simplify typing the command or selecting the groups from the lists in the dialog box Just define the groups via TAXSET commands in the input data file and use the taxon set names rather than actual taxon names or numbers in the GRPA GRPB GRPC and GRPD option settings Assessing Confidence using Bootstrap Analysis The search options for the bootstrap are the same as those for either a heuristic or branch and bound search A bootstrap search is stared using the BOOTSTRAP command or the Bootstrap menu command Because 1t randomly samples the data matrix with replacement you must also specify a starting seed for the analysis If you do not 1 is used for the first replicate the number is reset randomly for every subsequent replicate For example the following command will begin a branch and bound bootstrap search with the initial seed set to 12322 ten replicates and a confidence level of 50 bootstrap bseed 12322 nreps 10 method bandb conlevel 50 If you do not specify a starting seed for subsequent runs the seed defaults to the next number in the random number sequence initiated during the previous run You can try different addition sequences and heuristic search options or you c
316. s is the first POSSPLOT command and you do not specify a tree list reconstructions are shown for the first tree only ROOT OUTGROUP LUNDBERG MIDPOINT OUTROOT POLYTOMY PARAPHYL MONOPHYL ANCSTATES ancestral states name NO TCOMPRESS See Options Affecting Multiple Commands earlier in this chapter If no characters are specified for either a CHGPLOT or POSSPLOT command the characters are taken to be those plotted in the last invocation of either of these commands For example CHGPLOT 1 3 5 7 PAUP 3 1 USER S MANUAL 209 POSSPLOT will cause both commands to output information for characters 1 3 5 and 7 QUIT The QUIT command which has no options causes PAUP to terminate RANDTREES Use the RANDTREES command to randomly sample trees from the set of all possible trees and compute their lengths The results are shown in the form of a frequency distribution of tree lengths The syntax is RANDTREES options The following options are available TSEED starting seed Starting seed specifies an integer between 1 and 2147483646 2712 used to seed the random number generator If you do not specify a starting seed for the first random trees analysis 1 is used If you do not specify a starting seed for subsequent runs the seed defaults to the next number in the random number sequence initiated during the previous run NREPS number of replications Number of replications specifies the numbe
317. s potential support i e favored by at least one reconstruction for both However the group A B C D is not retained since neither reconstruction provides even potential character support for it Of course some other character might provide support for the A B C D group and if so the group would be retained Otherwise the tree would collapse as shown in tree G PAUP 3 1 USER S MANUAL 53 A BC ODE A BC DE 0 0 1 1 1 0 0 1 14 4 No M7 Ny MY N SNE a M orF b MO oF A BC ODE A BC ODE 0 0 1 1 1 0 0 1 14 4 NY M7 7 0 1 V c n OF d 0 0 F a E A D 0 Y E A BCD 0 0 1 1 i Na iz 9 V Demonstration of effect of different criteria for collapsing zero length branches a b The two MPRs for a hypothetical character on a given tree heavy lines indicate zero length interior branches c Possible state assignments to interior nodes consistent with at least one MPR heavy lines indicates mandatory zero length interior branch d Reconstruction A after collapsing all zero length interior branches e Reconstruction B after collapsing all zero length interior branches f Tree after collapsing all interior branches that are of zero length under at least one MPR g tree after collapsing all branches that have zero length under all MPRs retaining groups with potential but ambiguous in this case support TREE TO TREE DISTANCES As we discussed in the context of islands in the section on heuristi
318. s than or equal to a certain length 2 trees greater than or equal to a certain length 3 trees that satisfy a particular constraint specification 4 trees that fail to satisfy a particular constraint specification or 5 trees that are either less than or greater than a certain number of symmetric difference distance units from areference tree The filtering criteria are described in detail in the section Filtering Trees 264 PAUP 3 1 USER S MANUAL y Criteria for Filtering Trees O Constraints J Trees lt 2 symmetric difference units from tree m f o Do not retain a polytomous tree for which a more highly resolved compatible tree exists Tree numbers lt gt Tree numbers 2 Options O Retain trees that do NOT match the above criteria O Permanently delete trees not retained by filter Cea Ce The Filter Trees dialog box There are many potential applications for tree filtering For example you may choose to save nonminimal trees during a search via a keep length and then subsequently obtain consensus trees showing the information common to only the shortest trees trees within one step of the shortest and so on Also if a search finds a great many trees you can check via the constraints criterion 1f particular groups occur on all or some of them Finally you can select suites of similar trees using the symmetric difference distance criteria or alternatively
319. s to add an additional set of characters to the data matrix to signify the presence or absence of particular gaps The INTERLEAVE option provides a handy way of adding an additional set of characters to a matrix containing sequence data For example begin data dimensions ntax 4 nchar 24 format datatype dna gap interleave symbols 01 options gapmode missing matrix sequence data One ATGCT ATCCG TCATG A Two ATCCT AGCCG T AG A Three CT T AGCCG T AG TCCTA Four CT A CG A TA AGTCA insertion deletion data One 0001 Two OO ok Three 1010 Four AL 0 r endblock Since the two kinds of characters are in clearly defined blocks you could easily use other PAUP commands to assign different weights to insertion deletion vs substitution events or to compare trees calculated with and without the information from insertion deletion characters EQUATE macros The EQUATE macro facility allows translation of particular character state specifications in the data matrix to alternate character state specifications They are useful in several situations For example you PAUP do not contain any alignment capabilities the choice of alignment strategies is left entirely to the user 76 PAUP 3 1 USER S MANUAL might use two different symbols to refer to missing data one for unavailable and another for not applicable Because PAUP allows only a single symbol to refer to missing data you can convert one t
320. s trees D Berberidaceae O fartooa Characters Save as test data new The Save As dialog box By default saving a file via the Save As command will also change the type of a file created using another application to a PAUP document The Editor command in the Options menu can be used to request retention of the original creator setting After a file has been saved as a PAUP document double clicking on that document from within the Finder will start PAUP with the chosen file as the active file Revert Use Revert if you want to abandon changes to a document e g PAUP data file being edited without leaving the editor The following alert box will appear OK to revert to the last saved f version of the file discarding changes made since it was saved A ET The Revert dialog box If you answer Yes the most recently saved version of the file will be restored to the editor window Page Setup This command brings up the standard system Page Setup dialog box which lets you define printing features such as the size and orientation of the paper on which a document is printed The actual appearance of this dialog box will vary according to which printer has been selected via the Chooser See your Macintosh documentation for further information on using the Chooser and on the available options for each printer 228 PAUP 3 1 USER S MANUAL Print File Choose Print File
321. savefd file test out replace A sample frequency distribution for the Xenopus sample data file is shown below Frequency distribution of tree lengths BEE EEE EEE EEE EEE EEE EEE EEE EEE EEE EEE ELE EE 0 TE 4 gt 1 gt 7 gt 14 gt gt 79 gt 10 FEE EEE EEE EEE EEE EEL EEE EE ELE LEEK EEE EEE EEL EEL ELE mean 69 844156 sd 4 775776 gl 0 664305 g2 0 029334 The shape of this distribution has been suggested to provide a measure of the strength of phylogenetic signal Hillis 1991 Huelsenbeck 1991 PAUP 3 1 USER S MANUAL 115 Hillis and Huelsenbeck 1992 In general the more left skewed the distribution as quantified by increasing negativity of the gj statistic the greater is the amount of phylogenetic signal present in the data However as pointed out by Hillis and Huelsenbeck 1992 see also K llersj et al 1992 the existence of signal says little about the nature of the signal all of the signal may be confined to one or a few relatively uninteresting groups e g mouse rat in a study of tetrapod relationships Lake s Method of Linear Invariants Lake s method is only available if DATATYPE DNA or DATATYPE RNA To use the method you must specify either 1 a quartet of taxa to be evaluated 2 all possible quartets of taxa from the data matrix or 3 four groups of taxa where one taxon from each group makes up the quartet all such quartets are evaluated These options are select
322. sets in a wide variety of formats PAUP 3 1 USER S MANUAL 161 EXAMINING CURRENT STATUS PAUP provides several commands for displaying the current status of data trees and ancestors Data Matrix To display the current data matrix use the SHOWMATRIX command or the Show Data Matrix menu command By default ignored zapped characters will not be displayed They can be displayed 1f SET SHOWIGNORE is used or that option is selected in the Character Matrix Format dialog box Excluded characters are always displayed If SET CMSTATUS is used or status information is requested in the Character Matrix Format dialog box excluded uninformative and constant characters are flagged For example the following matrix has both ignored and deleted characters Input data matrix 1111111 Taxon 1234567890123456 Constant a Uninf E ER Excluded Ex taxonA 000000100101 000 taxonB 0100000011011111 taxonC 111000001101 000 taxonD 111110101101 000 taxonE 1101100111011000 taxonF 111101011111 111 taxonG 101111101011 000 taxonH 101111101011 000 The width of each column can be controlled with the SET CMDCOLWID command or by using the Character Matrix Format menu command The default is one column CMDCOLWID 1 If equate macros are used in the data file e g a 12 and character state reconstructions are requested the equate symbol will be assigned to internal nodes if SET CMSHOWEQ is chosen or this is selected in the Character Matrix Forma
323. sing the SINGLE VAR option That way you don t waste a lot of space describing characters which do not change over trees The output for LENFIT does not give the detail available from a DESCRIBE command in that it does not tell you where on the tree a particular character changed only the amount of change that had to be invoked to explain that character on each tree The advantage of using PAUP 3 1 USER S MANUAL 139 LENFTT is that you can quickly see which characters or sets of characters are consistent with each tree This can be very interesting if there are both minimal and non minimal trees in memory It may be that some characters are consistent only with topologies that are not found on minimal trees NOTE Minimum possible tree lengths are not easily determined when multistate polymorphic taxa are present and the character type is Dollo Consequently character fit measures are not evaluated in that case MANIPULATING TREES Rooting Trees for Output and Character Diagnosis The trees found by PAUP searches are unrooted unless there is some reason to explicitly root them in memory see Outgroups Ancestors and Roots in Chapter 1 These trees must be ouput from the tree buffer as rooted trees for character reconstructions and other analyses There are three rooting options outgroup Lundberg and midpoint Outgroup rooting which includes an assumed outgroup in the analysis is the most common method NOTE Remember that if
324. sired command For a complete list of available commands see the section Command Reference When the command line is open type in the appropriate command and press the Return or Enter keys to execute it or click on the Execute button The Clear button or the Escape key will clear the command line while the Recall button will bring up the last command typed SELECTING ITEMS IN LISTS Many dialog boxes contain scrollable lists from which you may select more than one item e g lists containing tree numbers character numbers names and taxon names To select multiple items you must either shift click or command click In lists that allow multiple selections if neither the shift key nor the command key is held down a click causes all current selections to be deselected and the item receiving the click to be selected If the shift key is held down then as long as the mouse button is held down the current selection is either expanded or contracted If the command key is held down then the selection status of a cell receiving a click is toggled Le if it was originally selected it becomes deselected and vice versa RUNNING UNDER MULTIFINDER OR SYSTEM 7 0 AND LATER PAUP is completely MultiFinder friendly In particular long operations such as tree searches can be run in the background under MultiFinder allowing you to use your computer for other work during this time Of course search times will not be as fast as when
325. solution of the consensus tree the more congruent a set of rival trees is the more highly resolved will be the consensus of those trees Some indices are also 60 PAUP 3 1 USER S MANUAL affected by the symmetry of the tree The consensus indices computed by PAUP are described in Rohlf 1982 and Swofford 1991 GOODNESS OF FIT STATISTICS PAUP outputs several indices that measure the fit of characters to particular trees Three main parameters are used to define these indices Kluge and Farris 1969 Farris 1989a Farris 1989b s length number of steps required by the character on the tree being evaluated m minimum amount of change that the character may show on any conceivable tree g maximum possible amount of change that a character could possibly require on any conceivable tree i e the length of the character on a completely unresolved bush The consistency index Kluge and Farris 1969 for a single character c equals m s Thus if a particular tree explains the data as well as any tree possibly could c 1 Unfortunately the lower bound on c is not 0 but is a function of the distribution of character states in the data matrix For example if two taxa have state 1 but all others have state 0 the maximum possible number of steps on any tree is 2 so that c can be no lower than 0 5 Farris 1989a 1989b proposed two new indices the retention index and the rescaled consistency index For a single character the r
326. status and information on whether the character is informative See the command CSTATUS and the sections Using a Subset of the Characters Including and Excluding Characters Specifying Character Types and the DATA and ASSUMPTIONS blocks for descriptions of this output Show Taxon Status This option produces output listing all deleted taxa and all taxa currently included in the outgroup Show Usertypes This command generates a listing of all user defined character types stepmatrices and character state trees Character state trees are shown in a graphical format so that you can see whether the tree you described is in fact what you intended Input of stepmatrices is less problematical but it is nonetheless useful to include a record of the costs assigned by the stepmatrix in the output See the section User Defined Character Types for further information on interpreting this output Show Data Matrix This command produces a listing of the data matrix The same dialog box shown by the Character Matrix Format command can be used to change output options Show Distance Matrix Use this command to request calculation and output of a matrix of pairwise distances between taxa The formulas used in calculating these distances are given in the section Distance Matrices in chapter 2 Show Ancestral States This command produces a listing of the currently defined ancestral states in a format similar to that used for showi
327. sted Calculation of branch lengths requested human chimp gorilla orang human gorilla chimp orang human orang chimp gorilla E F G Spectral distribution 0 1111 676 a 1222 21 b 1211 27 c 1121 30 d 1112 63 A 1333 1 B 1311 2 C 1131 3 D 1113 22 e 1122 14 f 1212 9 g 1221 13 E 1133 3 F 1313 G 1331 H 1233 h 1322 i 1344 J 1134 j 1123 4 k 1132 L 1323 1232 m 1343 N 1314 n 1213 2 p 1312 Q 1332 q 1223 r 1334 1 S 1341 s 1231 t 1321 u 1234 v 1324 w 1342 Total number of positions used 893 Spectral distribution of 3 taxon components and terms for Pattern WXY_ W_yZ WX_Z 0 111 761 705 709 a 122 35 35 30 b 121 38 43 40 c 112 48 72 17 A 133 2 4 1 B 131 3 3 2 C 113 6 25 29 u 123 0 5 3 v 132 0 0 0 w 134 0 1 2 1 2 3 1 2 3 3 2 3 6 20 26 R1 710 625 622 R2 716 641 642 branch lengths PAUP 3 1 USER S MANUAL L1 2 508 4 266 1 438 7 110 L2 3 726 4 160 2 773 2 876 L3 71 221 24 840 31 291 21 322 Branch lengths for E tree Branch Ist estimate 2nd estimate mean human 2 508 1 438 0 535 chi mp 3 726 2 173 3 250 gorilla 4 160 2 876 3 518 orang 24 840 21 322 23 081 central 4 667 Branch lengths for F tree Branch Ist estimate 2nd estimate mean human 2 508 4 266 3 387 chi mp 2 413 7 1
328. t dialog box SHOWMATRIX will always display the equate symbol If NOCMSHOWEQ is selected the possible character states will be displayed instead of the equate symbol Character Status To display the status of characters use the Show Character Status menu command Excluded zapped constant and ignored characters are 162 PAUP 3 1 USER S MANUAL flagged One column has a Y or N depending on whether the character is informative Output also includes current character type ordered etc character weight and observed states Current status of all characters Character Type Inform Status Wei gh States 1 Irrev Up Y 2 Unord Y 2 3 Unord Y 4 Unord Y 5 Ord Y 2 6 Ord Y 2 1 Unord Y 8 Unord Y 9 Unord N Ignored 10 mytype Y Excl uded 1 11 Unord Y Excl uded 1 12 N Const lgn 13 N Const lgn 14 Unord Y 15 Unord Y 2 There are 1 zapped characters User Defined Character Types To display the status of user defined characters use the SHOWUSER command or the Show Usertypes menu command This will output all stepmatrices or character state trees that have been defined for example Stepmatrix aa TO a b c d e FROM a 2 1 3 2 b 1 3 1 3 c 1 3 2 1 d 1 1 2 1 e 1 2 2 1 Character state tree typea 0 2 1 Ancestral States ANCSTATES To display the ancestral states use the SHOWANC command or the Show Ancestral States menu command This ouputs the name of the cur
329. t fails the 3 1 test Assigning e gives a tree length of 36 4 x 9 while assigning any other state gives a tree length of 39 3 x 13 plus 1 x 0 9 9 a Sete pb 9 9 d c a 0 13 b Sa ac b 13 13d Example of failure of 3 1 rule a optimal reconstruction with length 36 not found by 3 1 rule b reconstruction of length 39 allowed by 3 1 rule Of course this example is derived from a fairly unusual stepmatrix In reality the 3 1 rule will choose states that allow optimal reconstructions most of the time and the speedup in searches can be enormous because fewer states need be considered The Pairwise Homoplasy and Patristic Distance Matrices If you want to display a summary of homoplasy and patristic distances between taxa where patristic distances P sum of branch lengths on path between each pair of taxa D observed character difference and the PAUP 3 1 USER S MANUAL 137 homoplasy H P D check the Patristic distance matrix and or Homoplasy matrix items in the Describe Trees dialog box or use the HOMOPLASY or PATRISTIC option of the DESCRIBE command For example the command describe 1 patristic homoplasy will output patristic distance and homoplasy matrices for the first tree Note Multistate unordered and or stepmatrix characters excluded from patristic distance calculations Patristic distance matrix Below diagonal Adjusted character distances Above diagonal Patristic distances 1 2 3
330. t polytomous nodes indicate simultaneous divergence of multiple lineages rather than uncertain resolution i e hard polytomies vs soft polytomies Maddison 1989 Strict Consensus Strict consensus trees contain only those groups appearing in all of the rival trees Sokal and Rohlf 1981 Page 1989 This can be considered to be the most conservative estimate of consensus and is the simplest to interpret In the figure below tree fis the strict consensus of trees a b and c It preserves the two groups ABC and DEP but there is no resolution within either of those groups In the case of EDF two of the rival trees a and c agree on the resolution of D E F while the other tree b resolves this clade as E D F This conflict is reflected in the strict consensus by the trichotomy E D F In the case of ABC because only one tree resolves c resolves the clade and two do not the strict consensus must then be unresolved The liability of this technique stems from its demand for identity in each of the rival trees For example two trees may be identical except for the placement of a single taxon yet may have a completely unresolved consensus In this instance the consensus is too strict Adams 1986 Funk and Brooks 1990 in that none of the PAUP 3 1 USER S MANUAL agreement between the trees is preserved in the consensus when the unstable taxon is included Three rival trees a b c and their strict d semistr
331. t to the group ABCD but derived relative to some more inclusive group If on the other hand the trees are rooted by the attachment of a hypothetical ancestor possessing state O the left tree will be shorter 2 steps than the right tree 3 steps Ww r Wo S Na YO x No 2 Reconstructions for a Dollo character on the trees of the figure above under a rooted Dollo model The extra length in the right tree comes from the inclusion of the initial gain of state 1 in the tree length This example makes it clear that trees computed under Dollo parsimony are intrinsically rooted only if we assume we know the state at the outgroup node sensu Maddison et al 1984 A more formal definition of the unrooted Dollo criterion is the following A character state reconstruction satisfies the Dollo constraint if it is possible to trace a path between any pair of nodes possessing the same character state without passing through a node possessing a less derived state For example reconstruction A below while requiring fewer steps than reconstruction B is not a Dollo reconstruction as tracing the path connecting the two terminal nodes possessing state 1 requires passing through internal nodes possessing the less derived state 0 Reconstruction B on the other hand does satisfy the Dollo constraint PAUP 3 1 USER S MANUAL 11 0 0 A NI ES oe a b Mo ma a Two reconstructions for a single character on an unrooted tree Reconstruction
332. table of partition frequencies a Bootstrap consensus tree Retain groups with frequency gt i Retain groups compatible with 50 majority rule consensus f there are excluded or ignored characters Sample from included nonignored characters only Sample from all characters in data matrix j characters are not weighted equall O Ignore weights use equal weights for bootstrapping i Sample characters with equal probability but apply weights OTreat weights as repeat counts Defaults Cancel Continue The Bootstrap dialog box Specify the number of replications desired and change the random number seed if desired If there are excluded characters or if all characters are not weighted equally you must tell PAUP how to treat those characters during the bootstrap replicates Indicate whether you want to perform an exact branch and bound or heuristic search by clicking on the appropriate button then click on Continue A second dialog box will appear allowing you to specify search options The bootstrap procedure begins when you click the Bootstrap button of this second dialog box Random Trees Selecting Random Trees will evaluate the length of the number of random trees chosen The output will include a frequency distribution of tree lengths 260 PAUP 3 1 USER S MANUAL Random Trees Options Evaluate random trees O Save tree length frequency distribution data to file Class interval for f
333. ter states which satisfy the 3 1 rule are considered candidates for assignments at internal nodes See the section Stepmatrices Special Considerations in Chapter 2 for a detailed explanation of the 3 1 rule I The following two options are specific to the Macintosh version See the description of the Searching menu command in Chapter 4 for more information NO BACKGROUND Ordinarily PAUP continues processing when it is moved to the background Specify NOBACKGROUND to suppress background processing thereby giving more time to the foreground application NO CHECKEVTS NOCHECKEVTS disables event checking causing all mouse clicks an key presses to be ignored Speed of PAUP searches is improved somewhat but it will not be possible to stop the search without restarting the computer or to switch to a different application under MultiFinder The following options are described in the section Options Affecting Multiple Commands earlier in this chapter NO COLLAPSE NOJENFORCE CONSTRAINTS contraint tree name NO CONVERSE NO JINCLUDEANC NO STATUS PAUP 3 1 USER S MANUAL 217 ROOT OUTGROUP LUNDBERG MIDPOINT OUTROOT POLYTOMY PARAPHYL MONOPHYL NO TCOMPRESS OPT ACCTRAN DELTRAN MINF NO SHOWIGNORE NO CMLABELS NO CMCSTATUS CMDCOLWID column width SHOWANC Use the SHOWANC command to request a listing of the ancestral character states currently in effect see Defining Anc
334. th between terminal nodes B and D heavy lines h Subtree of node G heavy lines and boldface The total number of distinct unrooted terminally labeled strictly bifurcating trees for T terminal taxa is given by the formula T B T i 5 1 i 3 Cavalli Sforza and Edwards 1967 Felsenstein 1978b Table 2 shows the value of B T for several values of T Obviously the number of trees quickly becomes quite large as the number of taxa increases Table 2 The number of unrooted binary terminally labeled trees B T for T terminal taxa T BT 1 3 15 105 945 10 395 135 135 2x106 8 x 10 2 mare N0N0 Du W Uu Oo PAUP 3 1 USER S MANUAL 7 20 2 x 102 50 3 x 1074 CHARACTER TYPES PAUP implements different parsimony variants through the declaration of a character type for each character included in the data matrix Available types are ordered Wagner unordered Fitch Dollo irreversible Camin Sokal and user defined Any combination of character types can be assigned to the characters in the data matrix The character types and weights that are specified constitute the a priori assumptions that are in effect for a particular analysis Character types may be classified as either undirected or directed An undirected character is one in which for every pair of states a and b the cost in tree length is the same for the transformations A B and B gt A All of the standard character types are undirecte
335. the Print Trees command However for majority rule including Bootstrap consensus trees you can request that the group frequencies be drawn on branch labels Get Trees from File Use this command to read trees from a file containing a NEXUS format TREES block into memory Get trees from file S Sample data files y Hard Disk D 5 Taxon test Eject D Anacyclus D Angiosperms Desktop D Angiosperms trees D Berberidaceae D Cartoon Characters D cartoon phylip D command file Get Trees D Eucalyptus ue Options The Get Trees from File dialog box By default all trees in the file will be read in Alternatively by clicking the Options button you may specify a range of trees to import or you may use Boolean operations to select which trees are to be retained The Boolean operations are controlled using the Venn diagram provided The circle on the left represents the trees in memory the one on the right represents the trees in the file When a circle or a segment of it is dark that means the trees in that region will be kept For example selecting all the trees in the file the default would be achieved by the following PAUP 3 1 USER S MANUAL 271 Trees to be included from file All trees O Trees through Trees to be retained in memory Trees currently Trees from file in memory Cancel ox The Options dialog box This corresponds to MODE 3 in the GETTREES comm
336. the full set of binary trees all of PAUP s search algorithms therefore provide an option for collapsing interior branches of zero length Unfortunately however the criterion for collapsing zero length branches is complicated by the fact that for some characters more than one most parsimonious reconstruction MPR Swofford and Maddison 1987 may exist An MPR is the set of state assignments to each interior node that allow the length required by a character on a given tree to be minimized Consequently a branch may have zero length under one MPR but a length greater than zero under a different MPR Swofford and Maddison 1987 described an algorithm for finding the minimum and maximum lengths of a branch over all MPRs for a single character We PAUP 3 1 USER S MANUAL 51 can obtain the minimum possible and maximum possible branch lengths by summing over all characters the minimum and maximum branch lengths for each character Conceivably three rules could then be used to decide whether to collapse an interior branch 1 Collapse an interior branch if the minimum possible length of the branch is zero That is if there exists at least one MPR for every character such that no length is assigned to the branch the branch is collapsed 2 Adopt an ancillary criterion for choosing one MPR from the full set of MPRs for each character or choose one arbitrarily If no length is assigned to the branch for all characters then the branch
337. the ingroup depends upon how the outgroup taxa are related to each other see Maddison et al 1984 for a complete analysis of this problem If outgroup relationships are already well resolved you can use methods described by Maddison et al 1984 to assess character polarities a priori and include an ancestral taxon in the PAUP 3 1 USER S MANUAL analysis see also Donoghue and Maddison 1986 Otherwise it makes much more sense either to 1 include several outgroup taxa in the analysis with the root of the ingroup portion of the tree being determined by the location at which the outgroup taxa connect to the tree i e Maddison et al s 1984 simultaneous resolution of ingroups and outgroups taken to the extreme see also Farris 1972 or 2 compute an unrooted tree for the ingroup taxa only making no a priori assumptions about polarity and then attach an outgroup taxon or hypothesized ancestor to the tree a posteriori Lundberg 1972 Although clearly better than guessing at the polarity both of the above approaches have some disadvantages Maddison et al 1984 Meacham 1984 Donoghue and Maddison 1986 Meacham 1986 In order to be fully effective the first method requires the inclusion of characters that are informative with respect to outgroup relationships even if these characters are invariant within the ingroup Maddison et al 1984 On the other hand the second method can generate optimal trees that ar
338. the root The approach used in PAUP circumvents this problem the uncertainty in locating the root does not come into play until after the consensus is computed This approach can not be used for the Adams 1972 method which is defined only for rooted trees consequently trees are rooted prior to calculation of the consensus in this case In general a consensus tree for a group of minimal trees will be longer than any of those trees e g Miyamoto 1985 This is because consensus trees are usually less resolved than any of the rival trees which automatically means that some characters will have to change more times on the unresolved consensus than they would have to on the more highly PAUP 3 1 USER S MANUAL 57 resolved rival trees For example the figure below shows two rival trees of length four Their strict consensus is tree C Because this tree is unresolved both characters two and three must have two steps Its total length is then five steps one step longer than either of the rival trees Tree lengths of rival trees and their consensus a rival tree of length four b a different rival tree of length four c strict consensus of a and b length is now five The main point here is that the consensus tree is not an optimal tree for a particular data set and should not be treated as such It is a summary of agreement among trees but it should not be interpreted as a phylogenetic tree Such an interpretation assumes tha
339. the tree and the character polarities simultaneously rather 20f course a third possibility is that neither One and Two nor One and Three represent sister taxa with all shared derived states being homoplasies 4 PAUP 3 1 USER S MANUAL than going through the two stage process of assigning polarities first and then estimating the tree NOTE The concepts of character order and character polarity are often confused The former defines the nature of permitted character state transformations whereas the latter refers to the direction of character evolution Specifying how character states are ordered with respect to each other is not the same as determining which character state is ancestral For further amplification of these points see the sections entitled Character Types and Outgroups Ancestors and Roots later in this chapter TREE TERMS The most general terminology for describing the various components of trees is derived from the field of graph theory e g Harary 1969 Gould 1988 Some of this terminology is reviewed briefly here although formalism is minimized A graph consists of a set of vertices and a set of edges where each edge is a line joining a pair of vertices Two distinct vertices are adjacent 1f they are joined by an edge the edge is said to be incident to those vertices The degree of a vertex is the number of edges with which the vertex is incident A path is a sequence of distinct edges such that each ed
340. the tree requiring the least homoplasy would also be the shorter tree tree A 5 steps tree B 6 steps Clearly there is no real difference between computerized methods and manual methods for choosing trees under the parsimony criterion Two factors can cause the methods to differ in actual practice however The first lies in subjectivity introduced by the investigator who may prefer a tree requiring two or three extra steps in a character assumed to be unreliable or plastic over a tree requiring a single extra step in a better character The second is that the human brain is unable to deal effectively with the complexity of large and or noisy data sets The above example also illustrates one reason why excessive concern with a priori assessment of character polarities is unwarranted When incongruent character distributions occur as in this case the sharing of a derived character state among two or more taxa need not indicate close relationship of those taxa The problem is accentuated when the outgroup being used for the polarity assessment is heterogeneous In such cases parsimony can provide the basis for assigning character polarities but the most parsimonious state assignment s for the most recent common ancestor of the ingroup depends upon the overall structure of the tree Farris 1982 including the relationships of the outgroup taxa Maddison et al 1984 Thus the only fully satisfactory recourse is to infer the topology of
341. thin the ASSUMPTIONS block although you may also define them from within a PAUP block or via the command line Once a set has been defined you can use it in any subsequent command that recognizes sets of that type Character Sets CHARSETs The use of character sets allows you to refer to a group of characters by a single name The only restriction is that the name of the character set cannot be the same as the label assigned to an individual character for obvious reasons After defining a character set using a CHARSET command you can use the character set name in any place that you would ordinarily use a character name or number For example you could define two character sets as follows charset larval 1 10 26 30 charset adult 11 25 31 50 and then use the CHARSET names in subsequent commands exclude larval excludes larval characters weights l larval 2 adult assigns double weight to adult characters PAUP 3 1 USER S MANUAL 99 The preceding two commands are exactly equivalent to the commands xclude 1 10 26 30 weights 1 1 10 26 30 adult 11 25 31 50 Note that you may freely mix character names and or numbers with character set names For example the command weights 2 larval 12 14 45 49 assigns a weight of 2 to all of the larval and some of the adult characters Character sets are especially useful when analyzing molecular sequence data For example you might subdivide the regions of a protein
342. ting Compressing can save a great deal of space especially when there is a large number of taxa describe 1 would output an uncompressed tree 150 PAUP 3 1 USER S MANUAL F662 EEOC EAS AAA EEE ERE Erica tas A gt gt bie SO AR 1 GeO OO St Ge GH OS TO CEEC KO OR GO REE E gt gt ficcccccccld HI ERE OPER EEE CE OEE CME EEL EE MEM MEME EERE G gt gt fic ccc cece 3 FEE EEE EEE EE EEE EEE EEE EEE EEE D gt gt ficccccccc 2 AGO OOOO EEE EE EEK EACH EEE E gt gt ficcccccecell feo F gt gt ficcccccccld flecccccccce G fice eed ficcccccccce H while describe 1 tcompress will output a compressed tree A EEE EEE EEE EEE EEE EEE EE EEE EEE EEE EEE EEE EEE gt F amp O CLEC ER OEE O OEA O EMEA LEAMA ESA EA AB ficcccccecld AEE EEE ELE EEE EEE EE EEE EEE EEE ELE 0 fic ccc ce ee Ta B10 CEO OOEE CLE OO EE CE EEE EE LEE AMEN ECM AEE ND ficccccccc 2 MARA AREA E ficcccccccell fC F ficcccccccld flecccccccce G ferric peticii H Trees output by plotting or printing cannot be recovered for further manipulation and analysis for that they must be saved to a NEXUS file see Saving Trees to Files above Changing the Order of Taxa on Trees As is generally the case for phylogenetic trees PAUP s trees are formally unordered see Tree Terms in Chapter 1 however they may be ordered according to one of four available conventions for output purposes You can specify which ordering to use by the TORDER option of the SET co
343. ting tree number specifies the number of the first tree to get TO ending tree number If nonzero ending tree number specifies the number of the last tree to get MODE 11213141517 By default MODE 3 any preexisting trees in memory are replaced by the trees read from the file The MODE setting allows you to alter this behavior Let M the set of trees originally in memory and T the set of trees from the tree file The following mode values are then available 1 replace M by T M e keep trees from the file that are not originally in memory 2 replace M by T A M keep trees from the file that are also originally in memory 3 replace M by T i e replace all trees in memory by all trees from the file 4 replace M by M T 1 e keep trees in memory that are not also in the file PAUP 3 1 USER S MANUAL 199 HELP 5 replace M by M T e keep trees that are either currently in memory or in the file but not both places 7 replace M by M UT 1 e append trees from file to trees originally in memory with elimination of duplicates ROOT If directed character types are in effect for one or more characters and the trees being input are unrooted you will ordinarily be asked 1f you want to convert the trees to rooted trees otherwise you will not be able to use the new trees unless you change the character types or root them later If you specify ROOT explicitly this conversion will occur
344. tion keywords etc in any combination of upper and lower case characters the only exception pertains to the MATRIX command In addition PAUP allows abbreviation of command names and keywords to the shortest unambiguous truncation Note that other NEXUS conforming programs may not accept these abbreviations MacClade in particular does not allow abbreviations so 1f you want your data file to be MacClade compatible then commands visible to MacClade i e those not contained within a PAUP block should not be abbreviated Identifiers Identifiers are simply names given to taxa characters and other PAUP input elements such as character sets taxon sets and exclusion sets They may include any combination of upper and lower case alphabetic characters digits and punctuation If the identifier contains any of the following characters EACEA E O E or a blank the entire identifier must be enclosed in single quotes Underscores _ are translated to blanks unless the identifier is enclosed in single quotes For example the identifiers Homo_sapiens and Homo sapiens are equivalent but distinct from Homo_sapiens To include a single quote in the identifier you must use two consecutive single quotes Any trailing blanks are stripped before the identifier is stored leading blanks are preserved Examples of valid identifiers subterraneus Mus_musculus H _sapiens H sapiens 429 Fred s new sp rusticus 1 s
345. tions options deftype ord typeset mytypes unord 3 7 dollo 9 13 endblock The asterisk preceding mytypes is important It informs PAUP that you want the type set to go into effect immediately Otherwise the type set would not be used unless it was specifically invoked by an ASSUME command or a Choose Assumption Sets menu command NOTE Any characters not explicitly assigned a type in a TYPESET command use the character type specified by the DEFTPYE option in the OPTIONS command If you do not use an OPTIONS DEFTYPE __ directive the default character type is UNORD e By using the CTYPE command either from within a PAUP block or from the command line The same character types assigned in the example above could be specified using the command ctype ord all unord 3 7 dollo 9 13 NOTE Later specifications override earlier specifications in a CTYPE command so that characters 3 7 and 9 13 are set to UNORD or DOLLO as intended despite the preceding ORD ALL This sometimes saves effort since the alternative is to specify every character type explicitly e g ctype ord 1 2 4 6 8 14 unord 3 7 dollo 9 13 e By using the Set Character Types menu command CHARACTER WEIGHTING By default PAUP assigns equal weight to each character in the data matrix You may however prefer to attach greater weight to some characters than to others When you assign character weights PAUP simply computes a weighted sum of the single
346. tly set MSTAXA MacClade uses the uncertainty interpretation if the character state set is enclosed in curly braces and the polymorphism interpretation if the set is enclosed between parentheses If you set MSTAXA here this setting will override the interpretation implied by the curly brace vs parenthesis notation in MacClade ZAP character list All characters in the character list will be ignored see Using a subset of the characters in Chapter 2 GAPMODE MISSING NEWSTATE If GAPMODE MISSING gap characters in sequence data are treated as missing If GAPMODE NEWSTATE gap characters are treated as a fifth base or 21st amino acid See Alignment gaps in Chapter 2 for more information If you are satisfied with the defaults indicated above and you do not want to zap characters then you do not need to include an OPTIONS command STATELABELS The STATELABELS command is recognized but not interpreted by PAUP In MacClade it supplies names for the character states of each character TAXLABELS Use the TAXLABELS command to supply taxon names when the data matrix is entered in transposed format The syntax is TAXLABELS taxon label taxon label2 taxon labelNTAX PAUP 3 1 USER S MANUAL 175 If the data matrix is transposed and you do not provide a TAXLABELS command the integers 1 through NTAX will be used to identify taxa in both input and output The length limit on these labels is 16 characters althoug
347. to internal nodes so that you can determine whether or not a reconstruction contains ambiguities See Character State Optimization for information regarding the resolution of ambiguity in character state reconstructions OUTGROUPS ANCESTORS AND ROOTS In most cases trees computed by PAUP are unrooted and need be rooted for output and interpretational purposes only The absence of a defined root is a consequence of the fact that for undirected character types the tree length is the same regardless of the position of the root However rooted trees are computed when any of the following are true 1 characters of type irreversible are present 2 asymmetric stepmatrix characters are present or 3 a hypothesized ancestral taxon is explicitly included in the analysis at the request of the user For irreversible and asymmetric stepmatrix characters the length of the tree depends upon the position of the root so the trees must be stored as rooted trees When all characters are undirected the inclusion of a hypothesized ancestral taxon corresponds to the traditional but unnecessary practice Meacham 1984 Meacham 1986 of a priori character polarization This approach is reasonable when the polarity determination is unambiguous i e there is no heterogeneity in the outgroup for characters that are variable within the ingroup However when the outgroup is heterogeneous the most parsimonious assignment of an ancestral condition for
348. to a file before getting trees with this option Different tree number options and boolean operations are selected in the Get Trees from File dialog box by manipulating two partially intersecting circles one representing trees in memory one representing trees in the file Manipulating trees in this way gives you the ability to compare sets of trees such as those derived from different data sets different search options or different assumptions about characters e g weight One you have imported and kept the trees you want you can then save them to a different treefile This allows you to have different sets of trees in different treefiles and to extract different subsets of trees from each of those different treefiles Trees may be imported either rooted or unrooted You cannot have both types in memory at the same time If you attempt to retain one type in memory and import another type PAUP will warn you that rooted and unrooted trees cannot be mixed Also if characters require a rooted tree such as Dollo and irreversible characters and you import unrooted trees PAUP will warn you when you attempt to reconstruct characters on those trees In that case you must either root the trees or change character types PAUP 3 1 USER S MANUAL 145 See Outgroups Ancestors and Roots for a discussion of character types and rooted unrooted trees You can also use the Import File command to get trees created by other programs such as Henn
349. to print the document being viewed in the frontmost editor window generally a PAUP input or output file to the currently active printer You can use the Chooser desk accessory to change printers Note that Print File may only be used while in editor mode it is disabled otherwise The standard system printing dialog box will appear allowing you to specify things such as the number of copies page range etc The actual appearance of this dialog box will vary according to which printer has been selected via the Chooser See the preceding item for problems in printing PAUP output with semigraphics Note that PAUP makes no attempt to deal intelligently with files containing lines too long to print If for example a matrix contains very long lines PAUP will truncate them when they are sent to the printer One possible solution is to use the SHOWMATRIX command or Show Data Matrix dialog box and log the output to a file for subsequent printing Alternatively you can use those commands and then print the display buffer See the Log Output to Disk dialog box for a more detailed description See your Macintosh documentation for further information on using the Chooser and on the options found in the system printing dialog box Echo to Printer If you choose Echo to Printer all subsequent output to PAUP s main display window is also sent to the currently active printer Each time you choose Echo to Printer it will toggle the current printing status
350. tree B2 C2 1 C2 5 are constructed in turn Then we backtrack once again to tree A and proceed down the third path toward trees C3 1 C3 5 Eventually we will have constructed all of the possible trees culminating with tree C3 5 If the length of each tree containing all five taxa were evaluated at the time of its construction we would have performed an exhaustive search equivalent to that described in the above section B E B E D c c11 c D C31 A A E D EG B D B B as SS c FA D B1 B3 X A A BDC eke E E A A B DEC B CED 3 4 C1 4 A A B D c C1 5 E A DC C2 1 C2 2 C2 3 C2 4 C2 5 Search tree for branch and bound algorithm Suppose that L represents an upper bound on the length of the shortest tree s For the present we can obtain L for example by evaluating a random tree if we know that a tree of length L exists then the length of the optimal tree s cannot exceed this value If as we are moving along a 34 PAUP 3 1 USER S MANUAL path of the search tree toward its tips containing all T taxa we encounter a tree whose length exceeds L then we need proceed no further along this path connecting additional taxa cannot possibly decrease the length Thus we can dispense with the evaluation of all phylogenetic trees that descend from this node in the search tree and immediately backtrack and proceed down a different path By cutting off portions of the search tree in this manner we can greatly reduce the number of t
351. tree and are therefore also uninformative When missing data are present the same ambiguities referred to in the above paragraph with respect to the collapsing of zero length branches with uninformative characters also apply to invariant characters The type of a character can affect its informativeness For example a character for which states 0 1 and 2 were observed once five times and once respectively would be uninformative as an unordered character but informative as an ordered character PAUP does not attempt to evaluate the consistency index or informativeness of user defined stepmatrix characters due to the difficulty of determining the minimum possible lengths required by these characters In principle this minimum length could be established by performing separate parsimony analyses on each stepmatrix character considered separately but the amount of computation required would be horrendous If you absolutely must determine whether a given stepmatrix character is uninformative you could exclude all other characters and perform an exact or heuristic search to obtain or estimate the minimum possible length for that character This minimum length could then be compared to the maximum possible length available via the DESCRIBE command or the Describe Trees menu command See Chapter 3 for the complete list of commands available in the DATA block and a description of their syntax Additional examples can be found in the samp
352. trix was designated as the hypothetical ancestor In PAUP 3 1 the definition of the ancestor is no longer part of the matrix and is done using the ANCSTATES command This allows more flexibility in changing ancestral states for different analyses of the same data matrix e g to test the effect of alternative polarity assignments Entering the matrix in transposed format There is considerable flexibility in the form of the input matrix By 69 default rows correspond to taxa and columns to characters If you prefer to list all of the data for each character on a single row with columns corresponding to taxa you can use the TRANSPOSE option of the FORMAT command Since the taxon names are no longer defined in the data matrix itself you can use the TAXLABELS command to define the taxon names For example the DATA block below defines the same matrix as the one above begin d dimensions ntax 4 nchar 5 format transpose taxlabels taxonl taxon2 taxon3 taxon4 matr E o O GO endbloc Placing taxon and character names ata ix harl har2 har3 har4 har5 Ky 0011 0110 1100 1100 1710 Ordinarily taxon names precede the character state data for each taxon If you prefer to place your taxon names at the end of each row rather than t PAUP 3 1 USER S MANUAL he beginning you can use the LABELPOS RIGHT option of the FORMAT command i e the names are to the right rather than the left of the dat
353. tstrap revisited Cladistics 5 113 129 Sankoff D and R J Cedergren 1983 Simultaneous comparison of three or more sequences related by a tree Pages 253 263 in D Sankoff and J B Kruskal ed Time Warps String Edits and Macromolecules the Theory and Practice of Sequence Comparison Addison Wesley Reading Mass Sankoff D R J Cedergren and W McKay 1982 A strategy for sequence phylogeny research Nucleic Acids Research 10 421 431 Sankoff D and P Rousseau 1975 Locating the vertices of a Steiner tree in an arbitrary metric space Mathematical Programming 9 240 246 Sokal R R and F J Rohlf 1981 Taxonomic congruence in the Leptopodomorpha re examined Systematic Zoology 30 309 325 Swofford D L 1991 When are phylogeny estimates from morphological and molecular data incongruent Pages 295 333 in M M Miyamoto and J Cracraft ed Phylogenetic Analysis of DNA Sequences Oxford University Press New York N Y Swofford D L and S H Berlocher 1987 Inferring evolutionary trees from gene frequency data under the principle of maximum parsimony Systematic Zoology 36 293 325 Swofford D L and W P Maddison 1987 Reconstructing ancestral character states under Wagner parsimony Mathematical Biosciences 87 199 229 Swofford D L and W P Maddison 1992 Parsimony character state reconstructions and evolutionary inferences Pages 186 223 in R L Mayden ed Systematics Historical E
354. tted unless all characters are of type ORD or UNORD If you only want to know the length of the shortest tree s and do not care to obtain all trees of this length uncheck the Save all minimal trees MULPARS option The single tree found is guaranteed to be of mintmum length and the search often runs much faster If you check Tree length frequency distribution only PAUP will not save any trees but instead compute a frequency distribution of the number of trees of each length less than or equal to the length specified for Keep all trees lt length ___ This provides a way to obtain the number of trees that are one step from optimal two steps from optimal etc for data sets that are too large to allow a full computation of the tree length frequency distribution via exhaustive search The branch and bound run times are very data dependent It usually runs quickly for 12 or fewer taxa For between 13 and 18 taxa the run times vary considerably depending on the messiness of the data and other factors to complicated to go into here It has been successfully run for 20 or more taxa but for data sets this large extremely long run times are possible Don t even try it for data sets much beyond 20 taxa Exhaustive The Exhaustive Search command provides an alternative to the branch and bound algorithm for an exact search guaranteeing to find all minimum length trees If all you are interested in is the most parsimonious trees you should
355. u may also use the same command file on different data sets providing the number of characters and or taxa do not conflict PAUP 3 1 USER S MANUAL Batch Processing If you wish to run unattended a single batch file with multiple data sets and analyses you must use two SET options in the first PAUP block NOWARNRESET and AUTOCLOSE This is because by default PAUP will display a warning if you try and reset the data file and will not close windows unless told to do so thus you would never proceed beyond the first data set Batch files allow unattended multiple analyses useful 1f you have many data sets to run but do not wish to or cannot be present The format of a typical file with multiple data sets would be NEXUS begin data it ee 5 first data set endblock begin paup set nowarnreset autoclose outgroup 1 2 3 hsearch swap tbr addseg simple hold 10 endblock begin data gi a second data matrix here endblock begin paup PESENE paup commands new search endblock Each analysis may have its own TREES or ASSUMPTIONS block as these are reset when a new data set is processed Error Messages and Input Files Because of the extremely free format of data entry it is sometimes hard for PAUP to figure out exactly where a user erred when it tries to execute an input file This leads to bogus error messages in that what PAUP reports is not the real error but a reflection of a different error earlier
356. uding Dollo irreversible and character state tree characters Exclude from exported file Transform using additive binary coding Export unmodified Dialog box for specifying PHYLIP export options PAUP 3 1 USER S MANUAL 231 Neither Hennig86 nor PHYLIP accept multistate taxa If the characters are binary or ordered multistate they can be equivalently recoded but only under the interpretation of uncertainty rather than polymorphism using additive binary coding with missing values used to indicate the points of uncertainty If multistate taxa are present and you request export of data to Hennig86 the following self explanatory dialog will appear Hennig86 cannot handle multistate taxa Multistate taxa for ORDERED including Dollo and irreversible characters can be emulated using additive binary coding under the interpretation of uncertainty see Multistate Taxa Do this all characters will be recoded into additive binary Assign the state to multistate taxa Multistate taxa for UNORDERED including stepmatrix characters cannot be emulated in Hennig86 and will be assigned the state Dialog box for specifying Hennig86 options for exporting data containing multistate taxa Import File The Import File command allows you to import a file in one of four primary formats PHYLIP Hennig 86 tab delimited text or simple text A pop up menu will appear for you to select the file for
357. ups 1 2 3 specify constraints 1 2 3 Note that the outer pair of parentheses are necessary otherwise the converted specification would be 1 2 3 4 5 6 rather than 1 2 3 4 5 6 Constraints are enforced during a search by selecting this option in the search dialog box or by using the ENFORCE option as in this command hsearch enforce constraints testl This will result in the following output when the search is begun Keeping only trees compatible with constraint tree other Enforcing this constraint during the search will mean that any tree on which the human chimp group does not appear is automatically rejected regardless of its length For the monophyly constraint above the tree on the left below would be accepted while the tree on the right would not The rules for defining and enforcing backbone constraints are exactly the same as for monophyly constraint trees with the exception that backbone trees always omit some of the taxa When a backbone is specified PAUP 3 1 USER S MANUAL 157 omitted taxa are left off the subtree For example the following command will define the constraint tree bb1 as a backbone constraint constraints bbl backbone taxonA taxonB taxonC taxonD which is the equivalent of the following rooted constraint tree The tree below on the left is one of the rooted trees that satisfies the backbone constraint the tree on the right does not satisfy it See the section Se
358. ure between trees a and b which are identical if taxa B and G are removed It does not however imply that a monophyletic ACDEF is found on either of the rival trees The appearance of B and G at the unresolved portion at the base of the consensus indicates that their placement relative to the other taxa is different on each of the rivals It does not indicate that they both belong outside the ACDEF group since the rival trees suggest that either B or G may be contained within that group ABCDEFG AGCDEFB ACDEFGEB a b c Two rival trees a b and their Adams consensus c Note that the group AC appears in the Adams consensus yet is not found on either of the two rival trees the same is true of ACD ACDE and ACDEP This is a common occurrence with the Adams method and has been the source of much of the criticism concerning it This criticism is valid only if Adams trees are interpreted in the same way as are strict consensus trees On a strict consensus the appearance of an AC clade would indicate that clade appeared on all rival trees On an Adams consensus the appearance of an AC clade means instead that the rival trees indicate that A and C are more closely related to each other than either is to D E or F Consensus indices Several authors have developed indices that use the topology of the consensus tree to measure the degree of congruence of the rival trees Many of these measures are simple functions of the re
359. uristic search keeping no more than twenty five trees of fifty or more steps If KEEP and NCHUCK are used together once PAUP reaches the limit specified by NCHUCK and CHUCKLEN it stops looking for shorter trees that is it doesn t replace longer trees satisfying the KEEP limit with shorter trees satisfying the KEEP limit When that happens PAUP will either stop or go on to the next step depending on the circumstances The command hsearch nchuck 25 chucklen 50 keep 40 will try to keep all trees less than forty steps but if twenty five trees of length greater than or equal to fifty steps are found first PAUP will stop Part of the difficulty in using heuristic searches is that there are no settings that will get the best results for all data sets Each search will be more or less unique although different types of search will be similar The choice of options will almost always affect the outcome of the search This becomes critical when the data set is one which produces a very large number of trees In that instance there are several strategies for trying to get the best estimate of the set of shortest trees First a simple illustration of how option choice can affect the number of optimal trees found This example uses the sample Menidia data set distributed with the program If this data set is analyzed with the following command hsearch addseq asis swap nni hold 1 ten trees are found If we now choose to hold ten trees instead
360. visit an interior node for the last time optionally write the symbol for the character state If you omit the symbol for a state corresponding to an internal node of the character state tree PAUP will use an asterisk to label that state in the output For ease in interpreting the output however it is best to label all nodes on the character state tree Applying these rules to the example above the character state tree description would develop as follows PAUP 3 1 USER S MANUAL 87 Number in Description to sequence this point Explanation 0 leaving for internal node D s left descendant 1 leaving for internal node C s left descendant 2 A terminal node A visited 3 CA leaving for internal node C s right descendant 4 A B terminal node B visited leaving internal node C s rightmost descendant 5 A B C internal node C visited for last time 6 A B C leaving for internal node D s last descendant 7 A B C E terminal node E visited leaving internal node D s rightmost descendant 8 A B C E D internal node D visited for last time The character state tree could therefore be defined using a USERTYPE command in the ASSUMPTIONS block as follows begin assumptions usertype mycst cstree A B C E D endblock The name mycst can be any name you choose CSTREE is required because by default user defined character types are stepmatrices see below rather than character state trees Just for practice here is
361. want to root the trees AncStates v standard Rooting Options Cancel The Root Trees dialog box Similarly you may want to deroot trees if current character types do not require rooted trees Convert all trees in memory from B rooted to unrooted EE The Deroot Trees warning box Note that when you convert rooted trees to unrooted trees trees that were formerly distinct may become equivalent You can use the Condense Trees command to eliminate these duplicate trees Tree to Tree Distances This command produces a matrix of tree to tree distances computed according to the symmetric difference or partition metric Penny and Hendy 1985 Tree to Tree Distances p Compare all pairs of trees O tree to all other trees E Show all distances E Show frequency distribution The Tree to Tree Distances dialog box The Compare options allow you to output distances for all trees to all other trees or for all trees relative to a selected tree Clicking on the Show all distances checkbox will output a distance matrix for all selected trees while clicking on the Show frequency distribution will only output a frequency distribution of tree distances Trees that are most similar will PAUP 3 1 USER S MANUAL 263 have the lowest distance value The output is very useful in identifying classes of similar trees See the section Tree to Tree Distances and Swofford 199
362. wn below i i E B t F T A 7 E B C t F 4 B G F G Nearest neighbor interchanges around the branch partitioning taxa A B C from D E F In subtree pruning and regrafting a subtree is pruned from the tree e g the subtree containing terminal nodes A and B as indicated The subtree is then regrafted to a different location on the tree All possible subtree removals and reattachment points are evaluated PAUP 3 1 USER S MANUAL 41 m T RR F 6 N C A An example rearrangement via subtree pruning and regrafting a the starting tree b tree after pruning the subtree containing terminal taxa A and B c tree after regrafting this subtree to the peripheral branch incident to terminal taxon G In tree bisection and reconnection the tree is bisected along a branch yielding two disjoint subtrees The subtrees are then reconnected by joining a pair of branches one from each subtree All possible bisections and pairwise reconnections are evaluated 42 PAUP 3 1 USER S MANUAL E Example rearrangement via tree bisection reconnection Initial tree is bisected into two subtrees and then reconnected along a different pair of branches Of course the globally optimal tree s may be several rearrangements away from the starting tree If a rearrangement is successful in finding a better tree a round of rearrangements is initiated on this new tree So long as each round of rearrangements successfully find
363. xa so it is doubtful that exhaustive search strategies will be useful beyond 11 taxa Branch and bound algorithm Fortunately an exact algorithm for identifying all optimal trees that does not require exhaustive generation is available The branch and bound method frequently used to solve problems in combinatorial optimization was apparently first applied to evolutionary trees by Hendy and Penny 1982 This method closely resembles the exhaustive search algorithm described above In this procedure a search tree is traversed in a depth PAUP 3 1 USER S MANUAL 33 first sequence as illustrated in the figure below The root of the search tree A contains the only possible tree for the first three taxa We first construct one of the three possible trees obtained by connecting taxon 4 to tree A yielding tree B1 Then to this tree we connect taxon 5 yielding tree C1 1 If the data set contained more than five terminal taxa we would continue to join additional taxa in this manner until a tree containing all T taxa had been completed Now we backtrack one node on the search tree 1 e back to tree B1 and generate the second tree resulting from the addition of taxon 5 to tree B1 tree C1 2 When all five of the trees derivable from tree B1 C1 1 C1 5 have been constructed we backtrack all the way to tree A of the search tree and take the second path away from this node leading to tree B2 As before all five trees derivable from
364. xa 16 25 to the outgroup constraints ingrp myGenus define an ingroup monophyly constraint You cannot declare a TAXSET name that is the same as any taxon name in the data file Assumption Sets Assumption sets allow rapid assignment of character type and weight assumptions Each assumption set specifies a character type type sets weight weight sets or exclusion status exclusion sets for each character Any number of type weight or exclusion sets may be defined You can then change the assumptions assigned to all characters in the data set simply by invoking a new assumption set Assumption sets are normally defined in the ASSUMPTIONS block however you may also define them from within a PAUP block or via the command line Type sets TYPESETs Type sets assign a character type to each character For example if you wanted the first 10 characters to be of type UNORD the second 10 to be of type DOLLO and the remainder to be of type ORD you could define a corresponding type set as follows typeset mytypes unord 1 10 dollo 11 20 ord 21 If you do not explicitly assign a type to one or more characters the current default character type specified via DEFTYPE in the OPTIONS command of the ASSUMPTIONS block will apply Note that the default DEFTYPE is UNORD so ordinarily you will only need to assign types to those characters to which you want to assign a type other than UNORD Be careful to use t
365. xample is shown below A B C D E A D gt lt Xx A Ne a b Cc D A D A Cc SER Nyc B E B E c d Unrooted constraints a A rooted constraint tree b Unrooted equivalent of the constraint tree c A tree that satisfies the constraints imposed by tree b d A tree that violates the constraints imposed by tree b Topological constraints can also be used to restrict the solution space when searching using exact algorithms Sankoff et al 1982 Constantinescu and Sankoff 1986 If the monophyly of particular groups 1s indisputable the imposition of constraints that enforce these groupings can greatly reduce the number of possible trees that need to be examined thereby extending the limits of usefulness of the exact algorithms Of course the guarantee of optimality is then conditional on the validity of the assumed prior groupings but that sacrifice may be a relatively small price to pay in many situations Yet another use of topological constraints is to enforce ingroup monophyly or partial outgroup structure Recall that when potential outgroup taxa vary for characters that are informative with respect to ingroup relationships one recourse is to include several outgroup taxa in the analysis In this procedure the ingroup node incident to the branch PAUP 3 1 USER S MANUAL 47 partitioning the ingroup from the outgroup becomes the root of the ingroup portion of the tree If however the ingroup and outgroup taxa
366. xonD 111110101101 000 taxonE 1101100111011000 taxonF 111101011111 111 taxonG 101111101011 000 taxonH 101111101011 000 tread 1 tree s from PAUP taxonA taxonB taxonC taxonD taxonG taxonH taxonE taxonF procedure Notice that the outgroup designation in the PAUP data file does not get exported you must reset that yourself either within Hennig86 or by editing the Hennig86 data file The PHYLIP equivalent of this file looks like 8 16 taxonA 000000100101 000 taxonB 0100000011011111 taxonC 1110000011017000 taxonD 111110101101 000 taxonE 1101100111011000 taxonF 111101011111 111 taxonG 101111101011 000 taxonH 101111101011 000 1 taxonA taxonB taxonC taxonD taxonG taxonH taxnE taxonF Excluded and ignored characters are not exported To export all characters in the matrix include any excluded characters and reset the ignore option to ignore none The options for importing are more varied Besides PHYLIP Hennig86 and tab delimited text you can select two formats for molecular data GCG MSF or NBRF PIR If PHYLIP is chosen you have the option of selecting discrete data DNA sequence restriction site data or protein sequence Other data formats can be imported by selecting text in the Import File dialog box You can then select from standard format DNA RNA or protein Using this option gives you great flexibility in analyzing molecular data
367. xport facilities described in the section Importing 142 PAUP 3 1 USER S MANUAL and Exporting Trees and Data provide a more direct means of inputting trees found by PAUP to PHYLIP and Hennig86 Third data files and tree files are standard text files that are portable across computers For large and or complex data sets you may want to perform the searches on a larger more powerful computer and then download the trees to a microcomputer for further analysis graphics output etc You can use the SAVETREES command or the Save Trees to File menu command to save trees to a file the default treefile name is data file name trees You may save either all of the trees in memory or a range of trees that you specify For example the command savetrees from 1 to 10 file trees out brlen ttable will save trees one through ten to the file trees out along with branch lengths and a translation table that maps positive integers to taxon names In essence this creates nothing more than a TREES block contained in a single file Additionally trees may be saved rooted or unrooted although this may slow down the saving process For example savetrees from 1 to 10 root file trees out replace will replace the contents of the file trees out with rooted trees one through ten PAUP also includes information about the analysis that generated the trees should you need to replicate the search Here is a sample treefile resulting from a bran
368. y for a character with 3 or fewer states However stepmatrix characters require significantly more computation than other character types and should be used only if it is not possible to use one of the standard types Stepmatrices may also be used to describe character state trees For example the character state tree B O E AB C E D D can be described using the stepmatrix A BC D amp E Ao am ds Ts Zo lt 3 BZ 2 3 Cr sl 2 D 27 2 P oak E 3 3 2 However there are enormous computational advantages to using an explicitly defined character state tree rather than an equivalent stepmatrix Again use stepmatrices only when another equivalent type is unavailable PAUP 3 1 USER S MANUAL 19 See The ASSUMPTIONS block for details on how to define a stepmatrix in an input file It is possible to define stepmatrix characters that violate the triangle inequality This inequality is a property of distances in Euclidean space where the length of one side of a triangle is always less than the sum of the lengths of the other two sides This means the shortest distance between two points will always be a straight line If a stepmatrix is defined that violates this rule one side is actually longer than the sum of the other two sides For example the following triplet violates the inequality as the distance b c is longer than the sum of the distances a b and a c This would result from the following st
369. you specify outgroup rooting but do not specify an outgroup PAUP will take the first taxon in the matrix as the default outgroup Outgroup rooting is achieved by using the ROOT OUTGROUP option and perhaps specifying an OUTROOT option as well The different outgroup rooting options are also available from the Rooting menu command For example the following command would describe a tree and root it such that the outgroup is monophyletic relative to the ingroup describe 1 root outgroup outroot monophyl Other options for outgroup rooting output the outgroup in a basal polytomy OUTROOT POLYTOMY or as a group paraphyletic to the ingroup OUTROOT PARAPHYL Remember that if the tree cannot be rooted such the specified ingroup is monophyletic PAUP will include as many outgroup taxa as possible with the ingroup to form a monophyletic group though the primary outgroup is never included Lundberg rooting involves computing an unrooted tree for the ingroup taxa only and then attaching an outgroup taxon or hypothesized ancestor to the tree a posteriori Lundberg 1972 This is achieved by using either the ROOT LUNDBERG option or the appropriate selection in the Rooting dialog box The tree can be output with Lundberg rooting with a command like the following 140 PAUP 3 1 USER S MANUAL describe 1 root lundberg Lundberg rooting forces the use of MINF optimization the only available optimization that is independent of the location of the
370. zero length branches and keep only those trees that are unique after the collapsing is accomplished Selecting Condense Trees brings up the following dialog box wy Collapse branches for which maximum possible length is zero EX Eliminate duplicate trees The Condense Trees dialog box By default PAUP collapses branches whose maximum length is zero yielding polytomous trees However you may choose to override this default so that only completely bifurcating trees are produced during a search by clicking on the Collapse branches for which maximum possible length is zero checkbox You can also have PAUP delete duplicate trees in memory by clicking on the Eliminate duplicate trees checkbox Root Trees Deroot Trees Root Trees and Deroot Trees are specialized commands that are ordinarily unnecessary However if you have unrooted trees stored in memory and you change one or more character types to a type that requires rooted trees e g irreversible characters the trees will have to be rooted before they can be output Trees are rooted via outgroup rooting using the outgroup in effect at the time the command is issued If you are using Lundberg rooting you can select the ancestor with the pull down menu Selecting Rooting Options will bring up the standard Rooting dialog box 262 PAUP 3 1 USER S MANUAL Rooting trees is ordinarily needed only when directed characters are in effect Are you sure you

PAUP 3.1 Manual

Contents

Download Pdf Manuals

Related Search

Related Contents