BioClust User's Guide - Accueil Plateforme Biopuce de Toulouse
Contents
1. Systematic name only systematic name is used as gene ID Sys name short name both names are used and they are separated by semi column and a space Full gene name is self explanatory User s code idem One check box option can modify any choice made in menu append well will append the plate coords of well containing the gene e g ACTI 1A12 append user s code idem for user s code lFor other organisms GeneBank entries are often used as systematic names append fullname idem for full gene name Table content menu Options are Ratios on first column This option is equivalent to membrane mode in BioPlot An average value of gene intensities is used to calculate the final ratio of each column over the first one Ratios on coupled analysis channel This option is equivalent to slide mode in BioPlot A ratio is calculated for each spot for all couples of analysis A couple is made of analysis coming from the same slide The final ratio appearing in columns are the two stage average of spot ratios first stage over spot replicates second stage over slides Expressions Up 1 Equal 0 Down 1 According to this option the re sulting table will be composed of expression labels 1 0 and 1 The decision about over or under expression of each gene is done following the selection in BioClust gt Stats gt Expression changes based on Variable values This option makes the result table2. You can define a rectangular region by defining the starts and ends for rows and columns as fol lows lt row start gt lt row end gt lt space gt lt col start gt lt col end gt For exam ple 12 15 23 40 defines an array region having rows between 12 and 15 and columns between 23 and 40 Any of region border start or end is optional in which case the limit value 1 for start maximum for end is taken Thus 12 40 defines a region with rows from 12 to the max row and columns from to 40 Apply post normalization factor checkbox Yes If checked all spots in every anal ysis are multiplied by a user defined factor This is used in very particular cases Don t check this box if you don t need to correct normalized data by hand This factor one per analysis can be defined on the Browse analysis page accessible from a tool frame on the left of user s web space 10 Subtract background menu Apply or not background correction Options are e no e local background e average of negative spots e local negatives corrected by their bg Last two options may be useful for membranes with clones grown on them and having DNA sequences inserted in their own DNA Such array technique can lead to a situation where clones with a void insert have a signal above the back ground These negatives control should have signal as close to zero as possible so it is more appropriate to correct by the mean of negative controls than b
3. a user starting to work with BioClust it may be useful to start by BioPlot User s Guide http biopuce insa toulouse fr ExperimentExplorer doc BioPlot pdf which presents some theoretical considerations about transcriptome experiment BioClust does not provide any evolved graphical tool for clustering manipulation but it can be very helpful to prepare data which can be submitted to a software spe cialized on clustering or other multivariate statistical methods see for example Tigr Multi Experiment Viewer tmev on http www tigr org The rest of this guide is a detailed description of form fields which may be found in BioClust The reader is supposed to be familiar with theories and techniques employed in transcriptome experiments as well as with some advanced statistical techniques like clustering and Linear Discriminant Analysis LDA Chapter 1 BioClust Form Items In this chapter we describe the parameters that user can tune in BioClust to analyze transcriptome data corresponding to two or more conditions referred as X where i is condition number To analyze only two biological condition BioPlot provide reacher analysis environment including scatterplot zoom on image and link to external genome databases BioClust is a web service on http biopuce insa toulouse fr Users have personal accounts and protected access to their data When user connects and chooses BioClust in tool frame at left hand side a BioClust form is
4. contain the intensities not ratios The variable quantifying spot intensity may be chosen in Bio Clust gt Variable Norm gt Variable to use append P value checkbox if checked this option add columns with P values to columns requested as Table content Columns with P values are alter nated with others columns such that a gene has its ratio and corresponding P value for a given condition side by side Hide first column checkbox Yes Check this box if the selected option in previous menu is Ratios on first column Thus the first non informative column having only 1 or 0 depending on the option Variable Norm gt Take log10 is not shown Result type menu Options are html table with colors According to this option the cells of resulting html table will be colored in red or green depending on over or under expression of a given gene in a given biological condition If we cannot test the ex pression change because of data lack for example the corresponding cell will have gray background html table The same table but without colors in cells plain text This option is useful for exporting data from BioClust to any other software h clust by R hclust mva The result will be an image with a dendro gram corresponding to hierarchical clustering This clustering is done by a function hclust from the library mva which is part of R script language see http www r project org or h
5. involved in gene dye interaction Sorting criterion menu User can order data according to one of the following crite rion e Expression e Well s position e Membrane position by line e Membrane position by column e Gene id e Functional category e GO biological process e GO cellular component e GO molecular function Note that Expression in this menu means a discrete variable taking three values 1 0 and 1 depending on expression change test Ordering by Expression will form gene groups having the same expression profile These groups are easily detectable in html page if colors are used to highlight expression changes Ordering menu Options are e ascending e descending 1 7 An profiles tab Save current analysis parameters as text file name in which it will be stored the current set of analysis parameters such like type of normalization filtering rules and so on A set of analysis parameters is called analysis profile and storing it in a file can be useful to gain an effort when a user has to analyze an exper iment series in same conditions It diminishes also an error risk in choosing parameters Analysis profile files are stored in anprof bioclust subdirectory of user s space To choose an existent file or a subdirectory user can click on a button with ellipsis All files have kvh extension If user omits this exten sion in file name it will be added automatically There are two predefined
6. names The names entered in this field should be in accordance with the choice in Analysis gt Gene ID menu The names can be separated by comma by tabulation or by new line character It is perfectly possible to copy and past a gene list from an other application such as a spreadsheet The gene name can have a wild character which replaces any sequence of characters Gene names are not case sensitive For example y selects all genes starting by y or Y The wild character alone has a particular sense If entered it selects all spots corresponding to some gene id Spots don t having gene identity are excluded from selection By stored gene list text User can give a file name with stored gene list This will se lect the genes of this list User can store a gene list both in BioPlot and BioClust At the end of result table there is a form to fulfill to store currently shown gene list A button can be used for help in file choosing By Xrer X 2 cutoff when Xrer X 2 is menu number and other menu This is excluding field Here you define the spots to exclude from from results Very low spots may be annoying to see in the final results They can exhibit very high ratios but they are not very reliable When some real number is entered in text field the result of exclusion can be read as for example cutoff when Xrer X 2 is less or equal to lt some number gt in all columns Another choice i
7. profiles 16 predefined mbr_normtot_exp kvh and predefined slide_normtot_expr kvh for membrane and slide mode of averaging respectively These profiles enable e use of pixel mean as spot intensity e local background correction e log transformation of intensities e normalization by total mean of all spots e averaging of spot replicates by well e expression decision based on Student s test and ratio thresholds e no filtering rules e sorting by expression taking values in 1 0 1 NA in descending order Predefined parameters can be a start point for defining user s own profiles They cannot be overwritten When a file name is chosen user has to click on Save button to actually write the file Read analysis parameters from text file name from which an analysis profile will be read Delete analysis parameters text file name or a void subdirectory to be deleted 17 Troubleshooting Warning messages When user makes some inconsistent choice the result page can contain one or more warning messages which explicate the inconsistency Normally after reading the warn ing message in a particular context it becomes clear what selection should be changed to remove inconsistency Other problems In case of e the warning message is not sufficiently explicit e there is problem with BioClust that you cannot resolve by yourself e you would like to request a particular feature to be added to BioClust e you want to
8. report a bug in BioClust e there is a correction to do on this documentation you can contact Serguei Sokol by email sokol a insa toulouse fr or by phone 33 0 561 55 96 87 In case of a problem with your analysis it can be helpful to include in your email a copy of a link under the result table in BioClust if any You can click with right button on this link a choose Copy link location from contextual menu and past it in email composing window 18 Credits The work on BioClust and on documentation has benefited from various financial sup ports from INRA and CRITT Bioindustrie Platform director Jean Marie Francois and platform technical director V ronique Le Berre helped the author to understand tran scriptome experiments and by fruitful discussions contributed to this documentation 19
9. to apply log transformation So to prevent too low values they remain at lowest allowed value defined in this field Don t confound this feature with cut off by intensity No spot is cut here Conversely due to this feature low spots are preserved in final list To disable such preserving cancel any content of this field don t leave 0 1 4 Gene Selection tab Filters defined in different fields of this tab work like intersections i e if you select 66 spots coming from first plate and the genes whose name start by a you obtain genes 79 starting by a coming only from the first plate Conversely in the same field your 11 selection is considered like union i e if you select first and second plate to put on the result table than genes coming from these two plates will be shown An intersection in this case would be meaningless as the intersection of first and second plate is void When applicable any selection may become exclusion if preceded by tilde By plate s region text You can enter a comma separated list of plate coords using the same coord convention that for the field Variable Norm gt Control gene co ords By membrane s slide s region text Use a comma separated list to define regions of interest on your arrays The coord conventions are the same as for the field Variable Norm gt Control gene coords By name text You can select one or more genes by their
10. BioClust User s Guide by Serguei Sokol sokol a insa toulouse fr First release 24 05 2004 Last updated 07 09 2007 platform Biopuce INSA DGBA 135 av de Rangueil 31077 Toulouse cedex France https biopuce insa toulouse fr Copyright C 2004 platform Biopuce Toulouse Genopole Permission is granted to freely print copy translate or distribute this document for educational purposes pro vided this copyright notice is preserved The original and most upto date copy of this document can be found on the web page http biopuce insa toulouse fr ExperimentExplorer doc Contents 1 BioClust Form Items 4 1 1 Analysis dis 4 ao Ge ea ee ee aa ee eS 4 1 2 Cl stermpetaDle ia is a Se a RR Hos Bed SE A 6 1 3 Variable Norm tab s aaeeea e a ee 8 1 4 Gene Selectiontab E y E UE E eee 11 1 5 Category selection tab oaoa oo 13 1 07 Stats tabes m 4 2 a ag niet Beet BO MA Bom ad ote a aola GA a 14 1 7 Ad profiles tabio wiesen ne g Eee EE Ss 16 Introduction BioClust is a web service hosted by platform Biopuce which is a part of Toulouse Genopole and is located in INSA DGBA This service is aimed to help platform users to compare transcriptome data from multiple biological conditions and select significantly changed genes according to a desired expression pattern BioClust shares with BioPlot same data treatment methods such like normalization Student s test or gene selection For
11. Maximal Intensity Minimal Intensity Sum Intensity Background level Background Median Background Mean Spot Standart Deviation Spot Variance Background Standart Deviation Total Background pixels superior to background pixels superior to 1 5 background pixels superior to 2 background gaussian pixels superior to background gaussian pixels superior to background gaussian pixels superior to background statistical pixels superior to background statistical pixels superior to 1 5 background statistical pixels superior to 2 background pixels superior to Bg SD pixels superior to Bg 2 SD pixels saturated Variable Selected during Image Analysis is particular Some image analysis applications offer normalization features or other data treatments For this rea son this variable can not be corrected by background This option is deprecated and will be removed in the future Normalization type menu Choices are e no normalization e housekeeping genes spike control e stable majority by histogram e all spot s mean e all non negative mean e lowess e quantile Use spots marked Reference as control genes checkbox Yes This option used only when housekeeping genes spike control is selected in previous menu If checked the spots having attribute Reference set are used to calculate normal ization coefficient To mark spots as references see the BioPlot User s Guide for in
12. additional click on matching button back to th column button This button is visible only from the second column se lection and allows to return to the selection of precedent column Selected integer This is a recall of how many arrays are already selected for current column put in 1 st column button Click on this button when you have finished the selection of analysis for the first column You will be presented almost identical Analysis tab The multimenu for analysis selection will take name 2 nd column this button will become put in 2 nd column and a button back to 1 st column will appear So you will be able to select analysis for second third and so on column Finish button This button is visible in all tabs It should be clicked only after have set all parameters in all tabs To pass to other tabs click on the corresponding tab name 1 2 Clustering tab Gene ID menu Historically the gene naming is not something unique and standard So various choices are possible In this description we use the following terms for gene names short name like ACT for actin systematic name like YFLO39c for yeast actin full name like actin user s code whatever user has used to identify his spotted material The options of this menu are Default short name if exists systematic name otherwise Short name only short name is used If it is void than gene ID will be void in resulting list
13. ch is composed of at least two columns Thus columns belonging to the same group will have the same label The purpose of LDA is to find linear combination of observed variables here gene expressions which have the most discriminant power for submitted groups LDA maximizes Fisher statistic i e the ratio inter group variance over intra group variance The variances are calculated according to some linear com bination of observed variable LDA gives up a sorted list of genes according to their weight in linear combination which can be interpreted as importance or contribution to the discrimination The main LDA result is a sorted table of genes The ordering is done according to the contribution of a given gene to the first discriminant variable LD1 which maximize the ratio mentioned above Thus the most discriminant genes are in the top of the table Absolute values of values reported in the column LD1 can be considered as contribution weights of a given gene to LD1 We report a Fisher statistics i e ratio inter intra variance as well as corre sponding P value for every discriminant variable Naturally it is the first variable LD1 which is the most important for us For example if P value correspond ing to LD1 is low say lt 0 05 than LD1 is a good discriminator thus the most contributing genes to LD1 are in their turn good discriminators of biological diversity For more information on LDA see for example http www i
14. erplot pairs This kind of presentation may be useful for quality con trol User can select a set of arrays one by column and see the scatterplots of all possible column pairs The values reported on scatterplots depend on the choice in menu Table content upper in this tab For exemple if Table content gt Ratio on coupled analysis channel is chosen then log ratios on each slide will be compared with log ratios of all other se lected slides Correlations anti correlations or other kinds of point clouds observed on such scatterplots may be relevant for potential problems of labeling scanning and so on 1 3 Variable Norm tab Variable to use menu This menu gives the list of variable to quantify spots Most used variables are Mean Intensity and Median intensity Mean and median are calculated over pixels of a spot by image analysis software These and others spot measures are imported in data base from text files generated after image analysis As various applications are used for spot detection and quantification and they measures their own statistics not all choices are meaningful for all analysis The full list of options in this menu is e Variable Selected during Image Analysis e Mean Intensity Median Intensity Weighted Mean Intensity Mean Gaussian Intensity Median Gaussian Intensity Mean Statistical Intensity Median Statistical Intensity Most frequent pixel intensity Central Intensity
15. n the first menu is greater or equal which is complementary of the option cited in the example and can be useful to visualize the low spots that are cut The sec ond menu has an option in at least one column and mean that a spot will be excluded from the result if it is too low simultaneously in control and any even only one test condition By expression pattern menus All menus one per defined column have the same options relative to expression decision e All e Up 12 e Norm e Down Setting these menus to desirable values user can extract one or several expres sion profiles By intensity Put spots only if pixels gt bground real number checkbox If checked spots having the percentage of pixels above the background lower than defined threshold are excluded from scatterplot and from result table The statistics percentage of pixels above the background is measured and provided by im age analysis application This option is not very reliable to eliminate low spots It is preferable to use By X ef X 2 field By expression checkboxes This is another excluding field The choices are e Exclude over expressed genes e Exclude normally expressed genes e Exclude under expressed genes e Exclude unknownly expressed genes e Exclude unknownly normally expressed genes The decision if a gene is over or under expressed is related to the field Stats gt Ex pression changes based on Note that if y
16. ningful only when the replicate number is relatively high e g 11 like on Affymetrix slides 14 Spot aggregate menu Options of this menu are e by well e by gene name The most currently used option is by well However by gene name can be useful if plate plan of compared analysis are different Another potential application of this option is for eliminating of multiple entries in result table if some genes are present in more than one well This option is mandatory if BioClust gt Clustering gt Result type is set to h clust by R hclust mva as hclust doesn t support neither repeated nor unnamed entries Expression changes based on menu This option is without effect if no checkbox is checked in BioClust gt Gene Selection gt By expression Choices in the menu are e ratio thresholds e Student s test e ratio AND Student s test The first option ratio thresholds is intuitive and seems simple for biologists Unfortunately this method can result in a high number of false positives It is highly recommended to combine this filter with Student s test When Student s test is selected user is exposed to multiple tests problem dis cussed in BioPlot User s Guide So the combination of ratio thresholds and Student s test is probably the most appropriate choice To be able to apply Student s test user has to supply at least two independent array repetitions However the po
17. opened in main frame The names and comments in the form are intended to be self explanatory provided that user is familiar with transcriptome data treatment methods The form is organized in following tabs Analysis where analysis to treat can be selected and distributed among columns one biological condition per column Clustering where user makes his choice on data treatment method and on kind of result presentation Variable Norm where user chooses a quantification variable and selects options on background correction normalization and log transformation Gene Selection regroups options relative to filtering data Cat Selection Category selection is intended for gene filtering by ad hoc functional categories and by categories defined in Gene Ontologies GO Stats has fields concerning statistical data treatment BioClust has not extended graphical possibilities If you need an elaborated graph you have to export data from BioClust to some software having such possibilities In following sections we review all form fields 1 1 Analysis tab Group name for LDA text This entry is for labeling groups of columns It should be filled only if you want to proceed Linear Discriminant Analysis LDA treating columns like individuals and rows gene expressions like observed variables Usually for LDA you have to put one analysis per column and to label each col umn by a group label You need to define two or more groups each of whi
18. ou exclude for example normally expressed gene then only genes normally expressed in all conditions will be ex cluded If a gene has changed expression in at least one condition it will be preserved in results By type menu Options of this menu are e all e reference e negative e not negative This can be useful to visualize the spots genes corresponding to any chosen type To see reference genes may be of particular interest if the normalization by stable majority is used In this case a users don t know a priori what are the genes that are selected as stable majority So this option lets visualize them 13 1 5 Category selection tab Features of this tab are identical to those of BioPlot gt Cat Selection Category selection or for sake of brevity cat selection is composed of four fields one for gene filtering by ad hoc functional categories and three others for filtering by Gene Ontologies GO To use these options user has to provide an annotation information for his organism of interest The platform data base administrator has to treat and add this information prior to data analysis By functional category text and select button user can provide a list of category me names separated by or at his convenience he can use a select window which opens after clicking on button User has to use and not a coma to sep arate categories because some categories have a coma in their names Catego
19. ries are organized in a hierarchical tree like structure The tree branches are rolled or unrolled by clicking on a blue triangle By checking a box corresponding to a functional category user select all genes of this category and its subcat egories The name of the selected category is automatically added to the text field Unchecking a category removes its name from the text field User is al lowed to use a meta character to replace any portion of category name when entering the name in the text field By GO biological process text and select button By GO cellular component text and select button By GO molecular function text and select button All three fields are functioning in similar way to functional categories The main difference is that gene ontologies are defined by GO consortium in the same way for all organisms So there is no need to select an organism before selecting a category as it is the case for ad hoc functional categories Maximum hierarchical level to show menu As its name indicates this menu con trols the level of tree expanding when showing the category hierarchy both in select window and in tabulated results when ordering by a functional category or a gene ontology 1 6 Stats tab Spot aggregate function menu Options are e average e bi weighted mean Duplicate aggregation may be done in classic way by simple average or in more elab orated way using bi weighted mean This last method is mea
20. sip msstate edu publications reports isip_internal 1998 linear_discrim_analysis lda_theory_v1 1 pdf or in French http www lsp ups tlse fr Besse pub sdm2 pdf look for Analyse Factorielle Discriminante 1 st column multiselect menu In this menu user chooses biological conditions which may be interpreted as control conditions depending on the choice in BioClust gt Clustering gt Table content see description of this item here after for more details After you have clicked on put in 1 st column this multimenu will become 2 nd column and so on up to a desired column number When an option BioClust gt Clustering gt Table content described here after is set to Ratio on coupled analysis channel analysis chosen in this menu are treated as test analysis i e the ratios are calculated as intensities from selected analysis over the intensities coming from complementary channel on the same slide The last are detected automatically by BioClust matching button text field click on this button to make the menu contain only anal ysis whose names are matching the content of the neighbor text field This is a practical mean to reduce analysis choice This feature is particularly useful when the number of analysis become very important Note that on recent browsers Netscape 7 or InternetExplorer 6 the menu content becomes conform to the text field at each new entered letter such that you don t need
21. structions which can be found in Tab View description Checking this box puts the option housekeeping genes spike control as selected in previous menu Control genes coords text This option used only when housekeeping genes spike control is selected in Normalization type menu and previous checkbox is not checked This field can contain a comma separated list of well or spot coords corresponding to genes that should be used as references in normalization Well or plate coords are given as lt plate nb gt lt letter gt lt number gt e g 1A14 means plate 1 well A14 lt letter gt and lt number gt are optional If number is omitted the whole line is used If lt letter gt and lt number gt are omitted the whole plate is selected It is possible to define a rectangular region on a plate by introducing the left upper and right low conner well coords For example 2C3 D12 defines a region on the plate 2 containing wells in rows C to D and between columns from 3 to 12 Spot or array coords are of the form lt row gt lt space gt lt column gt where lt row gt and lt column gt are integers starting at 1 For instance 11 23 defines a spot at row 11 and column 23 All arrays in our data base are oriented to have a spot corresponding to the well 1A1 in upper left corner This orientation is 90 rotated compared to vertically oriented slide scans Usually on vertical slide images the well 1A1 has its spot in upper right coin
22. ttp www lsp ups tlse fr Besse pub TP r tpintro ps for a brief introduction in French R func tions may need some number of parameter settings which are not accessi ble from BioClust The resulting dendrogram corresponds to default op tions i e Euclidean distance and average link Here the dendrogram is provided only for quick visual analysis and is limited to 500 genes A thorough study should be conducted in R or any other statistically ori ented software like tmev on www tigr org For more information on hclust see help hclust in R For an introduction in clustering see for example http www statsoftinc com textbook stcluan html or http www lsp ups tlse fr Carlier Hyper polyclass nodel html in French heatmap by R heatmap mva Hierarchical clustering is performed both on genes and conditions like in previous item The results are presented in so called heatmap form where each table cell is coded is some color green for low values red for high values Such presentation facilitates a visual detection of dissimilarities between genes and condition This can be useful for a subjective quality control e Lin Discr An by R lda MASS This option defines a result type which is provided by a Ida function from library MASS available in R mentioned above The purpose of Linear Discriminant Analysis is briefly presented in the description of the field BioClust gt Analysis gt Group name for LDA on page 4 Scatt
23. wer of statistical test on only two repeti tions may be deceptive You can estimate the number of repetitions on http biopuce insa toulouse fr microarray exp_numb php Overexpression threshold gt 0 positive real number This field is used when one of checkboxes BioClust gt Spot Selection gt By expression is checked and an option selected in the precedent menu concerns ratio Uderexpression threshold gt 0 positive real number Same remarks P threshold in Student s test 0 lt P lt 1 positive real number less than 1 This field is used when one of checkboxes BioClust gt Spot Selection gt By expression is checked and an option selected in the menu Expression changes based on in this tab concerns Student s test Error estimate without gene dye interaction radio check boxes Three options of this field can be used to choose the way to treat an error estimate in dye switch experiment e No default corresponds to classical Student s test with classical error estimate for all genes 15 e Yes for all genes means that error will be estimated without gene dye effect contribution for all genes e Yes for genes giving lower P value corresponds to automatic choice between two kind of error estimate The error estimate canceling gene dye effect is used only if it gives better P value otherwise a classical Student s test is used This treatment avoid the loss of statistical power for genes not
24. y local background In other situations local background should be good choice Use average of spots marked Negative as bg level to subtract checkbox Yes This checkbox is taken into account only when one of the last two options of prece dent menu is chosen For marking spots as negative see BioPlot User s Guide When this option is checked it puts average of negative spots as selected in precedent menu Negative gene coords text If average of negative spots or local negatives cor rected by their bg options of Subtract background menu are selected this field can define the spots that should be used as negative controls User can enter a comma separated list of well or array coords following the same coord conventions as in Control gene coords field on this tab Take log10 radiobox Choices are No Yes and Yes but final results are in orig inal scale If the last option is chosen the log transformation is applied and all statistical treatment is done on log transformed values Only final results will be back transformed from log to original scale for easier interpretation of ratio values It is highly recommended to apply log transformation to bring better statistical properties to data passed through statistical tests Lowest value limit real number It may happen after the background subtraction that a spot intensity become zero or negative This is not desirable if we want
Download Pdf Manuals
Related Search