LMGene User's Guide
Contents
1. 3 View the data structure of the sample data and the details of exprs and phenoData slots in the data gt slotNames sample eS 1 exprs se exprs description annotation notes 6 reporterInfo phenoData gt dim sample eS exprs 1 613 32 gt sample eS exprs 1 3 pidO pidi pid2 pid3 p2d0 p2d1 p2d2 p2d3 p3d0 p3di p3d2 p3d3 p4d0 p4di p4d2 g1 216 149 169 113 193 172 167 168 151 179 142 156 160 214 157 g2 334 311 187 135 514 471 219 394 367 390 365 387 318 378 329 g3 398 367 351 239 712 523 356 629 474 438 532 427 429 574 419 p4d3 p5d0 pddi p5d2 p5d3 p6 dO p6di p6d2 p6d3 p7dO p7di p7d2 p7d3 p8d0 p8d1 g1 195 165 144 185 162 246 227 173 151 796 378 177 278 183 285 g2 450 293 285 390 428 645 631 324 343 852 451 259 379 259 386 g3 564 438 321 519 488 824 579 416 489 1046 501 375 388 373 509 p8d2 p8d3 gl 275 202 g2 361 333 g3 468 436 gt sample eS phenoData phenoData object with 2 variables and 32 cases varLabels patient patient dose dose gt slotNames sample eS phenoData 1 pData varLabels varMetadata Data generation If you don t have exprSet class data you need to make some LMGene provides a function that can generate an object of exprSet class assuming that there are array data of matrix class and experimental data of list class 1 The package has sample array and experimental data sample mat and vlist gt data sample mat gt dim sample mat 1 613 32 gt data vlist2. g84 g85 11 g86 g93 g102 g123 g139 g155 g178 g179 g205 g250 21 g256 g271 g277 g304 g310 g319 g327 g336 g372 g375 31 g384 g399 g405 g406 g407 g408 g409 g410 g411 g412 41 g413 g414 g415 g423 g425 g460 g461 g462 g463 g477 51 g503 g520 g524 g528 g566 g607 g612 dose 1 No significant genes patient dose 1 No significant genes The routine LMGene requires the multtest package References 1 Durbin B P Hardin J S Hawkins D M and Rocke D M 2002 A variance stabilizing transformation for gene expression microarray data Bioinformatics 18 S105 S110 Durbin B and Rocke D M 2003a Estimation of transformation parameters for microarray data Bioinformatics 19 1360 1367 Durbin B and Rocke D M 2003b Exact and approximate variance stabilizing transfor mations for two color microarrays submitted for publication Geller S C Gregg J P Hagerman P J and Rocke D M 2003 Transformation and nor malization of oligonucleotide microarray data Bioinformatics 19 1817 1823 Rocke David M 2004 Design and Analysis of Experiments with High Throughput Biological Assay Data Seminars in Cell and Developmental Biology 15 708 713 Rocke D and Durbin B 2001 A model for measurement error for gene expression arrays Journal of Computational Biology 8 5
3. gt trsample eS lt transeS sample eS tranparmult lambda tranparmult alpha gt trsample eS exprs 1 3 1 8 pido pidi pid2 p1d3 p2do p2d1 p2d2 p2d3 g1 5 686954 5 424873 5 449682 4 549380 5 590642 5 418542 5 268332 5 347915 g2 6 272797 6 308464 5 592073 4 915159 6 811348 6 710929 5 693269 6 492140 g3 6 488757 6 493737 6 388361 5 832776 7 173087 6 830052 6 345199 7 029530 It s also possible to estimate the parameters using the more accurate lowess normalization as opposed to uniform normalization gt tranparmult lt tranest sample eS ngenes 100 mult TRUE lowessnorm TRUE gt tranparmult lambda 1 655 1817 alpha 1 72 33459 54 85916 59 95882 67 84961 65 76304 72 80078 74 63962 8 59 91484 54 35704 69 41260 71 14899 62 35461 61 59616 75 55289 15 59 27235 86 55559 61 67145 59 51983 63 81498 62 58284 60 05210 22 99 95130 58 68123 61 70271 171 23313 119 38508 57 23296 73 56054 29 65 25109 98 40390 67 89693 63 68758 It is even possible now to estimate parameters using a specified model For example if we think that the interaction of variables in vlist is important we can add interaction to the model gt tranpar lt tranest sample eS model patient dose patient dose gt tranpar lambda 1 860 0836 alpha 1 55 68625 The model is always specified in the same way as the right hand side of an Im model In the example above we set the parameters to minimize the mean squared error f
4. gt vlist patient 1311112222333 344445555666677778888 Levels 1234567 8 dose f1101230123012301230123012301230123 2 Generate exprSet class data using neweS function gt test eS lt neweS sample mat vlist gt class test eS 1 exprSet attr package 1 Biobase gt identical sample eS test eS 1 TRUE c f If you have different types of array data such as RGList marrayRaw and so on you can convert them into exprSet class by using as method after installing convert package 3 G log transformation 1 Estimating parameters for g log transformation The linear model is not applied to the raw data but to transformed and normalized data Many people use a log transform LMGene uses a log like transform involving two parameters We estimate the parameters A and a of the generalized log transform log y a y a sinh t 52 log A using the function tranest as follows gt tranpar lt tranest sample eS gt tranpar lambda 1 726 6187 alpha 1 56 02754 The optional parameter ngenes controls how many genes are used in the estimation The default is all of them up to 100 000 but this option allows the use of less A typical call using this parameter would be gt tranpar lt tranest sample eS 100 gt tranpar lambda 1 874 2464 alpha 1 55 54313 In this case 100 genes are chosen at random and used to estimate the transformation param eter The
5. 57 569 Rocke D and Durbin B 2003 Approximate variance stabilizing transformations for gene expression microarray data Bioinformatics 19 966 972
6. LMGene User s Guide Geun Cheol Lee John Tillinghast and David M Rocke October 25 2006 Contents 1 Introduction 1 Data preparation 1 3 G log transformation 3 4 Finding differentially expressed genes 6 N 1 Introduction This article introduces usage of the LMGene package LMGene has been developed mainly for analysis of microarray data using a linear model and glog data transformation in the R statistical package 2 Data preparation LMGene takes objects of class exprSet which is the standard data structure of the Biobase package Hence if data which is exprSet class is ready the user can jump to further steps like diagnostic plotting or g log transformation Otherwise the user needs to generate new exprSet class data For more detail please see the vignette Textual Description of Biobase in the Biobase package Note exprSet In this package an object of exprSet class must contain exprs and phenoData slots with proper data Example LMGene includes a sample array data which is of class exprSet Let s take a look this sample data 1 First load the necessary packages in your R session gt library LMGene Loading required package Biobase Loading required package multtest Loading required package survival Loading required package splines Loading required package survival gt library Biobase gt library tools 2 Load the sample exprSet class data in the package LMGene gt data sample eS
7. or a regression of transformed gene expression against patient log dose and their interaction Be very careful of using interactions between factor variables If you do not have enough replications you can easily overfit the data and have no errors to work with Naturally it s possible to use mult lowessnorm and model all together 4 Finding differentially expressed genes 1 Transformation and Normalization Before finding differentially expressed genes the ar ray data needs to be transformed and normalized gt trsample eS lt transeS sample eS tranparmult lambda tranparmult alpha gt ntrsample eS lt lnormeS trsample eS 2 Finding differentially expressed genes The lmgene routine computes significant probes using the method of Rocke 2003 A typical call would be gt sigprobes lt LMGene ntrsample eS There is an optional argument level which is the test level 05 by default A call using this optional parameter would look like gt sigprobes lt LMGene ntrsample eS level 0 01 The result is a list whose components have the names of the effects in the model The values are the significant genes for the test of that effect or else the message No significant genes As with tranest it s possible to specify a more complex model to LMGene gt sigprobes lt LMGene ntrsample eS model patient tdoset patient dose gt sigprobes patient 1 g2 g3 g9 g10 g14 g15 g49 g54
8. routine returns a list containing values for lambda and alpha G log transformation Using the obtained two parameters the g log transformed expres sion set can be calculated as follows gt trsample eS lt transeS sample eS tranpar lambda tranpar alpha gt sample eS exprs 1 3 1 8 pidO pidi pid2 pid3 p2d0 p2d1 p2d2 p2d3 gi 216 149 169 113 193 172 167 168 g2 334 311 187 135 514 471 219 394 g3 398 367 351 239 712 523 356 629 gt trsample eS exprs 1 3 1 8 pido pidi pid2 p1d3 p2do p2d1 p2d2 p2d3 g1 5 779555 5 254781 5 441132 4 804638 5 627829 5 466407 5 423932 5 432568 g2 6 325217 6 239533 5 584240 5 101309 6 822052 6 723790 5 797778 6 519446 g3 6 531151 6 436654 6 384164 5 911558 7 180511 6 841453 6 400864 7 045494 Tranest options multiple alpha lowessnorm model Rather than using a single alpha for all samples we can estimate a separate alpha for each sample This allows for differences in chips in sample concentration or exposure conditions gt tranparmult lt tranest sample eS mult TRUE gt tranparmult lambda 1 689 2819 alpha 1 69 67146 37 02711 54 13904 69 35728 60 33270 60 75301 71 72965 8 64 55506 58 63427 65 73625 48 40173 59 43778 76 34568 78 81046 15 82 20326 96 19938 77 60070 79 48089 73 63257 73 41650 33 86029 22 69 26448 55 75460 54 29840 139 89493 91 36521 46 46158 59 02056 29 73 60255 89 48728 57 13887 64 98866 For vector alphas transeS uses exactly the same syntax
Download Pdf Manuals
Related Search