Hello, I have five groups of samples that I have microarray data for and I would like to know what genes make that group distinct from the other groups. The initial approach I took was to use limma to do a comparison between one group and the rest, for example: design1 <- model.matrix(~pData(alignedSetlur)$group ==1) fit <- lmFit(alignedSetlur,design=design1) fitone <- eBayes(fit) topTable(fitone,coef=2,n=Inf,p.value=0.05) I did this for each group, but I found that there was a large amount of overlap between the significant genes in one group and another (over 50% in some cases) and this doesn't seem correct. What is the best way to answer this sort of question? Some sort of multivariate analysis or PCA? Many thanks Dan -- ************************************************************** Daniel Brewer, Ph.D. Institute of Cancer Research Molecular Carcinogenesis Email: daniel.brewer at icr.ac.uk ************************************************************** The Institute of Cancer Research: Royal Cancer Hospital, a charitable Company Limited by Guarantee, Registered in England under Company No. 534147 with its Registered Office at 123 Old Brompton Road, London SW7 3RP. This e-mail message is confidential and for use by the a...{{dropped:2}}

On Wed, Apr 22, 2009 at 11:42 AM, Daniel Brewer <daniel.brewer@icr.ac.uk>wrote: > Hello, > > I have five groups of samples that I have microarray data for and I > would like to know what genes make that group distinct from the other > groups. > > The initial approach I took was to use limma to do a comparison between > one group and the rest, for example: > > design1 <- model.matrix(~pData(alignedSetlur)$group ==1) > fit <- lmFit(alignedSetlur,design=design1) > fitone <- eBayes(fit) > topTable(fitone,coef=2,n=Inf,p.value=0.05) > > I did this for each group, but I found that there was a large amount of > overlap between the significant genes in one group and another (over 50% > in some cases) and this doesn't seem correct. > > What is the best way to answer this sort of question? Some sort of > multivariate analysis or PCA? > Hi, Dan. You might want to look at decideTests() in limma. Also, some of the machine learning algorithms (randomForests, for example) will generate class-specific importance measures for features. Sean [[alternative HTML version deleted]]