Dear all, I have a question on a subject that I think was not discussed here before. I am using limma package for the detection of significant differential expression in an affy experiment: 3 "Healthy" (group1) indiv. before treatment and the same indiv. after treatment (group4) 3 "ill-low" (group2) " " " " " (group 5) 2 "ill-high" (group3) " " " " " (group 6) the interest is comparing effects of the treatment (i.e. group4-group1, group5-group2, etc). I used these commands: >library(affy) >library(limma) >library("hgu133a") >x<-ReadAffy() >eset<-rma(x) >design <- model.matrix(~ -1+factor(c(1,1,1,2,2,2,3,3,4,4,4,5,5,5,6,6))) >colnames(design) <- c("group1","group2","group3","group4","group5","group6") >fit<-lmFit(eset,design) >contrast.matrix <- makeContrasts(group2-group1,group4-group1,group5-group2, group6-group3,levels=design) >fit2 <- contrasts.fit(fit, contrast.matrix) >fit2 <- eBayes(fit2) The question is: How has to be taken into account that the individuals are the same before and after the treatment? I red about block in lmFit but I am not sure how to do that. Here it would be some correlation, but no so high as in the case of real technical replicates. Thanking you in advance, Susana Kalko [[alternative HTML version deleted]]

Why not follow the section in the limma User's Guide on technical replication? Blocking on individuals and technical replication is really the same thing. Gordon > Date: Mon, 21 Feb 2005 11:41:53 +0100 > From: <skalko@clinic.ub.es> > Subject: [BioC] limma and paired data > To: <bioconductor@stat.math.ethz.ch> > > Dear all, > > > > I have a question on a subject that I think was not discussed here > before. > > I am using limma package for the detection of significant differential > expression in an affy experiment: > > > > 3 "Healthy" (group1) indiv. before treatment and the same indiv. after > treatment (group4) > > 3 "ill-low" (group2) " " " > " " (group 5) > > 2 "ill-high" (group3) " " " > " " (group 6) > > > > the interest is comparing effects of the treatment (i.e. group4-group1, > group5-group2, etc). I used these commands: > > > >>library(affy) > >>library(limma) > >>library("hgu133a") > >>x<-ReadAffy() > > > >>eset<-rma(x) > >>design <- model.matrix(~ -1+factor(c(1,1,1,2,2,2,3,3,4,4,4,5,5,5,6,6))) > >>colnames(design) <- > c("group1","group2","group3","group4","group5","group6") > > > >>fit<-lmFit(eset,design) > > > >>contrast.matrix <- > makeContrasts(group2-group1,group4-group1,group5-group2, > group6-group3,levels=design) > > > >>fit2 <- contrasts.fit(fit, contrast.matrix) > >>fit2 <- eBayes(fit2) > > > > The question is: How has to be taken into account that the individuals > are the same before and after the treatment? > > I red about block in lmFit but I am not sure how to do that. Here it > would be some correlation, but no so high as > > in the case of real technical replicates. > > > > Thanking you in advance, > > Susana Kalko

Dear Gordon, Thank you very much for your answer. If I understood well, your suggestion means to include two commands (shown with **) below? --- >eset<-rma(x) >design <- model.matrix(~ -1+factor(c(1,1,1,2,2,2,3,3,4,4,4,5,5,5,6,6))) >colnames(design) <- c("group1","group2","group3","group4","group5","group6") **>cor<-duplicateCorrelation(eset, design, block=c(1,2,3,4,5,6,7,8,1,2,3,4,5,6,7,8)) **>fit<-lmFit(eset, design, block= c(1,2,3,4,5,6,7,8,1,2,3,4,5,6,7,8), correlation=cor$consensus) >contrast.matrix <- makeContrasts(group2-group1,group4-group1,group5-group2, group6-group3,levels=design) >fit2 <- contrasts.fit(fit, contrast.matrix) >fit2 <- eBayes(fit2) --- A second question is, what would be the consequences if you do not calculate "duplicateCorrelation" and use only "block" argument in "lmFit" (where the default for correlation is 0.75)? That may signify that I am forcing high correlation between arrays? Is that incorrect, or a defensible approximation? By the way, I am working with human arrays. Thank you again for your help and advice Best, Susana -----Mensaje original----- De: Gordon K Smyth [mailto:smyth@wehi.EDU.AU] Enviado el: 23 February 2005 13:58 Para: KALKO, SUSANA (IDIBAPS) CC: bioconductor@stat.math.ethz.ch Asunto: [BioC] limma and paired data Why not follow the section in the limma User's Guide on technical replication? Blocking on individuals and technical replication is really the same thing. Gordon > Date: Mon, 21 Feb 2005 11:41:53 +0100 > From: <skalko@clinic.ub.es> > Subject: [BioC] limma and paired data > To: <bioconductor@stat.math.ethz.ch> > > Dear all, > > > > I have a question on a subject that I think was not discussed here > before. > > I am using limma package for the detection of significant differential > expression in an affy experiment: > > > > 3 "Healthy" (group1) indiv. before treatment and the same indiv. after > treatment (group4) > > 3 "ill-low" (group2) " " " > " " (group 5) > > 2 "ill-high" (group3) " " " > " " (group 6) > > > > the interest is comparing effects of the treatment (i.e. group4-group1, > group5-group2, etc). I used these commands: > > > >>library(affy) > >>library(limma) > >>library("hgu133a") > >>x<-ReadAffy() > > > >>eset<-rma(x) > >>design <- model.matrix(~ -1+factor(c(1,1,1,2,2,2,3,3,4,4,4,5,5,5,6,6))) > >>colnames(design) <- > c("group1","group2","group3","group4","group5","group6") > > > >>fit<-lmFit(eset,design) > > > >>contrast.matrix <- > makeContrasts(group2-group1,group4-group1,group5-group2, > group6-group3,levels=design) > > > >>fit2 <- contrasts.fit(fit, contrast.matrix) > >>fit2 <- eBayes(fit2) > > > > The question is: How has to be taken into account that the individuals > are the same before and after the treatment? > > I red about block in lmFit but I am not sure how to do that. Here it > would be some correlation, but no so high as > > in the case of real technical replicates. > > > > Thanking you in advance, > > Susana Kalko