I'm using edgeR to analyze a RNAseq data set which has many biological replicates and several variables. I have tried to use both the exact test pipeline and the glm function.
- Which of the three approaches (exactTest, glmFit, glmQLFit) is more appropriate for my data set?
- I noticed that by updating R to the 3.3.1 version and edgeR to 3.14 version the results of the glmQLFit are different from the ones obtained with R 3.2.2 and edgeR 3.12 version (see below). Is there a difference in the glmQLFit between the two versions?
My code is as follows:
y <- calcNormFactors(y)design <- model.matrix(~0+Group) colnames(design) <- levels(Group) y <- estimateDisp(y, design) qlfit <- glmQLFit(y, design) my.contrasts <- makeContrasts( D1.DvsD0.D= D1.D-D0.D, D2.DvsD1.D= D2.D-D1.D,levels=design) lrt <- glmQLFTest(qlfit, contrast=my.contrasts[,"D1.DvsD0.D"]) d1<- topTags(lrt, n=Inf) Version 3.12 result: summary (dup<- decideTestsDGE(lrt)) [,1] -1 3566 0 6725 1 2714 Version 3.14 result: summary (dup<- decideTestsDGE(lrt)) [,1] -1 4323 0 5535 1 3147
Any advice is much appreciated!