Hi, I am using edgeR to analyze a knockout cell line vs WT control and expect a major fraction of transcripts differentially expressed. Using default settings with edgeR I could confirm significant downregulation of my knocked out transcript. However, given the following statement in the edgeR manual I am worried I am using incorrect assumptions for the analysis:

"TMM is recommended for most RNA-Seq data where the majority (more than half) of the genes are believed not differentially expressed between any pair of the samples. The following commands perform the TMM normalization and display the normalization factors."

What would be an acceptable strategy for data where most genes are believed to be differentially expressed?

The problem only occurs when all the DE genes are changed in the same direction. There is no automatic RNA-seq analysis procedure that can accurately handle cases where all or most genes are DE in the same direction. This is an intrinsic limitation of the RNA-seq technology.

In most cases, the best procedure is just to do the usual analysis and see what you get. It would be very unusual for (say) 75% of all genes in the genome to be truly DE in the same direction but, even in this case, the edgeR results will not be disastrous. In the worst case that 75% of genes are truly up-regulated (say) and none are truly down-regulated, then edgeR will slightly underestimate the amount of up-regulation and will introduce a few extra down-regulated genes that might be spurious.