I have been using DESeq2 to analyze small RNA expression across several treatments. I am surprised to find in my results some sequences that have very high variance listed as significantly differentially expressed. I was under the impression that one of DESeq2's strengths is dealing with variance. Is there an additional command I need to run?
An example of normalized expression values for 4 replicates that come up differentially expressed from another treatment:
I am running the following:
dds.4dpd <- DESeq(dds.4dpd, test="LRT", full=~treatment, reduced=~1) <font face="sans-serif, Arial, Verdana, Trebuchet MS"> </font>res4_1 <- results(dds.4dpd, name="treatment_E_vs_A") res4_1<-subset(res4_1, padj<0.01) res4_1a<-subset(res4_1, log2FoldChange>2) res4_1b<-subset(res4_1, log2FoldChange< -2) res4_1full <-rbind(res4_1b, res4_1a) res4_1_final <-subset(res4_1full, baseMean>25)