I have done several DE analyses at gene and transcript levels using edgeR and EBSeq and I found ~30% of transcriptome (~2100 genes out of ~7000, and ~3300 transcripts out of ~11000) is depleted from my experimental samples. The total read counts for these genes is about 5 million reads (out of ~17 million total reads in each control sample). This large depletion causes an overestimation of other genes (e.g. housekeeping genes) in the DE analysis.
Three questions:
1) I can correct for this overestimation by modifying the normalization factors of these samples. I do not know how to do this correction in EBSeq. Any suggestions would be appreciated.
2) As far as I know, normally in DE analysis we assume that majority of the genes are unchanged but this is not the case here. Any thoughts about whether this would invalidate the whole analysis?
3) Is there another package that I can use in R for DE analysis other than limma and DESeq (specially for analysis at transcript level) that provides me with a way to correct for this overestimation?
Thanks!
