I have Illumina RNA Seq data from human biopsies (2 groups, 50 samples each). My pipeline is trimmomatic-star-featureCounts-edgeR-voom/limma (big fan of limma, being a veteran from array days..).
I am looking into the optimal ways to filter my data to increase detection power. Thus far, I have used the method from the limma manual to filter by cpm in >50 samples. I have stumbled over discussions of alternative ways to filter, e.g. htsfilter, or filtering by variance. I recall, however, that certain ways to filter do not work well with limma.
What is your recommendation? Is it appropriate to try the alternative ways to filter here, or would you recommend me to stick with cpm-based filtering, or no filtering at all?