I'm running DESeq to compare two genotypes responding to a treatment. I turn that was not able to see interactions, at least after apply Benj-Hoch for multiple testing correction. I'm thinking in remove genes that are too lowly expressed with the hope that will increase my power. Now my question is, where is the logical threshold..
I like to do something like
dds <- dds[ rowSums(counts(dds)) > x, ]
which will be a good number for "x", if any?
This will reduce the number of tests made and therefore improve the multiple testing correction of Benjamin-Hochberg. I hope!