Hello.

I have read that one shouldn't use FPKM data as input for limma. I did so and only found 5 DE genes between two groups between which a lot more DE genes would be expected. I was wondering whether this anomalous result can be due to use FPKM data as input. Is that possible?

Not ideal, but, if I were to do it, I would transform the FPKMs via log [base 2] FPKM + 0.1, and then perform the differential expression analysis via the limma-trend pipeline (eBayes(..., trend = TRUE)),

I would absolutely not run any sort of differential expression analysis on the FPKMs without any further transformation

Kevin