The proposed solutions are either generate synthetic null data to measure FDR (assumption laden), or use permutation (transcriptomics usually to low n... I think?). So I don't know what the answer is