I am trying to compare two conditions (WT and KO) across multiple tissues in terms of number of significant genes using DESeq2. Some tissues had good distributions of p-values (0-enriched) but some showed skewed to right where I could adjust using fdrtool. In the case, do I need to apply fdrtools to all the tissues (including the tissues with good distribution of p-values obtained directly from DESEq2) and compare the number of significant genes (alpha<0.01) across tissues ? I feel that that is fair to compare the number of significant genes across tissues ...
I'll leave the fdrtool recommendation to Bernd Klaus.
Do you happen to have batches, and if so are you controlling for them? Also, when you look at the p-value histogram, are you removing very low count genes (which do not follow a uniform)?
Note that while you can compare number of significant genes across tissue, you may have differential power across tissues (sample size, distribution of effect sizes, distribution of mean count, etc.) which you should keep in mind.
