Entering edit mode
9.9 years ago
Zhenyu Zhang
★
1.1k
Say, I have two tissues, from each one I called eQTLs, with effect sizes and p-values. My question is, how to find eQTLs only exist in tissue A, but not B. Obviously, a simple comparison is not appropriate, because "not detected" as an eQTLs in one tissue does not mean it is not (it could simply be lack of power).
I would think it as some sort of meta analysis and thus have checked some papers, but couldn't find what I want. Could anyone suggest a test, or a paper doing that, or a package? My feeling is that this question fits better in the bayesian framework? Thanks.
Wouldn't you just add "tissue" to your statistical model and possibly look for interactions?
I could do that. But the interaction could only tell me whether the effects are the same for two tissues, not whether one has effect while the other has not; in addition, I could also compare the interaction to zero, but it backs to my old point that if there are no detectable effect, it does not mean no effect and could just be no power.