Very interesting that this meta analysis revealed a strong linkage. However, I have several doubts about the significance P used as the threshold in the meta GWAS. First, I didn't see a discussion on the effect sizes of the diseases scanned in the original GWASs; for valid meta analysis, the effect sizes should be as nearly equal as possible. Second--this is my thought; please someone correct me if I am wrong--if one scans "n" SNPs for "m" diseases, the threshold for the P-value to be accepted is uncorrected P divided by (n x m), not just "n", because there are n x m hypotheses being tested (if all n and m are independent). I like how the authors linearly weigh their significance P, which to some extent covers the problem of effect sizes, but the threshold P should really be corrected by the factor 1/(n x m). Am I wrong?
