The independent COGA sample was used to attempt to replicate these top signals in the discovery GWAS. Two of these SNPs, rs115455482 and rs74722579, were available in the COGA data; neither of these were significantly associated with trait, although their β-effects were both in the same direction (0.05 and 0.02, respectively). Further meta-analysis of both Yale-Penn discovery EA samples with the COGA EA replication sample (Table 1) for these SNPs increased the statistical significance of the association signals (meta-P-values of 1.32 × 10−10 and 2.60 × 10−9 for rs115455482 and rs74722579, respectively). In addition, meta-analysis for rs74722579 (MAF ~ 0.10) including all available AA samples from Yale-Penn 1 and 2, and COGA (total cases = 362 and controls = 3404) yielded a nominally significant meta-P-value (meta-P = 4.58 × 10−2); the analysis for rs115455482 was not significant. For the latter variant, the minor allele is rare in AAs (MAF = 0.01, meta-P = 1.37 × 10−1) (Table 1). Taken together, the robust association signals for the two top SNPs in EAs and the nominal significance for rs74722579 in AAs (Fig. 1 and Table 1) support the validity of the GWAS results.
