the testing step, (1) two independent samples were used to replicate each other, which significantly reduced the chance of false positive findings (i.e., false discovery rate). (2) We aimed to detect replicable regions, not individual markers. Thus, more than one risk marker were detected in the risk regions, which reduced the chance of false positive associations too. (3) Functional analysis in distinct tissues as confirmation of association finding further reduced the chance of false positive findings (including co-localization of association signals and eQTL signals randomly), although using different independent samples in one study might increase the false negative rates due to sample heterogeneity. (4) -log(P) value distributions across EAs, AAs and meta-analysis were compared using Pearson correlation analysis. The consistency between them would significantly reduce the chance of false positive findings. Therefore, in the testing step, when an association was replicable across EAs, AAs and meta-analysis, α could be set at 0.05 (except for the exon-level cis-eQTL findings that needed to be corrected for the number of exons and the types of tissues). Accordingly, the power of the discovery (α=5×10-7) and replication (α=0.05) samples to detect the significant genetic effects was analyzed using the power analysis package in R.
