Whole genome data in the discovery sample and the region-wide imputed genotype data in other 20 cohorts were analyzed. Before association analysis was conducted, we stringently cleaned the phenotype data and then the genotype data within each ethnicity. Detailed procedures of data cleaning were described previously (Zuo et al., 2012). After cleaning, our subjects had high levels of ancestral homogeneity within each phenotype group (QQ plots were presented for discovery and replication samples previously (Zuo et al., 2011, 2012); λ=1.07 and 1.01 in EA discovery sample and Australian replication sample, respectively). This selection process yielded 805,814 SNPs in the discovery sample (1,409 cases and 1,518 controls) and 300,839 SNPs in 6,438 Australian replication samples. Cleaned SNP numbers in other datasets are shown in Table S24.
