in the discovery cohort (6.2×10−9 ≤ p ≤ 0.049) were replicated in the Australian replication cohort (0.001 ≤ p ≤ 0.049) (Table 2 and Figure 2D), with the same directions of gene effects in both cohorts. Meta-analysis showed that all of these 34 replicable SNPs were associated with disease (9.6×10−10 ≤ p ≤ 0.021; Table 2), including four genome-wide significant SNPs, i.e., rs7445832 (p=9.6×10−10), rs13361996 (p=8.2×10−9), rs62380518 (p=2.3×10−8) and rs7714850 (p=3.4×10−8). In the African-American replication cohort, 77 SNPs were nominally associated with alcohol and nicotine co-dependence (0.002 ≤ p ≤ 0.049; data not shown). Eleven risk SNPs in the discovery cohort (8.1×10−6 ≤ p ≤ 0.042) were also risk SNPs in the African-American replication cohort (0.032 ≤ p ≤ 0.049) (Table 3). However, all of these 11 SNPs but one had opposite directions of gene effects between the discovery cohort and the African-American cohort. Meta-analysis showed that only this exceptional one SNP was associated with disease (rs10042968: OR=1.41, p=8.1×10−6 in European-Americans; OR=1.54, p=0.041 in African-Americans; OR=1.42, p=8.0×10−7 in meta-analysis; Table 3). Among these SNPs, rs690957 was a risk SNP across three cohorts (p=0.008, 0.004 and 0.047 in European-Americans, European-Australians and African-Americans, respectively). Rs690957 was also the most significant one in
