Data from two dbGaP samples with individuals of EA and AA ancestry that were ascertained for alcohol and substance dependence were utilized for replication of GWS SNPs. These included the Study of Addiction: Genes and Environment (phs000092.v1.p1, SAGE: non-overlapping individuals numbered EA: 630 cases and 1,020 controls; AA: 387 cases and 415 controls)24 and the Yale-Penn AA sample (phs000425.v1.p1) with 1,525 cases and 485 controls25. Any overlapping participants as well as the first and second degree of relatives (π ≥ 0.2) of COGA members in SAGE or Yale-Penn were excluded from the replication samples. Cases and controls were defined as described above. Covariates included sex and the first 3 principal components. For SAGE, birth cohorts as defined in COGA were included as covariates while for Yale-Penn AA, age was used (as recommended in prior publications of this sample51,52). Effect sizes across COGA and replication samples were meta-analyzed in METAL 47. For SNPs associated with ANYDEP in the EA families, we also examined summary statistics for association with alcohol intake frequency in 452,264 individuals from the UK Biobank (http://geneatlas.roslin.ed.ac.uk/, accessed 11/26/2018)26 and with cannabis use from the current largest GWAS of the phenotype [n=184,76527].
