and four ancestry PCs were included as covariates. The second dataset was from the Collaborative Study on the Genetics of Alcoholism (COGA). COGA is a multi-center study of families with alcohol dependence (AD). African American (AA) and European American (EA) subsamples of COGA were included in analysis. For each COGA subsample, all individuals (N=1,527 in AA; N=4,717 in EA) and COGA prospective samples (N=326 in AA: N=822 in EA), which were offspring of COGA families that were born after 1982 (Bucholz et al., 2017) to match the ALSPAC/S4S samples were tested separately. PRSice-2 (Euesden et al., 2015) was used to calculate PRS. Sex, the first four ancestry PCs, and genotyping array indicators were included as covariates. For analyses of all individuals, birth cohorts were also included as covariates. Linear mixed models were fit to adjust family clustering using SAS9.4 (SAS Institute INC., Cary, NC, USA). For both the UCSF and COGA samples, we tested for associations at 8 p-value thresholds: 0.001, 0.01, 0.05, 0.1, 0.2, 0.3, 0.4, and 0.5.
