PGRS for alcohol dependence were created in GS:SFHS using PLINK as previously described in detail (Purcell et al. 2009) using the summary GWAS data from the two independent (SAGE and Yale‐Penn) GWAS of alcohol dependence (Gelernter et al. 2013). Both the Yale‐Penn and SAGE alcohol dependence GWAS used an ordinal model to test for association. The imputed SNP allele dosage was the dependent variable and DSM‐IV symptom counts for alcohol, cocaine and opioid dependence (adjusted for age, sex and ancestry principal components) were ordinal predictors. For 5 708 204 high‐quality SNPs with P‐values in both the Yale‐Penn and SAGE datasets, inverse variance meta‐analysis was performed using METAL (Willer, Li & Abecasis 2010). These summary data were used to create a meta‐analysis PGRS for alcohol dependence (n = 5127) in GS:SFHS. PGRS were created using SNPs associated with alcohol dependence with P‐value thresholds of 0.01, 0.05, 0.1, 0.5 and 1 in the Yale‐Penn and SAGE GWAS; however, only data from the P‐values for the ≤ 0.5 threshold are presented in the tables as these generally explained the largest amount of variance in the dependent variable (Fig. 1 shows significance and variance explained at all P‐value thresholds).
