A linear association model was applied for association analyses between SNPs genome-wide for the log-transformed MaxDrinks. In this analysis, MaxDrinks was the dependent variable and each SNP was a predictor. Each model was adjusted for age, sex and the first three ancestry PCs. Consider body weight may influence alcohol distribution and consumption, we conducted a separate analysis with body mass index (BMI) as an additional covariate. The p-value for genome-wide significance was set at 5.0 x 10-8. To further limit the confounding factors, we tested models for moderately associated SNPs (p < 1 × 10−4) for dependence on cocaine, opioids, and alcohol, three major comorbid substance use in our samples
