from dbGaP. Since COGA had more phenotypic information, if a sample in the COGA dataset was also in SAGE and/or YalePenn, it was only analyzed as part of the COGA data. SAGE and YalePenn were mixes of related and unrelated individuals, although most were unrelated. Only AA samples from COGA, SAGE, and YalePenn were used. Across all three datasets, AUD was defined as meeting lifetime criteria for DSM-IV alcohol dependence [32] or DSM-5 alcohol use disorder [33]. All other individuals were considered as controls.
