In this study, we found that gene-based PRS (PRSgene) calculated using 858 variants from 410 genes were significantly associated with AUD in both AA and EA, and outperformed the PRS calculated using all variants (PRSall) in AA. Compared to the bottom decile, those at the top PRSgene decile were nearly twice as likely to be AUD cases (OR = 1.76) in AA. The 410 genes included in calculating PRSgene were enriched in 54 GOBPs, and many of them are likely to be AUD-related. They were also enriched in brain tissues. In addition, four genes were targets of drugs in Phase II or III clinical trials to treat AUD; 22 genes were targets of drugs approved by the FDA or in clinical trials to treat other diseases but may be repurposed to treat AUD. Together, these findings showed that biologically meaningful polygenic scores can be characterized in non-European ancestry individuals by leveraging methods that focus on intragenic signals with concordant directions of effects across ancestries. Furthermore, the process identified drugs already under development that could be evaluated for their potential to treat AUD.
