One source of statistical power in PRS analyses lies in the sample size of the discovery genome-wide association study (GWAS). With the publication of multiple large-scale GWAS of AUD-related phenotypes (Kranzler et al., 2019, Sanchez-Roige et al., 2019, Zhou et al., 2020, Walters et al., 2018), it is now possible to perform PRS analysis of AUD with discovery GWAS of sufficient statistical power. However, current PRS for AUD explained a small proportion of the variance in AUD related traits, and these estimates are often lower than the variance attributable to family history (Kendler et al., 2012, Kiiskinen et al., 2020, Liu et al., 2019, Walters et al., 2018, Wray et al., 2014, Zhou et al., 2020). For example, PRS derived from the largest AUD related GWAS to date only explained ≤2.12% variations (Zhou et al., 2020). In this study, we propose a new strategy to calculate PRS using variants that had the same directions of effects in discovery GWAS. Since study-specific variants and variants having small P-values due to random variations were excluded, we hypothesized that the performance of PRS
