still potentially cause problems (Ge et al., 2019). Lastly, the PRS we used was from a meta-analysis of MVP-AUD (Kranzler et al., 2019) and UKBB-AUDIT-P (Sanchez-Roige et al., 2019). As noted, these are from somewhat different although correlated phenotypes (Sanchez-Roige et al., 2019) and from different populations, which may reduce the performance of the PRS in our sample even we only kept the variants that have the same directions of effects.
