Several limitations of this study need to be considered in interpreting the findings. First, we only focused on participants of European ancestry in COGA because the discovery samples in the GWAS study were primarily of European descent and mismatch in ancestry between discovery and target samples can result in bias in GPS prediction (Martin et al., 2017). Thus, it is not clear whether our findings are generalizable to non-European populations. Second, social support and alcohol use were both measured at T2 and these measures lack time precedence. Thus, inferences about causal relations need to be made cautiously and it is possible that the association between social support and alcohol use is bidirectional. Third, we examined sensation seeking as an intermediate phenotype that links alc-GPS to alcohol use. However, there are alternative models that could also explain the association between alc-GPS, sensation seeking, and alcohol use, whereby sensation seeking and alcohol use are indicators of a common underlying factor (e.g., externalizing, Krueger et al., 2002) that is influenced by genetic factors (e.g., alc-GPS). Future research is warranted to test these alternative
