Background Peer drinking is one of the most robust predictors of college students' alcohol use and can moderate students' genetic risk for alcohol use. Peer effect research generally suffers from 2 problems: selection into peer groups and relying more on perceptions of peer alcohol use than peers' self-report. The goal of the present study was to overcome those limitations by capitalizing on a genetically informed sample of randomly assigned college roommates to examine multiple dimensions of peer influence and the interplay between peer effects and genetic predisposition on alcohol use, in the form of polygenic scores. Methods We used a subsample (n = 755) of participants from a university-wide, longitudinal study at a large, diverse, urban university. Participants reported their own alcohol use during fall and spring and their perceptions of college peers' alcohol use in spring. We matched individuals into their rooms and residence halls to create a composite score of peer-reported alcohol use for each of those levels. We examined multiple dimensions of peer influence and whether peer influence moderated genetic predisposition to predict college students' alcohol use using multilevel models to account for clustering at the room and residence hall level. Results We found that polygenic scores (beta = 0.12), perceptions of peer drinking (beta = 0.37), and roommates' self-reported drinking (beta = 0.10) predicted alcohol use (all ps < 0.001), while average alcohol use across residence hall did not (beta = -0.01, p = 0.86). We found no evidence for interactions between peer influence and genome-wide polygenic scores for alcohol use. Conclusions Our findings underscore the importance of genetic predisposition on individual alcohol use and support the potentially causal nature of the association between peer influence and alcohol use.