only variant selected in ADH1B in calculating PRSgene, resulting in no contribution of ADH1B when calculating PRSgene in EA individuals from the Indiana Biobank. However, for PRSall, multiple common EA variants in ADH1B (e.g., rs2066701, rs1042026, and rs2075633) were included, thus increasing the performance of PRSall. Third, we limited inclusion to variants within gene boundaries. Although PRSintergenic and PRSgene with extended boundaries analyses showed that overall including intergenic concordant variants did not increase the PRS performance, however, some AUD variants are not located within gene boundaries and this may affect AA and EA disproportionately. For example, rs1229978, which is located between ADH1B and ADH1C, is much more common in EA (MAF = 0.39) than in AA (MAF = 0.15) (https://www.ncbi.nlm.nih.gov/snp/rs1229978?vertical_tab=true); therefore, not including this variant in PRS calculations had a larger impact in EA than in AA. Nevertheless, the significance of PRSgene in both AA and EA suggested that most of these genes were AUD-related in these two populations.
