with AD symptom counts (meta-analyzed p-value from 3.40*10−9 to 1.10*10−5). In EAs, we identified 106 variants with significant z-scores. All of these are located in the regions surrounding rs10031423 (Supplementary Table 10). According to RegulomeDB, none of them showed notable evidence of functional regulatory effects, and the strongest increase of significance was observed for rs1229984 (z = 13.369; meta-analyzed p-value from 0.726 to 0.189). In AAs, the analysis of ADH1C rs6846835 (significant in AAs) revealed 15 variants with significant z-scores. All of these are located on chromosome 4, as rs6846835 (Supplementary Table 11). Among them, only ten variants present an r2 with rs6846835<0.2. Considering the non LD variants, rs12639887 showed significant effects on both ADH1B rs1693457 and ADH1C rs6846835. In the analysis of ADH1C rs6846835, rs12639887 showed a strong effect on the association with AD symptom count (z = −6.015; meta-analyzed p-value from 5.65*10−9 to 0.237). Stratifying the AA and EA samples for rs12639887 genotype, we observed the same trend in both ancestry groups for the association between ADH1C rs6846835 and AD symptom count (Supplementary Table 12). Regarding the analysis of ADH1C rs6846835 in EAs, we observed 72 variants with significant z-scores. They are all located on chromosome 4 (Supplementary
