To verify the effect of common variants with high ΔF on genome-wide significant associations, we analyzed two independent datasets used in a recent published AD GWAS [4]. Specifically, the datasets used in the present analysis comprise our Yale–Penn samples and SAGE samples, the latter obtained via dbGAP application. The Yale–Penn dataset includes 3318 AAs and 2379 EAs. The SAGE dataset includes 1195 AAs and 2528 EAs. Information about the genotyping, quality control and imputation analysis was published previously [4]. Considering the outcomes obtained by our previous AD GWAS, ten variants were found to be genome-wide significant (p < 5*10−8) for AD symptom count in an ancestry group with nonsignificant results for the other one (Supplementary Table 2). Due to the strong linkage disequilibrium (LD) among some significant associations in this AD GWAS, we selected three independent variants from the total of ten: PDLIM5 rs10031423, ADH1B rs1693457 and ADH1C rs6846835. Specifically, PDLIM5 rs10031423 showed R2 = 0 with respect to ADH1B rs1693457 and ADH1C rs6846835 in AAs and EAs, whereas ADH1B rs1693457 and ADH1C rs6846835 showed R2 = 0 in EAs
