We performed additional QC checks centrally. For each study, we checked the sample size and the total number of SNVs (monomorphic and polymorphic) and assessed the beta distribution. Within each cohort’s results, all monomorphic SNVs were checked to have non-available results. In order to detect potential strand-flip issues, the cohort-coded effect allele frequencies (EAF) of each SNV were compared with the meta-analysed EAF of a group of CHARGE cohorts (AGES, ARIC, CHS, FHS and WHI). Any discordant SNVs showing cohort-EAF ∼ 0 in at least one study, but meta-EAF ∼ 1, or vice versa, were excluded from the central meta-analysis. A set of approximately 11 000 SNVs that were known to have QC issues from central CHARGE QC were also excluded from the meta-analysis. Quantile–Quantile plots were produced to inspect each cohort. After all QC steps were completed 235 677 SNVs remained. The results from all cohorts were then combined into a discovery meta-analysis using the SeqMeta R package.
