Phasing and imputation

Each cohort and array was phased and imputed separately (Fig. 1). Before phasing, variants were excluded that were duplicates, monomorphics, singleton variants, had poor intensity clustering, or had allelic differences between the array and reference panel. Samples not genotyped on both the OmniExpress and Exome chip arrays were excluded, as well as variants with MAF <5% genotyped on the OmniExpress. Samples were phased using SHAPEIT v2.r778 (ref. ³³) and imputed using IMPUTE v2.3.1 (refs ^{33, 34}). Following imputation, any variant with HWE P<1.00 × 10⁻⁴ or imputation information score <0.4 was excluded. There was good genotype concordance between the 249 overlapping samples in the imputed and WGS (Supplementary Note 3).