ancestry groups (EA and AA), separate analyses were run on subjects genotyped on the HumanOmni1-Quadv1.0 and Human Core Exome microarrays. SNPs were filtered out if the minor allele frequency (MAF) was <5%, or if the imputation INFO score was <0.7. Meta-analyses were then performed within ancestry groups using METAL, which was also used to remove SNPs with heterogeneous effect estimates across the two microarrays, and to adjust summary statistics based on the genomic inflation factor (λ).39 LocusZoom was used for regional association plot generation.40 We defined the cutoff for genome-wide significance using the criterion of P = 5.0 × 10−8.
