We meta-analyzed GWAS summary statistics from 109 ancestrally diverse cohort datasets with 688,808 MD cases and 4,364,225 controls (see STAR Methods, Table S1, Methods S1, and see key resources table). These studies had power equivalent to a case-control study of 1,004,459 cases and 1,004,459 controls, with 23% in diverse/non-European ancestries (Table 1). For cohorts with diverse ancestries, associations were assessed using tools that explicitly model population structure, admixture, and relatedness (GENESIS). For a subset of cohorts with ancestrally diverse samples, we compared the sample size using the commonly used strategy of assigning individuals into ancestry groups followed by logistic regression (N = 24,859) to our joint approach (N = 47,642) and found a 92% sample size increase. Our final sample size of 163,611 cases and 1,001,890 controls with diverse ancestries (Methods S1) led to an increase in the discovery of genome-wide significant loci compared with the European-only ancestry studies analysis. Using conditional-and-joint GCTA-COJO12 analysis with threshold p ≤ 5 × 10−8 within 10 Mb windows for the combined meta-analysis, we identified 697 significant independent single-nucleotide polymorphisms (SNPs) in 635 genomic
