To estimate h2SNP in admixed individuals and compare h2SNP across different ancestries, individual-level genotype data was analyzed using an unweighted linear mixed model81 as implemented in the LDAK software82. For each ancestry group (EUA and AFA, respectively), imputed individual-level genotype data were filtered to bi-allelic SNPs with MAF ≥1% in the corresponding 1KGP phase 3 superpopulation. Imputed genotype probabilities ≥0.8 were converted to best-guess genotype calls, and for each ancestry group, studies were merged and SNPs with <95% genotyping rate or MAF <10% removed. Next, to estimate relatedness between subjects, a genetic relatedness matrix (GRM) was constructed based on autosomal SNPs that were LD pruned at r2 > 0.2 over a 1MB window, and an unweighted model with α = −1, where α is the power parameter controlling the relationship between heritability and MAF. To prevent bias of h2SNP due to cryptic relatedness, strict relatedness filters were applied. For pairs with relatedness values > the negative of the smallest observed kinship (−0.014 for EUA and −0.045 for AFA, respectively), one subject was randomly removed. PC’s were then calculated in the
