We used RFMix (v.2.03-r0)68, a discriminative modeling approach for rapid and robust local-ancestry inferences, to infer local ancestry in our admixed samples using the European and African ancestry samples from the 1000 Genomes Project69 as reference. We extracted the posterior probability of African ancestry at each SNP per haplotype from the forward–backward output of RFMix. Local ancestry for a genomic region was then estimated as the average African ancestry across all SNPs within the region. As RFMix also computed and output a global ancestry estimate for each sample, we compared global ancestry estimates between STRUCTURE and RFMix and observed a high correlation between estimates from the two programs (Spearman rho = 0.99).
