Recent population history inferred from more than 5,000 high-coverage South Asian genomes:

Next, we developed a novel method for estimating the genome-wide average divergence time between a single individual and a focal group. This method focuses on extremely rare variants, which should be the most informative about very recent demographic events, and is robust to demographic events affecting the particular individual studied. We focused this work on samples from Birbhum district, West Bengal due to the presence of additional metadata on caste and religion. We used 704 general-caste individuals from Birbhum as the focal group, and estimated divergence times for all other individuals. Mean divergence times ranged from ~2,600 years for the Santal, an Austro-Asiatic language speaking tribal group, to 850 years for “scheduled castes” (i.e., Dalits), 625 years for Bangladeshis and 225 years for “Other Backward Castes” (OBC) individuals. The recent divergence times for OBC individuals confirms that this category is more of a political construct than a long-lived social grouping, while the other divergence times suggest a substantial amount of gene flow between groups. Finally, we extended our approach to thousands of other genomes from around the world. We show how patterns of rare variation can be used to detect asymmetrical migration, and document evidence for more migration from East Asia into Bengal than the converse.