Modelling rare genetic variation to investigate recent human history

By studying human genetic variation, recent studies have provided key insights into human history, including major migration events and population turnovers. However, for many historical questions, the events of interest involve populations that are very closely related with each other, which pushes many traditional methods to the limits of resolution. Here I will present a new method that we developed and recently published, rarecoal, which infers population history and identifies fine-scale genetic ancestry from rare variants. The method efficiently computes the joint site frequency spectrum of rare variants, by probabilistically modeling the coalescent tree of rare derived mutations under a population model with split times, branch population sizes and admixture edges. I will present two applications of rarecoal and related methods. In the first study, we analyse the genomes from 10 ancient English samples, excavated close to Cambridge and ranging from the late Iron Age to the middle Anglo-Saxon period. By analysing shared rare variants with hundreds of modern samples from Britain and Europe, we estimate that on average the contemporary East English population derives 38% of its ancestry from Anglo-Saxon migrations. Using rarecoal we find that the Anglo-Saxon samples are closely related to modern Dutch and Danish populations, while the Iron Age samples share ancestors with multiple Northern European populations. In the second study, we analyse hundreds of modern samples from Siberia and America and gain insights into the prehistory of Native Americans. In particular, we detect a gene flow event from Siberia into Athabaskan speaking Native Americans around 7,000 years ago, that is separate from first American ancestry and later immigrations of Eskimo-Aleuts. We also show that this Athabaskan-specific Siberian ancestry is closely related to the ancient Saqqaq individual, which confirms a previous hypothesis of Palaeo-Eskimo ancestry in Athabaskan speakers.