Diffusion Approximations for Demographic Inference: DaDi

Abstract

Models of demographic history (population sizes, migration rates, and divergence times) inferred from genetic data complement archeology and serve as null models in genome scans for selection. Most current inference methods are computationally limited to considering simple models or non-recombining data. We introduce a method based on a diffusion approximation to the joint frequency spectrum of genetic variation between populations. Our implementation, DaDi, can model up to three interacting populations and scales well to genome-wide data. We have applied DaDi to human data from Africa, Europe, and East Asia, building the most complex statistically well-characterized model of human migration out of Africa to date.