Distinguishing multiple-merger from Kingman coalescence using the two-site frequency spectrum.
Authors: Eliot Fenton 1; Daniel Rice 2; John Novembre 3; Michael Desai 1
Affiliations: 1) Harvard University, Cambridge, MA; 2) National Center for Biotechnology Information, Bethesda, MD; 3) University of Chicago, Chicago, IL
Keywords: Theory & Method Development
Many demographic inference methods in population biology use the site frequency spectrum (SFS) to fit past population sizes to present-day genetic data. These methods typically assume the population ancestry is well represented by a Kingman coalescent in which only two lineages can coalesce at once and all lineages are equally likely to coalesce at any time. However, real populations often violate the assumptions of the Kingman coalescent. For example, natural selection and highly skewed offspring number distributions can each lead to “multiple-merger” coalescence events in which three or more lineages simultaneously coalesce. In both cases, features of the SFS can often be reproduced by a Kingman coalescent with changing population size, making it difficult to infer the evolutionary and demographic forces shaping a population from the SFS alone. Here, I present results on distinguishing changing-population-size Kingman coalescents from multiple-merger coalescents using the two-site frequency spectrum (2SFS) constructed from pairs of linked sites. Using a combination of forward- and backward-time simulations, we show that our method distinguishes fluctuating population sizes from natural selection and skewed offspring number distributions. I additionally present a pipeline for analyzing real-world genomic data and results from applying this technique to Drosophila melanogaster strains.