Ghosts of a structured past: Impacts of ancestral isolation by distance on divergence time estimation

Abstract

Isolation by distance is a widespread pattern in nature that describes the reduction of genetic correlation between subpopulations with increased geographic distance. In the population ancestral to modern sister species, this pattern may hypothetically inflate population divergence time estimation due to the potential for allele frequency differences in subpopulations at the ends of the ancestral population. In this study, we analyze the relationship between the time to the most recent common ancestor and the population divergence time when the ancestral population model is a linear stepping-stone. Using coalescent simulations, we compare the coalescent time to the population divergence time for various ratios of the divergence time over the product of the population size and the deme number. Next, we simulate whole genomes to obtain SNPs, and use the Bayesian coalescent program SNAPP to estimate divergence times. We find that as the rate of migration between neighboring demes decreases, the coalescent time becomes significantly greater than the population divergence time when sampled from end demes. Divergence-time overestimation in SNAPP becomes severe when the divergence-to-population size ratio < 10 and migration is low. We conclude that studies estimating divergence times be cognizant of the potential ancestral population structure in an explicitly spatial context or risk dramatically overestimating the timing of population splits.