Inferring species divergence times using pairwise sequential Markovian coalescent modelling and low-coverage genomic data

Abstract

Understanding when species diverged aids in identifying the drivers of speciation, but the end of gene flow between populations can be difficult to ascertain from genetic data. We explore the use of pairwise sequential Markovian coalescent (PSMC) modelling to infer the timing of divergence between species and populations. PSMC plots generated using artificial hybrid genomes show rapid increases in effective population size at the time when the two parent lineages diverge, and this approach has been used previously to infer divergence between human lineages. We show that, even without high coverage or phased input data, PSMC can detect the end of significant gene flow between populations by comparing the PSMC output from artificial hybrids to the output of simulations with known demographic histories. We then apply PSMC to detect divergence times among lineages within two real datasets: great apes and bears within the genus Ursus Our results confirm most previously proposed divergence times for these lineages, and suggest that gene flow between recently diverged lineages may have been common among bears and great apes, including up to one million years of continued gene flow between chimpanzees and bonobos after the formation of the Congo River.This article is part of the themed issue 'Dating species divergences using rocks and clocks'.

Keywords: PSMC; bears; great apes; species divergence.

Figure 1.

Results of simulation experiments designed to test the accuracy of hPSMC in inferring divergence time under three varying demographic scenarios: (a) the influence of using phased (dashed lines) versus unphased (solid lines) data to infer divergence times at seven different depths of divergence; (b) the influence of pre-divergence effective population size on the ability of hPSMC to detect divergence between unphased data; (c) the influence of post-divergence migration between populations. In (b,c), divergence between populations occurs 1 Ma, and the dashed horizontal lines indicate the pre-divergence effective population size.

Figure 2.

An approach to pinpoint the transition (divergence) time using simulation. Here, the hPSMC plot generated for the artificially created chimpanzee/bonobo hybrid genome (blue line) is compared with 11 simulated datasets with divergence times ranging from 0 to 500 ka. Divergence is inferred to have occurred between the simulated divergence times of 300–400 ka (red shaded region), as these are the closest simulations with transition times that do not intersect the transition time of real data. All simulations assume a pre-divergence effective population size of 18 000, which was estimated from the plot of the real data. The horizontal lines delineate the range of ancestral effective population size estimates that correspond to 1.5–10 times the pre-divergence N_e (27 000–180 000). Plots resulting from all other comparisons are provided as electronic supplementary material, figures S1–S17.

Figure 3.

Results of hPSMC analyses of (a) five species of great apes (human, chimpanzee, bonobo, gorilla, orangutan) and (b) three species of bears in the genus Ursus (American black, brown, polar). Within the great apes (a), we observe the expected pattern of divergence in which orangutans diverge most anciently followed by gorillas and then humans, and chimpanzees and bonobos diverge most recently. Within bears (b), we also find the expected order of divergence, where the American black bear is the most ancient divergence, followed by brown bear/polar bear divergence (light brown) and brown bear/brown bear divergence (dark brown) The polar bear/polar bear divergence (blue) is inferred to have occurred very recently and may be an artefact of the small effective population size of polar bears (figure 1b).