Full Bayesian Comparative Phylogeography from Genomic Data

Abstract

A challenge to understanding biological diversification is accounting for community-scale processes that cause multiple, co-distributed lineages to co-speciate. Such processes predict nonindependent, temporally clustered divergences across taxa. Approximate-likelihood Bayesian computation (ABC) approaches to inferring such patterns from comparative genetic data are very sensitive to prior assumptions and often biased toward estimating shared divergences. We introduce a full-likelihood Bayesian approach, ecoevolity, which takes full advantage of information in genomic data. By analytically integrating over gene trees, we are able to directly calculate the likelihood of the population history from genomic data, and efficiently sample the model-averaged posterior via Markov chain Monte Carlo algorithms. Using simulations, we find that the new method is much more accurate and precise at estimating the number and timing of divergence events across pairs of populations than existing approximatelikelihood methods. Our full Bayesian approach also requires several orders of magnitude less computational time than existing ABC approaches. We find that despite assuming unlinked characters (e.g., unlinked single-nucleotide polymorphisms), the new method performs better if this assumption is violated in order to retain the constant characters of whole linked loci. In fact, retaining constant characters allows the new method to robustly estimate the correct number of divergence events with high posterior probability in the face of character-acquisition biases, which commonly plague loci assembled from reduced-representation genomic libraries. We apply our method to genomic data from four pairs of insular populations of Gekko lizards from the Philippines that are not expected to have co-diverged. Despite all four pairs diverging very recently, our method strongly supports that they diverged independently, and these results are robust to very disparate prior assumptions.

The comparative biogeography of Philippine geckos challenges predictions from a paradigm of climate‐driven vicariant diversification across an island archipelago

“…Our goal is to treat the number of divergence events shared (or not) among the pairs of populations, and the assignment of the pairs to those events, as random variables to be estimated from the aligned sequence data. For eight pairs, there are 4140 possible divergence models (i.e., there are 4140 ways to partition the eight pairs to divergence events; Bell ; Oaks , ). Although divergences caused by sea‐level rise would not happen simultaneously, we expect that on a timescale of the lizards' mutation rate, treating them as simultaneous should be a better explanation of data generated by such a process than treating them as independent.…”

“…Given the large number of models, and our goal of making probability statements about them, we used a Bayesian model‐averaging approach. Specifically, we used the full‐likelihood Bayesian comparative biogeography method implemented in the software package ecoevolity version 0.1.0 (commit b9f34c8) (Oaks ). This method models each pair of populations as a two‐tipped “species” tree, with an unknown, constant population size along each of the three branches, and an unknown time of divergence, after which there is no migration.…”

“…This method can directly estimate the likelihood of values of these unknown parameters from orthologous biallelic characters by analytically integrating over all possible gene trees and mutational histories (Bryant et al. ; Oaks ). Within this full‐likelihood framework, this method uses a Dirichlet process prior on the assignment of our pairs to an unknown number of divergence times.…”

“…Data that satisfy this assumption include single‐nucleotide polymorphisms (SNPs) that are well spaced across the genome. However, by analyzing simulated data, Oaks () showed the method performs better when all linked sites are used than when data are excluded to avoid violating the assumption of unlinked sites. We simulate datasets based on our gekkonid data to confirm these results hold for our sampling design (see below).…”

“…We simulate datasets based on our gekkonid data to confirm these results hold for our sampling design (see below). Furthermore, when analyzing the RADseq data from three of the Gekko population pairs we are analyzing here, Oaks () found biologically unrealistic estimates of divergence times and population sizes when only unlinked variable sites (i.e., SNPs) were analyzed. Using additional simulations, Oaks () found these unrealistic estimates were likely due to data‐acquisition biases, which are known to be common in alignments from reduced‐representation genomic libraries (Harvey et al.…”