An approximate likelihood method reveals ancient gene flow between human, chimpanzee and gorilla

Gene flow and incomplete lineage sorting are two distinct sources of phylogenetic conflict, i.e., gene trees that differ in topology from each other and from the species tree. Distinguishing between the two processes is a key objective of current evolutionary genomics. This is most often pursued via the so-called ABBA-BABA type of method, which relies on a prediction of symmetry of gene tree discordance made by the incomplete lineage sorting hypothesis. Gene flow, however, need not be asymmetric, and when it is not, ABBA-BABA approaches do not properly measure the prevalence of gene flow. I introduce Aphid, an approximate maximum-likelihood method aimed at quantifying the sources of phylogenetic conflict via topology and branch length analysis of three-species gene trees. Aphid draws information from the fact that gene trees affected by gene flow tend to have shorter branches, and gene trees affected by incomplete lineage sorting longer branches, than the average gene tree. Accounting for the among-loci variance in mutation rate and gene flow time, Aphid returns estimates of the speciation times and ancestral effective population size, and a posterior assessment of the contribution of gene flow and incomplete lineage sorting to the conflict. Simulations suggest that Aphid is reasonably robust to a wide range of conditions. Analysis of coding and non-coding data in primates illustrates the potential of the approach and reveals that a substantial fraction of the human/chimpanzee/gorilla phylogenetic conflict is due to ancient gene flow. Aphid also predicts older speciation times and a smaller estimated effective population size in this group, compared to existing analyses assuming no gene flow.