Determining the probability of hemiplasy in the presence of incomplete lineage sorting and introgression

Abstract

AO_SCPLOWBSTRACTC_SCPLOWThe incongruence of character states with phylogenetic relationships is often interpreted as evidence of convergent evolution. However, trait evolution along discordant gene trees can also generate these incongruences - a phenomenon known as hemiplasy. Classic phylogenetic comparative methods do not account for discordance, resulting in incorrect inferences about the number of times a trait has evolved, and therefore about convergence. Biological sources of discordance include incomplete lineage sorting (ILS) and introgression, but only ILS has received theoretical consideration in the context of hemiplasy. Here, we derive expectations for the probabilities of hemiplasy and homoplasy with ILS and introgression acting simultaneously. We find that introgression makes hemiplasy more likely than ILS alone, suggesting that methods that account for discordance only due to ILS will be conservative. We also present a method for making statistical inferences about the relative probabilities of hemiplasy and homoplasy in empirical datasets. Our method is implemented in the software package HeIST (Hemiplasy Inference Simulation Tool), and estimates the most probable number of transitions among character states given a set of relationships with discordance. HeIST can accommodate ILS and introgression simultaneously, and can be applied to phylogenies where the number of taxa makes finding an analytical solution impractical. We apply this tool to two empirical cases of apparent trait convergence in the presence of high levels of discordance, one of which involves introgression between the convergent lineages. In both cases we find that hemiplasy is likely to contribute to the observed trait incongruences.