Model-based Assessments of Differential Introgression and Linked Natural Selection During Divergence and Speciation

Demographic changes such as fluctuating population size and differential introgression can mask the effects of natural selection, and affect rates of genome evolution, local adaptation, reproductive isolation, and eventual speciation. Besides identifying differentially introgressing genes (and genomic regions) that are \"labeled\" to be retroactively causal to adaptive evolution and speciation, there is significant impetus to understand, and perhaps estimate the underlying demography that affects current genomic diversity. Using model-based likelihood methods to directly estimate, and decouple the effects of differential intogression and demography across genomic loci offers an ideal solution to detect differential introgression, and population demography and build hypotheses around its underlying evolutionary processes. We describe a computationally efficient parallelized implementation of mixture-model based isolation with migration (IM) analyses to assign loci to classes based on their shared coalescent histories (population sizes, or migration rates). We apply this method to several genomic data sets (great apes - chimpanzees and bonobos, Anopheles mosquitoes, threespine sticklebacks, Mullerian mimics of Heliconius butterflies, mice, European rabbits, and fruitflies), that have been previously characterized (perhaps erroneously) using genome-wide scans of differentiation. We show that we cannot reject a model of differential introgression, or linked selection across a majority of species analyzed, with two species showing the combined effects of differential introgression and linked natural selection across multiple, non-independent genomic loci.