2022 Winner: Inferring non-additive multi-locus selection in introgressed populations using hidden Markov models

Project Information
Inferring non-additive multi-locus selection in introgressed populations using hidden Markov models
Corbett-Detig Lab
Admixture is a phenomenon where genetic material from potentially disparate source populations combines and is thought to be a major source of adaptive novelty. As such, multi-locus and non-additive selection on introgressing mutations is potentially common in natural admixed populations. However, existing tools for inferring adaptive introgression only account for additive selection at a single site, overlooking phenomena such as interference among selected loci that are located proximally along a chromosome and dominance between alleles on sister chromosomes. Furthermore, most existing applications and methods assume that the landscape of local ancestry along the genome can be inferred prior to searching for selection, ignoring the fact that the local ancestry landscape is shaped by natural selection. To meet this important need, we present AHMM-MLS, a hidden Markov model based tool for inferring and identifying multiple selected sites on a chromosome. This tool numerically calculates the expected local ancestry landscapes in an admixed population for a given MLS model, and then optimizes the model to fit the data. It uses read pileup data in an introgressed population to identify selected sites and estimate a multi-locus selection model. In applying our method to a suite of simulated admixed populations, we find that the estimated strength of selection can be affected by ignoring the contributions of other sites and that our method can often identify the number of selected sites and their dominance coefficients. In applying our method to real data from admixed populations of Drosophila melanogaster we find that the selection coefficients of some selected sites have been overestimated in the past, and that some selected sites show evidence for dominance. This method will enable more accurate and detailed analyses of selection in admixed populations than has been possible previously.
PDF icon 1493.pdf
  • Nicolas Maggiani Ayala (Crown)