Affiliations: 1) Department of Ecology and Evolution, University of Chicago, Chicago, IL; 2) Department of Human Genetics, University of Chicago, Chicago, IL
Keywords: Theory & Method Development
With the rapid accumulation of ancient DNA (aDNA) genomes and evolve-and-resequence (E&R) data, more time-stratified population genomic datasets are emerging. Such time-series data allow us to examine the temporal dynamics of natural selection and can lend power to detecting its footprints. Few selection-detecting methods, however, are tailored to jointly consider multiple samples of the same population at different time, especially when more than one locus is involved. Meanwhile, increasing evidence is supporting the polygenic nature of most traits under selection, underscoring the need for approaches that account for multiple loci. Here, we constructed a hidden-Markov model (HMM) framework based on the Wright-Fisher diffusion model explicitly for directional or stabilizing selection on polygenic traits. To reduce the computational load, we use a normal approximation for the step-wise transition between generations in the underlying diffusion model. We implemented this framework as a Python package with its command-line-interacting script, and benchmarked its performance with both forward and backward simulations. Further, for each major complex trait in the UKBioBank, we extract variation records of their significant loci in historical British populations during the past ~3500 years from the Allen aDNA Resources dataset, and assign these loci their respective effect size estimates in UKBioBank. With the composite likelihood incorporating all significant loci, we were able to obtain estimates of the predominant mode of selection, i.e., directional or stabilizing, and its intensity, for each trait, gaining insights into recent human evolution.