Inferring Natural Selection from Genomic Data Local adaptation in indigenous Siberians
Adaptation via positive natural selection is one of the key evolutionary processes shaping genetic and phenotypic variation. The basic idea is familiar: a mutation in the genome causes a protein or regulatory change, that change results in altered biological processes, and those altered processes result in increased reproductive success for those carrying the mutation. Over time, advantageous variants rise in frequency, leading to characteristic patterns of genomic diversity around selected sites. But while classical population genetics provides a solid theoretical foundation to understand these evolutionary processes, how do we actually find those genes that have undergone positive natural selection in real populations?
Recent advances in sequencing technologies have resulted in the ability to sequence much of the genome and to query all genes at low cost. In response, a number of selection scan statistics have been developed, where one computes a statistic in sliding windows across the genome, and then uses the loci in the tails of the empirical distribution of the scan statistic as candidates. In humans, dozens of papers have published hundreds of candidate genes using these methods. To date, however, only several have additional evidence strongly supporting natural selection.
In this talk I will provide an overview of the population genetic theory underlying popular statistical methods to infer demography and natural selection from genomic data. I will then illustrate the shortcomings of these methods, highlighting both the high rate of false positives, as well as how adaptive variants may not be discovered by these approaches. Throughout the talk I will use examples from my dissertation work on cold adaptation in indigenous Siberians, and present a more comprehensive approach to identifying adaptive genetic variants.