14,000 years of natural selection

Razib Khan's Unsupervised Learning

The full episode is available on: https://www.razibkhan.com/p/14000-years-of-natural-selection

On this episode of Unsupervised Learning Razib talks about what we have learned from a blockbuster new preprint, Pervasive findings of directional selection realize the promise of ancient DNA to elucidate human adaptation. Synchronously released was the Ancient Genome Selection browser, which allows you to trace the allele frequency of variants of interest over the last 10,000 years. Razib covers:

• The relationship of selection to adaptation and the Darwinian understanding of evolution

• Non-genetic selection

• Types of biological selection like positive, negative, background and balancing selection

• Hard vs. soft sweeps and their relevance to detecting selection in the genome

• Older forms of natural selection detection between species (dN/dS, Tajima’s D)

• Newer forms of selection detection within species with haplotype structure, outlier SNP analysis and site frequency spectra

• The Generalized Linear Mixed Model used to model allele frequency change over time, and estimates of selection in cases where population structure and drift are not sufficient

• Specific examples of SNPs whose variation can be examined in the browser and are clearly cases of selection

• Survey of traits that were revealed under selection, including blood groups, pigmentation and intelligence

• Critiques of the methods due to not accounting for drift or population structure, and its limitations in relation to the ability to port across populations due to LD structure