Statistics Seminar: Evolutionary Analysis of Biobank-Scale Data to Study Population History, Mutation Rates, Natural Selection, and the Genetics of Complex Traits

Presenter: Joshua Schraiber, Research Scientist, University of Southern California and Cornell University

Description: The advent of biobank data has presented an unprecedented opportunity to understand the evolutionary history of the human species. By collecting genotype data from hundreds of thousands of individuals, we can look closely at recent evolutionary history, including recent demography, mutation rates, and the action of strong natural selection. Moreover, biobank data with phenotypes enables probing into the details of the genetic architecture of complex traits. However, the large sample sizes and low frequency alleles found in such datasets often result in violations of common modeling assumptions in population and statistical genetics. To address this gap, and gain the most possible information out of this data, I will present biobank-scale evolutionary models. I will first show that ultra-rare mutations often violate the assumption that every mutation only occurred once in the history of the sample. By developing a model that explicitly accounts for recurrent mutation, we are able to use ultra-rare mutations to learn fine-scale details about recent demography history and mutation rates. We then extended this model to account for natural selection, and developed a Bayesian approach to share information between genes and learn the distribution of fitness effects at the level of individual genes. Finally, I will show how rare variants violate the standard assumptions of models of complex trait architecture and demonstrate a modeling approach that can account for effect sizes among rare variants. This will show that the distribution of effect sizes among common variants in the human genome is approximately scale-free. Together, these findings enhance our understanding of human biology and evolution in the biobank era.

About the speaker: Joshua Schraiber is a Research Scientist working with Doc Edge at USC and Matt Pennell at Cornell University. His background is in evolutionary modeling and statistical methods development. Prior to his current role, he was a Research Scientist at AncestryDNA and Illumina before returning to academia.

This seminar is hosted by Professor Yunyi Shen.