AM Seminar: Genetically Admixed Groups as a Laboratory for Mathematical Modeling and Discovery
Presenter: Dr. Aw, Department of Genetics, University of Pennsylvania
Description: Admixed groups consist of individuals descended from two or more historically separated ancestral populations, and they are underrepresented in biomedical studies. Admixed individuals are unique in that they carry mosaics of ancestral segments within their genome, so their genetic information is typically summarized as a pair of genotype and local ancestry data matrices. I will present two research projects on admixed groups: one applying statistical models to study genetic architecture and polygenic risk, and another using biomedical data analysis to motivate combinatorial and probabilistic questions. In the “math to genetics” direction, we describe structural causal models that show local and global genetic ancestry are instruments for epistasis. These models of genetic architecture imply that certain polygenic scores can differentiate between cis and trans epistases, and highly similar cross-ancestry genetic effects do not rule out pervasive gene-gene or gene-environment interactions. In the “genetics to math” direction, we study the enumeration of genotype and local ancestry data matrices — which we call admixed arrays — subject to constraints that arise naturally in biomedical applications. Using saddle-point approximation and complex martingale techniques, we show that admixed arrays admit a different independence heuristic than the closely related binary contingency tables (e^(–1/4) vs e^(–1/2) correction factor). If time permits, we will discuss ongoing work on designing algorithms for performing exact and approximate enumeration.
About the speaker: Alan Aw is a mathematical scientist specializing in human statistical and population genomics. He is currently a postdoctoral researcher with the Department of Genetics at the University of Pennsylvania, having obtained a PhD in Statistics at UC Berkeley and studied Applied Mathematics as an undergraduate. His research centers on the genetics and mathematical modeling of underrepresented groups. This includes statistical modeling and analyses of Biobank-scale admixed cohorts to better understand the genetic architecture of biomedical traits and improve genetic risk prediction, developing non-parametric hypothesis testing methods for genomics, and interdisciplinary approaches to studying European demographic history inclusive of indigenous Siberians. He is a member of the PRIMED Consortium and a trainee under a National Institutes of Health T32 Grant in Genomic Medicine.
Hosted By: Applied Mathematics