Biological Insights From Population Differentiation

Abstract

Population genetics studies the genetic variation within and between populations to gain understanding of human history and insight into underlying biological processes. My dissertation introduces three distinct methods: a linear-time principal components analysis (PCA) algorithm, a scan for natural selection along continuous principal components (PCs), and a relationship between the cross-population correlation of genetic effects at all single nucleotide polymorphisms (SNPs) and the correlation of genetic effects at typed SNPs. These methods are all related to the statistical concept of the correlation matrix. The first two build off the genetic correlation matrix across individuals (also known as a genetic relationship matrix, or GRM), and the last on the correlations between SNPs in a population (also known as the linkage disequilibrium matrix, or LD matrix). With the PCA algorithm, we are now able to study the population structure of European American and British populations with finer resolution using very large population samples (55k and 113k samples, respectively). These PCs were fed into the scan for natural selection that detected signals of selection at known loci as well as several novel loci, including a gene protective against alcoholism in Europeans and several genes associated with blood pressure in the British. Lastly, using the SNP LD patterns in several populations we computed a factor that can be used in cross-population heritability scans to correct for the differential tagging efficiency within those populations.