On inferring and interpreting genetic population structure - applications to conservation, and the estimation of pairwise genetic relatedness

Abstract

<p>The presence of population structure is ubiquitous in most wild populations of species. Detecting genetic population structure and understanding its consequences for the evolutionary trajectories of species has shaped a lot of our understanding of the process of evolution. This delineation of subdivision within a population plays an important role in several allied fields, including conservation genetics, association studies, phylogeography, and quantitative genetics. This dissertation addresses methods to infer and interpret subpopulation structure. In this regards, I discuss the standing motivation for developing new analytic tools, a classic population</p> <p>genetics study of the imperiled freshwater turtle, <em>Emys blandingii</em>, the development of a fast, likelihood based estimator of subpopulation structure, MULTICLUST, and a likelihood based method to infer pairwise genetic relatedness in the presence of subpopulation structure.</p> <p>Our analyses of population structure in midwestern populations of <em>Emys blandingii</em> detected considerable genetic structure within and among the sampled localities, and revealed ancestral gene flow of <em>E. blandingii</em> in this region north and east from an ancient refugium in the central Great Plains, concordant with post-glacial recolonization timescales. The data further implied</p> <p>unexpected links between geographically disparate populations in Nebraska and Illinois. Our study encourages conservation decisions to be mindful of the genetic uniqueness of populations of <em>E. blandingii</em> across its primary range.</p> <p>Analyses of both simulated and empirical data suggests that MULTICLUST infers structure consistently (reproducible results), and is time effcient, compared to the popular Bayesian</p> <p>MCMC tool, STRUCTURE (Pritchard et al. (2000b)). The new likelihood estimator of pairwise genetic relatedness also has the least bias, and mean squared error in estimating relatedness</p> <p>in full-sibling, half-sibling, parent-offspring, and a variety of other related dyads, compared to the methods of Anderson and Weir (2007), Queller and Goodnight (1989), Lynch and Ritland (1999).</p> <p>Overall, this dissertation lays the grounds for several interesting biological and statistical questions that can be addressed with a robust framework for identification of subpopulation structure.</p>