Genetic Variation and Modern Human Origins.

Abstract

This dissertation focuses on the use of mathematical modeling and statistical inference applied to within- and between-species genetic data. I develop tools for analyzing genetic variation within populations, use these tools to identify evolutionary models that are consistent with worldwide human genetic variation, and develop and analyze the performance of methods for estimating population and species relationships from genomic data.

Typical human population-genetic datasets contain genotypes from closely related individuals. Because relatives share recent common ancestors, their presence in such datasets biases estimates of genetic diversity. Thus, using pedigree analysis techniques, I derive an estimator that corrects this bias, enabling me to accurately measure genetic diversity in a study of human origins. Further, I extend the estimator to accommodate samples of arbitrary ploidy, solving the most general case and enabling investigators to apply my estimator to any species.

Next, I apply my estimator as well as other measures of variation to worldwide human genetic data to investigate the extent to which models of human origins are consistent with observed patterns. I develop models of human demographic history to represent two main hypotheses, the ``out-of-Africa'' and ``multiregional'' hypotheses, for modern human origins. Using both simulations and analytical formulas, I compare measures of genetic variation observed from human data with those predicted by my models. I find that the model representing the out-of-Africa hypothesis produces patterns that mimic those observed from human data, whereas the model representing the multiregional hypothesis produces opposite patterns. These results lend support to the out-of-Africa hypothesis.

Finally, I investigate sources of inter-population or inter-species genetic variation that can cause evolutionary relationships estimated from different genetic loci to disagree and that therefore make it difficult to estimate population or species relationships from genetic data. I develop a method that circumvents this difficulty in phylogenetic tree reconstruction and demonstrate that it is both fast and consistent and that it performs well on genome-scale data. I also analyze the performance of various methods for inferring population and species trees when relationship estimates conflict across loci. These results provide investigators with tools for facilitating accurate inference of population and species relationships.