Simulation-based statistical methods for reconstructing the history of hybrid human populations

Erkan Buzbas, Statistical Science

A large number of human populations are founded by contribution from multiple source populations in recent history, and therefore are highly admixed. However, complex mechanisms of admixture underlying the observed genetic patterns in humans are unknown. Understanding prominent mechanisms through which human populations have admixed is a major step towards building reliable models of human history and evolution, with a wide range of implications for population genetics, anthropology, and personalized medicine. Although mechanistic models of admixture histories for human populations have been described, statistical inference about biologically interpretable parameters have proven difficult to perform due to complexity of these models. The goal of this proposal is to obtain funding to develop simulation-based statistical methods to perform inference under complex models of admixture using a framework known as approximate Bayesian computation. Using these novel methods, we further aim to analyze genetic data from world-wide admixed human populations, including African Americans, Central African Pygmies, Native Americans, and Cape Verdeans. Our ultimate goal is to reconstruct the history of these populations, and describe the prominent mechanisms of admixture in humans.