The 84th Annual Meeting of the American Association of Physical Anthropologists (2015)


Using coalescent simulations to understand population dynamics of the admixture process

ANTHONY J. KOEHL and JEFFREY C. LONG.

Anthropology, University of New Mexico

March 26, 2015 3:45, Lindbergh Add to calendar

The purpose of our study is to investigate the accuracy of maximum likelihood estimates of genetic ancestry under different models of the admixture process. We use Fastsimcoal to simulate two models of admixture. The first model, termed Single Admixture Event (SAE), depicts a single event of interbreeding between two isolated source populations. The second model, termed One-Way Gene Flow (OGF), depicts a constant influx of genes from one source population into an admixed group. We simulated varying levels of contributions from the parents in both models, used maximum likelihood to estimate ancestry in the simulated populations, and then compared the estimates to the known values from the simulations. We allowed for the misspecification of source populations to increase the realism of our statistical analyses. To do this, we simulated proxies for the source populations and substituted them in our analyses for the true source populations. Proxies are populations related to the source population, but did not contribute to the admixture event. We found that maximum likelihood estimation had low bias with a single admixture event and equal contributions from the true source populations. However, they were biased when in versions of the SAE wherein one source contributed more than the other to the admixed population, overestimating the minor contributor. The bias increases nearly threefold when we substituted proxies for the true source populations. One-way gene flow created biases similar to those created by single event admixture. Thus, a more complex population mixing process did not require a more sophisticated analysis.