Bayesian methods for inferring selection and demographic from historical and contemporary DNA sequences

Abstract

In this thesis, I propose Bayesian methods to infer selection coefficients and allele age using time-series data and uncover the demographic history given contemporary whole-genome data.

Approximate Bayesian computation and Markov chain Monte Carlo method are widely used in solving population genetics problems. Time-series allele frequency problems often are modeled by the Hidden Markov Model, which is complex to make accurate inferences from. Here I employ a particle marginal Metropolis-Hastings method to make co-estimates of selection coefficients
and allele age based on the single-locus Wright-Fisher model and the two-locus Wright-Fisher model. In addition, I also propose an EP method with the ABC algorithm to extract demographic information from whole-genome contemporary data.

For each method, I make simulation studies to present the accuracy of the method and apply the method to re-analysis of published data to show the method can achieve effective and accurate estimates for genetic parameters of interests.

Date of Award	12 May 2020
Original language	English
Awarding Institution	University of Bristol
Supervisor	Mark A Beaumont (Supervisor)