Inference of demographic histories in structured populations with an application to human evolution.

Overview & Synopsis

Inferring the demographic history of species is one of the greatest challenges in populations genetics. Such histories are often represented as histories of size changes, ignoring population structure. Alternatively, when structure is assumed, they are defined a priori as population trees and not inferred. This thesis aims to contribute methods and tools for reconstructing the demographic history of populations that have been structured into an unknown number of sub-populations for long periods of time.We present two approaches to demographic inference in the presence of structure. The first is based on the IICR (Inverse Instantaneous Coalescence Rate) which is a function of the demographic model and sampling scheme, and can be estimated for a single diploid individual. The proposed method fits observed IICR curves with exact IICR curves obtained under piecewise stationary symmetrical island models, and infers the number of islands, their common size, and the amount of gene flow in different periods of time. Our application to a set of five human PSMCs yielded demographic histories that are in agreement with previous studies suggesting ancient human structure.The second approach assumes multiple genetic samples, and is centered on the use of the SFS (Site Frequency Spectrum) as a summary statistic for demographic inference. We focus on the efficient computation of the exact expected SFS under a general model of population structure, and show that, for sample sizes up to 26 haploids, it is possible to achieve good numerical accuracy and performance in symmetrical island models by exploiting the sparsity pattern of the transition matrix for the associated Markov process.