Inference of genetic forces using a Poisson random field model with non-constant population size

Abstract

Aligned DNA sequences have been widely used for quantitatively analyzing and interpreting evolutionary processes. By comparing the information between intraspecific polymorphism with interspecific divergence in two sibling species, Poisson random field (PRF) theory offers a statistical framework with which various genetic parameters such as natural selection intensity, mutation rate and speciation time can be effectively estimated. A recently developed time-inhomogeneous PRF model has reinforced the original method by removing the assumption of stationary site frequency, but it keeps the assumption that the two sibling species share same effective population size with their ancestral species. This paper explores a relaxation of this biologically unrealistic assumption by hypothesizing that each of the two descendant species experienced a sudden change in population size at the times of the divergence from their most recent common ancestor. The newly developed PRF model with non-constant population size is applied to a set of 91 genes in African population of Drosophila to make statistical inference of the various genetic parameters under a hierarchical Bayesian framework and carried out with a multi-layer Markov chain Monte Carlo sampling scheme. In order to meet the intensive computational demand, a R program is integrated with C++ code and a parallel executing technique is designed to run the program with multiple CPU cores.