Diffusion processes and coalescent trees

TL;DR

This paper explores the duality between coalescent trees and diffusion processes in population genetics, highlighting connections with orthogonal polynomials and subordinated Wright--Fisher diffusions, including novel time-subordinated structures.

Contribution

It provides a detailed analysis of the duality between coalescent processes and diffusion models, introducing new subordinated Wright--Fisher diffusions and related structures.

Findings

Connection between coalescent trees and diffusion processes clarified

Introduction of inverse Gaussian subordinators in Wright--Fisher diffusions

Novel time-subordinated forest structure in Kingman coalescent

Abstract

We dedicate this paper to Sir John Kingman on his 70th Birthday. In modern mathematical population genetics the ancestral history of a population of genes back in time is described by John Kingman's coalescent tree. Classical and modern approaches model gene frequencies by diffusion processes. This paper, which is partly a review, discusses how coalescent processes are dual to diffusion processes in an analytic and probabilistic sense. Bochner (1954) and Gasper (1972) were interested in characterizations of processes with Beta stationary distributions and Jacobi polynomial eigenfunctions. We discuss the connection with Wright--Fisher diffusions and the characterization of these processes. Subordinated Wright--Fisher diffusions are of this type. An Inverse Gaussian subordinator is interesting and important in subordinated Wright--Fisher diffusions and is related to the Jacobi Poisson Kernel in orthogonal polynomial theory. A related time-subordinated forest of non-mutant edges in the Kingman coalescent is novel.