The historical Moran model

TL;DR

This paper develops a Markov process framework to analyze genealogical evolution in a multi-type Moran model with selection and mutation, providing new insights into ancestral lineages and genealogical distances.

Contribution

It introduces a path-valued Markov process for genealogical lines and a novel backward process to characterize ancestral distributions and their limits under selection.

Findings

Convergence of ancestral lines as time tends to infinity.

Representation of stationary type distribution via the backward process.

Genealogical distances are stochastically smaller under selection.

Abstract

We consider a multi-type Moran model (in continuous time) with selection and type-dependent mutation. This paper is concerned with the evolution of genealogical information forward in time. For this purpose we define and analytically characterize a path-valued Markov process that contains in its state at time $t$ the extended ancestral lines (adding genealogical distances) of the population alive at time $t$. The main result is a representation for the conditional distribution of the extended ancestral lines of a subpopulation alive at a fixed time $T$ (present time) given the type information of the subpopulation at time $T$ in terms of the distribution of the sample paths (up to time $T$) of a special Markov process (different from the ancestral selection graph) to which we refer as backward process. This representation allows us both to prove that the extended ancestral lines converge in the limit $T \to \infty$ if the type information converges in the limit $T \to \infty$ and to study the resulting limit of the extended ancestral lines by means of the backward process. The limit theorem has two applications: First, we can represent the stationary type distribution of the common ancestor type process in terms of the equilibrium distribution of a functional of the backward process, where in the two type case we recover the common ancestor process of Fearnhead if we let the population size tend to infinity. Second, we obtain that the conditioned genealogical distance of two individuals given the types of the two individuals is distributed as a certain stopping time of a further functional of the backward process which is a new approach towards a proof that genealogical distances are stochastically smaller under selection.