Continuous time multi-allelic stepping stone models in population genetics (Q794569)

The paper considers a general class of one-locus multiallelic stepping- stone genetic models in continuous time incorporating geographical differentiation with mutation among colonies [the colonies form a countable set S], mutation, selection, and random genetic drift. It extends to the multiallelic case a previous paper of the author [ibid. 20, 213-242 and 723-733 (1980; Zbl 0494.92012 and 92013, respectively)] concerned with two alleles, which is summarized here in section 3. Mathematically, this gives a class of multidimensional diffusion processes defined on \(X_ d=(\Delta_ d)^ S\), where \(\Delta_ d=\{(x_ 1,...,x_ d): x_ 1,...,x_ d\geq 0\); \(x_ 1+...+x_ d=1\}\) is the simplex of all possible gene frequencies of d alleles for a particular colony. Diffusion terms come from random genetic drift and drift terms from migration, mutation, and selection. Section 2 considers the mathematically more interesting infinite-allelic model, which ''provides a concrete and analyzable example of infinite dimensional diffusion processes'' defined on \(X_{\infty}=(\Delta_{\infty})^ S\); under suitable mild assumptions and following the semi-group technique of previous work on the subject, the existence and uniqueness of a solution is established. Sections 4, 5 and 6 are important - at least from the population genetics point of view - and new. They consider the d-allelic case with d finite. Excluding selection to avoid quadratic drift terms, existence and uniqueness of solutions is established under mild assumptions and the set of steady-state distributions is shown to be non-empty, compact and convex (the extremal elements are also characterized). Excluding now both selection and mutation, a nice multidimensional Ornstein-Uhlenbeck process results as the scaling limit both in time and space (Section 5) or only in space (Section 6); these results can provide useful approximations.