Random collision model for random genetic drift and stochastic difference equation (Q1057792)

The author studies by Monte Carlo simulation a one-locus multiallelic genetic model contemplating random genetic drift, mutation (Wright-Kimura infinite allele model), and overdominant selection (all heterozygotes have fitnesss 1-s, with \(s>0\), and all homozygotes have fitness 1). It can be generalized to other fitness schemes. Rather than taking the approach of \textit{T. Maruyama} and \textit{M. Nei} [Genetic variability maintained by mutation and overdominant selection in finite populations. Genetics 98, 441-459 (1982)] of solving analytically (case \(s=0)\) or numerically a stochastic difference equation, the author uses simulated direct sampling based on an ingenious random collision model. To project from time t to time \(t+\Delta t\), collisions between alleles occur at a rate proportional to \(\Delta\) t for each allele. For random genetic drift, collisions of two alleles are considered [to form the newborn zygotes that replace the dead ones]. For selection, collision of four alleles are required (corresponding to the choice of two mates, each one with two alleles). There are other technical details of some importance concerning the decomposition of the variance-covariance matrix. The numerical results obtained for the approximately asymptotic time-average heterozigosity in a variety of situations are very similar to those obtained by Maruyama and Nei using stochastic difference equations.