Stochastic individual-based models with power law mutation rate on a general finite trait space (Q2243919)

The authors study a stochastic individual-based model for the evolution of a haploid population reproducing asexually. In this model, the space of possible traits in the population is given by the vertices of a directed finite graph \(G = (V,E)\). In this model, an individual of trait \(v\) gives birth at rate \(b_v\) and (naturally) dies at rate \(d_v\). The population is kept constant of order \(K\) thanks to a competition among individuals, i.e., each particle of type \(w\) will kill a particle of type \(v\) at rate \(c_{v,w}/K\). The children of an individual of trait \(v\) are typically of trait \(v\), except at some (rare) mutations event with probability \(\mu_K\). The trait of a mutant child is chosen according to some probability distribution among the neighbors of \(v\). The authors consider a mutation rate \(\mu_K := K^{-1/\alpha}\) for \(\alpha > 0\). For \(\alpha \geq 1\) this results in several mutant traits being present at the same time and competing for invading the resident population. They describe the evolution in time of the order of magnitude of the number of individuals of each trait on the \(\log K\) scale as \(K \to \infty\). More precisely, they show that under some generic assumptions on the initial distribution, the processes defined by \[ \beta_v(t) = \lim_{K\to\infty}\frac{\log 1 + N_v(t)}{\log K}\text{ in probability} \] is a.s. deterministic, where \(N_v(t)\) is the number of individuals of trait \(v\). The process \((\beta_v,v \in V)\) can be described by an explicit algorithm using relative fitness of each trait under the current equilibrium distribution of the population. Heuristically, the dominant traits in the population (which are carried by a number of order \(K\) of individuals) almost instantly arrange in an equilibrium state corresponding to the Lotka-Volterra equilibrium of the associated deterministic ODE system. Then, every trait with more that \(K^{1/\alpha}\) individuals will create mutants among its neighbors in the trait space, every trait with more that \(K^{2/\alpha}\) individuals create mutants in a ball of diameter \(2\) centred around its positions, etc. The non-dominant traits will then grow at an exponential rate depending on their natural birth and death rate, as well as the competition imposed by the dominant traits. Every time a new trait becomes dominant, the equilibrium of dominant traits is modified. This results in \((\beta_v)\) being continuous piecewise affine functions. Several examples of trait spaces are considered, which exhibit paradoxical behavior for the evolution. For example, it is possible to construct a trait space in which mutations occur clockwise, such that the observed macroscopic population dominant trait goes counter-clockwise. It is also possible to construct trait spaces exhibiting Braess paradox, i.e., such that closing a possible mutation route decreases the time taken by the population to go from the less fit to the fittest population, or to construct examples in which the macroscopic population suddenly gains an arbitrarily large number of mutations.