Dynamics of fitness distributions in the presence of a phenotypic optimum: an integro-differential approach**This work has been carried out in the framework of Archimède Labex of Aix-Marseille University. The project leading to this publication has received funding from Excellence Initiative of Aix-Marseille University—A*MIDEX, a French ‘Investissements d’Avenir’ programme, from the European Research Council under the European Union’s Seventh Framework Programme

Abstract

We propose an integro-differential description of the dynamics of the fitness distribution in an asexual population under mutation and selection, in the presence of a phenotype optimum. Due to the presence of this optimum, the distribution of mutation effects on fitness depends on the parent’s fitness, leading to a non-standard equation with ‘context-dependent’ mutation kernels.Under general assumptions on the mutation kernels, which encompass the standard n − dimensional Gaussian Fisher’s geometrical model (FGM), we prove that the equation admits a unique time-global solution. Furthermore, we derive a nonlocal nonlinear transport equation satisfied by the cumulant generating function of the fitness distribution. As this equation is the same as the equation derived by Martin and Roques (2016) while studying stochastic Wright–Fisher-type models, this shows that the solution of the main integro-differential equation can be interpreted as the expected distribution of fitness corresponding to this type of microscopic models, in a deterministic limit. Additionally, we give simple sufficient conditions for the existence/non-existence of a concentration phenomenon at the optimal fitness value, i.e. of a Dirac mass at the optimum in the stationary fitness distribution. We show how it determines a phase transition, as mutation rates increase, in the value of the equilibrium mean fitness at mutation-selection balance. In the particular case of the FGM, consistently with previous studies based on other formalisms (Waxman and Peck, 1998, 2006), the condition for the existence of the concentration phenomenon simply requires that the dimension n of the phenotype space be larger than or equal to 3 and the mutation rate U be smaller than some explicit threshold.The accuracy of these deterministic approximations are further checked by stochastic individual-based simulations.