Genetic algorithms and evolutionary landscapes

The metaphor of fitness landscapes is common in evolutionary biology, as a way to visualise the change in allele or phenotypic frequencies of a population under selection. Understanding how different factors in the evolutionary process affect the trajectory of the population across the landscape is of interest to both theoretical and empirical evolutionary biologists. However, fitness landscape studies often have to rely heavily on mathematical methods that are not easy to access by biologically trained researchers. Here, we used a method borrowed from engineering - genetic algorithms - to simulate the evolutionary process and study how different components affect the path taken through a phenotypic fitness landscape. In a simple study, we compare five selection models that reflect different degrees of dependency of fitness on trait quality: this includes strengths of selection, trait-quality dependent reproductive hierarchy and the amount of stochasticity in the reproductive process. We include an analysis of other evolutionary variables such as population size and mutation rate. We analyse a game theory problem, as a test landscape, that lends itself to analysis through a deterministic mathematical simulation, which we use for comparison. Our results show that there are differences in the speed with which different models of selection lead to the fitness optimum. Author summaryEvolution and adaptation in biology occurs in fitness landscapes, multidimensional spaces representing all possible genotypic or phenotypic combinations, where population adapt by following the cline of the fitness dimension. The study of adaptation on complex fitness landscapes has so far been limited by the need for mathematically heavy methods. Here, we present a simulation modelling framework, genetic algorithms, that can be used for evolutionary simulations of a population on a fitness landscape of chosen features and with custom evolutionary parameters.