Quantitative system drift

We consider a biological system composed of multiple genetically variable components, the combined result of which is a quantitative trait under stabilizing selection for an optimal value. We show mathematically that, while the mean value of the system is ultimately constrained to remain near its optimum, the mean contributions of individual components are free to drift far from their initial values. Each component's drift, though qualitatively identical to neutral drift, is slower by a factor that depends on the fraction of the system's genetic variance contributed by the component. We further show that symmetric mutation between alleles that increase and decrease components' contributions to the system imposes a weak long-term brake on components' drift. Our results provide a population-genetic basis for "system drift," the concept that individual components of a biological system can evolve despite selective constraint on their combined product. A special case is a single polygenic trait under stabilizing selection, where our results predict that the mean contributions to the trait of different subregions of the genome, such as the chromosomes, can drift despite constraint on the genome-wide value. To indicate the broad applicability of our results, we explore their implications for the evolution of gene expression, selection against interspecific hybrids, selection against turnovers of sex-determining systems, and the division of labor in mutualisms.