Haploids, polymorphisms and fluctuating selection

I analyze the joint impact of directional and fluctuating selection with reversible mutation in finite bi-allelic haploid populations using diffusion approximations of the Moran and chemostat models. Results differ dramatically from those of the classic Wright-Fisher diffusion. There, a strong dispersive effect attributable to fluctuating selection dissipates nascent polymorphisms promoted by a relatively weak emergent frequency dependent selective effect. The dispersive effect in the Moran diffusion with fluctuations every birth-death event is trivial. The same frequency dependent selective effect now dominates and polymorphism is promoted. The dispersive effect in the chemostat diffusion with fluctuations every generation is identical to that in the Wright-Fisher diffusion. Nevertheless, polymorphism is again promoted because the emergent frequency dependent effect is doubled, an effect attributable to geometric reproduction within generations. Fluctuating selection in the Moran and chemostat diffusions can also promote bi-allelic polymorphisms when one allele confers a net benefit. Rapid fluctuations within generations are highly effective at promoting polymorphism in large populations. The bi-allelic distribution is approximately Gaussian but becomes uniform and then U-shaped as the frequency of environmental fluctuations decreases to once a generation and then once every multiple generations. Trade-offs (negative correlations in fitness) help promote polymorphisms but are not essential. In all three models the frequency dependent effect raises the probability of ultimate fixation of new alleles, but less effectively in the Wright-Fisher diffusion. Individual-based forward simulations confirm the calculations.