Effects of rescaling forward-in-time population genetic simulations.

Forward-in-time population genetic simulations enable modelling of a wide array of complex evolutionary scenarios. Simulating small genomic regions, and rescaling by reducing population size by a scaling factor while maintaining population-scaled parameters, are common approaches for improving computational tractability. However, simulating whole chromosomes in large populations remains computationally prohibitive, and it is unclear what scaling factors produce consistent evolutionary dynamics between rescaled and unscaled populations. Here, we thoroughly test the effects of rescaling in populations experiencing direct and linked effects of selection for various scaling factors and chromosomal lengths in Drosophila melanogaster-like populations by comparing them with theoretical expectations. We find that while rescaling has minimal effects with most length and scaling factor combinations, simulating long regions with large scaling factors can introduce biases in summary statistics including neutral diversity, allele frequencies, linkage disequilibrium, and nonneutral divergence, even within expected diffusion limits. These deviations occur particularly when the crossover rate exceeds 0.44 per individual/generation because substantial multiple-crossover events are likely to occur within individuals, reducing the effect of recombination. In addition, when the genome-wide deleterious mutation rate is high (U≫1) in highly rescaled long regions, we observe increased Hill-Robertson interference effects and progeny skew, the extent of which was strongly dependent on the fitness effects of selected mutations. We find that hitchhiking effects near functional regions are relatively unaffected by rescaling the population, even with large scaling factors. Our findings expose potential pitfalls when simulating long regions with rescaling and highlight parameter spaces within which expected evolutionary dynamics are conserved.