Preprint identifies mitigation strategies for rescaling biases in forward-in-time population genetic simulations

Researchers have posted a preprint to bioRxiv addressing a methodological concern that has attracted attention in the population genetics simulation community: the accuracy of parameter rescaling in forward-in-time simulations. Forward-in-time simulators such as SLiM are widely used to model evolutionary processes, including natural selection, demographic history, and mutation accumulation. Because simulating realistic population sizes over realistic timescales is computationally prohibitive, rescaling — reducing population size and generation count by a common factor — is a common pragmatic shortcut.

Several recent studies have documented biases introduced by this approach, particularly when using the SLiM simulator. The new preprint argues that many of these reported biases are not intrinsic to rescaling as a principle but are instead consequences of particular implementation choices. The authors show that switching to a different forward-in-time simulator substantially mitigates the biases described in the literature, suggesting that the concerns may be more tractable than previously implied.

The work is primarily of technical interest to researchers who rely on forward-in-time simulations for evolutionary analyses — including those studying population history, adaptation, and the generation of synthetic data for method benchmarking. The preprint has not yet been peer reviewed. No lead author or institutional affiliation is identified in the feed lede; full details are at the bioRxiv record.