Skip to main content
Dryad

Data for: Assessment of estimating selection coefficients in non-Wright-Fisher populations

Cite this dataset

Rêgo, Alexandre; Stelkens, Rike; Zhivkoplias, Erik (2023). Data for: Assessment of estimating selection coefficients in non-Wright-Fisher populations [Dataset]. Dryad. https://doi.org/10.5061/dryad.zgmsbccgb

Abstract

Selection coefficients are a useful parameter in evolutionary studies. Given that most tools to estimate selection coefficients assume population dynamics of a Wright-Fisher model, but most real-world populations violate Wright-Fisher model assumptions, the accuracy of these tools under more realistic scenarios of evolution is debatable. In this study, we test a common tool (WFABC) to assess estimates of selection coefficients in simulated populations which do and do not adhere to typical Wright-Fisher assumptions. Specifically, we look at simulated non-Wright-Fisher populations which experience evolutionary rescue. We see that fluctuating demography plays an important role in our ability to infer natural selection. Increased rates of sweeps due to bottlenecks increase error in estimates of selection by up to 5.45 times relative to typical WF-adhering populations. In a moderately polygenic model of adaptation, more severe bottlenecks produce harder sweeps such that selected sites interact with surrounding sites, including other selected sites, to increase error in estimates. This may lead to an erroneous excess of the estimated number of sites under selection within a population experiencing evolutionary rescue. Our results place an emphasis on thoughtful analyses of estimates of selection obtained from real-world populations which don’t adhere to typical model assumptions.

Methods

The population demographic and genomic time-series data was generated solely through simulations using the SLiM 3.0 software. The data was then input into WFABC software to generate estimates of selection coefficients and subsequently analyzed in R.

Usage notes

Data can be accessed using R, python 3, and SLiM 3.0.

Funding

Swedish Research Council, Award: 2017‐04963