Detecting epidemic-driven selection: a simulation-based tool to optimize sampling design and analysis strategies

Throughout history, populations from numerous species have been decimated by epidemic outbreaks, like the 19th-century rinderpest outbreak in Cape buffalo ({approx} 90% mortality) and Black Death in humans ({approx} 50% mortality). Recent studies have raised the enticing idea that such epidemic outbreaks have led to strong natural selection acting on disease-protective variants in the host populations. However, so far there are few, if any, clear examples of such selection having taken place. This could be because so far studies have not had sufficient power to detect the type of selection an epidemic outbreak must induce: strong but extremely short-term selection on standing variation. We present here a simulation-framework that allows users to explore under what circumstances it is possible to detect epidemic-driven selection using standard selection scan methods like FSTand iHS. Using two examples, we illustrate how the framework can be used. Furthermore, via these examples, we show that comparing survivors to the dead has the potential to render higher power than more commonly used sampling schemes. And importantly, we show that even for outbreaks with high mortality, like the Black Death, strong selection may have led to only modest shifts in allele frequency, suggesting large sample sizes are required to obtain appropriate power to detect the selection. We hope this framework can help in designing well-powered future studies and thus lead to a clarification of the role epidemic-driven selection has played in the evolution of different species. Significance StatementOur study introduces a simulation-based framework, SimOutbreakSelection (SOS), which enables researchers to design studies that have power to detect epidemic-driven selection while taking sampling time points and demographic history into account. We use rinderpest in African Buffalo and the Black Death in Medieval Sweden as examples to showcase the framework. Via these examples we also show that large sample sizes are needed even for severe epidemics like the Black Death and that the often used sampling strategy where samples from before the epidemic and samples from after are compared is not always optimal.