Abstract
Both rare and common genetic variants contribute to human disease, and emerging evidence suggests that they combine additively to influence disease liability. However, the non-linear relationship between disease liability and disease prevalence means that risk variants may have more severe phenotypic consequences in high-risk polygenic backgrounds and minimal impact in low-risk backgrounds, resulting in uneven selection across the population. As a result, selection coefficients may be better modeled as distributions that differ across populations, time, environments, and individuals than as single values. As the number of genes contributing to a trait and epistasis between alleles increases, so does phenotypic variance, pushing more individuals to extreme phenotypes and enhancing negative selection. Because disease-relevant phenotypes may be masked in certain genetic backgrounds, we argue that the polygenic background should be considered when designing experiments to characterize the molecular underpinnings of complex traits.