Distinct genetic architecture in the tails of complex traits

Abstract

Complex traits are highly polygenic, with heritability explained by many hundreds of common variants of small effect together with rare variants of large effect1. Yet how this genetic architecture varies along the trait continuum has been underexplored, as has the role of natural selection in shaping this variation. Here we developed an approach based on polygenic risk scores that reveals widespread departures from common-variant architecture in one or both of the tails of 74 quantitative traits. These observations were replicated across ancestries, cohorts and repeated measures and using an alternative family-based approach2. Incorporating rare variants identified from sequence data resulted in marked reductions in these deviations, suggesting that rare alleles of large effect are key drivers of trait-tail architecture. Forward simulations showed that stabilizing selection could generate the observed patterns, whereas modelling reproductive success provided empirical support for the role of selection. These findings show that although complex traits are polygenic in the population at large, they have a distinct and less polygenic architecture in their tails due to selection. This has implications for rare-variant discovery and complex trait and disease prediction.