Evidence for evolutionary shifts in the fitness landscape of human complex traits

Selection alters human genetic variation, but the evolutionary mechanisms shaping complex traits and the extent of selection9s impact on polygenic trait evolution remain largely unknown. Here, we develop a novel polygenic selection inference method (Polygenic Ancestral Selection Test Encompassing Linkage, or PASTEL) relying on GWAS summary data from a single population. We use model-based simulations of complex traits that incorporate human demography, stabilizing selection, and polygenic adaptation to show how shifts in the fitness landscape generate distinct signals in GWAS summary data. Our test retains power for relatively ancient selection events and controls for potential confounding from linkage disequilibrium. We apply PASTEL to nine complex traits, and find evidence for selection acting on five of them (height, BMI, schizophrenia, Crohn9s disease, and educational attainment). This study provides evidence that selection modulates the relationship between frequency and effect size of trait-altering alleles for a wide range of traits, and provides a flexible framework for future investigations of selection on complex traits using GWAS data.