Longitudinal modelling reveals widespread non-additive genetic effects underlying developmental plasticity

Abstract

Developmental plasticity, or the capacity of genotypes to produce different phenotypes under varying contexts, is poorly characterized in humans. Here we developed a longitudinal variance and a genotype-by-time model able to disentangle genetic effects from between-subject variability. Applied to early growth (infant length and BMI) and cognitive traits (math and reading) of 45,000 to 65,000 individuals, our models discovered 76 lead putative plasticity loci. Nearly all loci are novel; six replicate signals from independent interaction studies. We found evidence that additive genetic variation captures marginal effects from interactions, challenging the assumption that effect estimates from genome-wide association studies (GWAS) are purely additive. Notably, over 90% of putative plasticity loci are involved in distal intra-chromatin interactions implicating regulatory activity. Our findings provide molecular evidence that non-additive genetic variation contributes to complex traits and highlight the utility of longitudinal models in uncovering the biological underpinnings of trait plasticity.