Longitudinal genome-wide analysis reveals putative non-additive loci in trait development.

Abstract

Complex traits emerge from reciprocal interactions among genotype, environment, and developmental processes. Yet, standard genetic models assume purely additive effects, potentially obscuring non-additive effects. Here, we introduce a longitudinal log-linear variance and genotype-by-time model to detect associations from within-individual variation departing from additivity, i.e., putative non-additive effects. Applied to early growth (infant length and BMI) and cognitive traits (math and reading) of 45,000 to 65,000 individuals, we report 76 lead putative non-additive loci that are enriched 16-fold for cis-regulatory interactions. Of the 76, 6 overlap prior interaction studies (anthropometric) and only 3 loci overlap prior genome-wide association study (GWAS) (cognitive). Accounting for scale effects and linkage disequilibrium (LD), we observe that additive effects are correlated with putative non-additive effects, i.e., "effect pleiotropy." These results are consistent with non-additive genetic contribution to trait development, which may partly be absorbed by effects estimated under standard GWAS parameterization that assumes strict additivity.