"Why I am crazy about linear models: adventures into the contribution of genetics to complex traits"
Biobank-scale genotyping has enabled the routine use of statistical genetics applications, such as Mendelian randomization, polygenic scores and heritability estimations. However, many questions remain less well explored, such as the contribution of gene-by-environment interactions (GxE) and the role of rare coding variants (RV) in complex traits.
|
Leveraging the speed, versatility and robustness of large linear models, we sought to address these two questions. First, we introduce MonsterLM, a multiple linear regression method that does not rely on model specification and provides unbiased estimates of variance explained by GxE. Identification of GxE is crucial to understand the interplay of environmental effects on complex traits. However, current methods evaluating GxE on biobank-scale datasets have limitations.
|
We demonstrate robustness of MonsterLM through comprehensive genome-wide simulations using real genetic data from 325K UK Biobank participants. We estimate GxE using waist-to-hip-ratio, smoking, and exercise as the environmental variables on 13 outcomes and find significant GxE for 8 environment-outcome pairs. The majority of GxE variance involves SNPs without strong marginal or interaction associations. Second, we developed a novel framework, the Rare variant heritability (RARity) estimator, to assess RV heritability (h2RV) without assuming a particular genetic architecture. We applied RARity to 31 complex traits in the UK Biobank (N=167K) and showed that gene-level RV aggregation suffers from 79% (95% CI: 68-93%) loss of h2RV. Using unaggregated variants, 27 traits had h2RV>5%, with height having the highest h2RV at 21.9% (95% CI: 19.0-24.8%).
|
|
The total heritability, including common and rare variants, recovered pedigree-based estimates for 11 traits. We also used RARity to reveal 11 novel gene-phenotype relationships. Finally, we demonstrated that in silico pathogenicity prediction (variant-level) and gene-level annotations do not generally enrich for RVs that over-contribute to complex trait variance, and thus, novel methods are needed to predict RV functionality.
|
|
|
|
