Pathway-specific polygenic scores substantially increase the discovery of gene-adiposity interactions impacting liver biomarkers.

Abstract

Polygenic scores (PGSs) are appealing for detecting gene-environment interactions due to the aggregation of genetic effects and reduced multiple testing burden compared to single-variant genome-wide interaction studies (GWISs). However, standard PGSs reflect many different biological mechanisms, limiting interpretation and potentially diluting pathway-specific interaction signals. Previous work has uncovered a significant genome-wide PGS×Adiposity signal impacting liver function, but there is an opportunity for additional and more interpretable discoveries. Here, we leveraged pathway-specific polygenic scores (pPGSs) to discover mechanism-specific gene-adiposity interactions. We tested for body mass index (BMI) interactions impacting three liver-related biomarkers (ALT, AST, and GGT) using (1) a standard, genome-wide PGS, (2) an array of pPGSs containing variant subsets derived from KEGG pathways, and (3) a GWIS. For ALT, we identified 49 significant pPGS×BMI interactions at a Bonferroni corrected p < 2.7 × 10^-4, 80% of which were not explained by genes close to the 8 loci found in the associated GWIS. Across all biomarkers, we found interactions with 83 unique pPGSs. We tested alternate pathway collections (hallmark, KEGG Medicus), finding that the choice of pathway collection strongly impacts discovery. Our findings reinforced known biology (e.g., glycerolipid metabolism and hepatic lipid export affecting ALT release) and captured additional phenomena (e.g., actin cytoskeleton remodeling-associated variants alter the liver's robustness to lipid mechanical stress and thus GGT release). These results support the use of pPGSs for well powered and interpretable discovery of pPGS×E interactions with adiposity-related exposures for liver biomarkers and motivate future studies using a broader collection of exposures and outcomes.