Age-dependent genetic effects on body mass index in the uk biobank: a single-nucleotide polymorphism level approach

Human phenotypes, including gene expression, often exhibit changes over the lifetime. However, in standard genome-wide association studies (GWAS), a possible age dependency is not taken into account. So far, few studies have investigated age-dependent genetic effects in the context of GWAS. Therefore, this study set out to establish whether age-by-genotype information can be used to better understand the genetic basis of such traits, what the best way to model and discover age-dependent genetic effects is, and how to fully leverage age-by-genotype information. Body Mass Index (BMI) in the UK Biobank (UKB) was selected as the trait of interest because BMI, as well as its genetic and environmental influences, changes with age, making BMI a good candidate for uncovering potential age-dependent genetic effects. First, we performed age-stratified GWAS of BMI for three independent age intervals (40-52, 53-61, and 62-73) in the UKB. We then computed global genetic correlations using linkage disequilibrium score regression (LDSC) and local genetic correlations using Local Analysis of [co]Variant Association (LAVA), as well as effect size correlations between the different age intervals. Second, we utilize a modified version of the genotype/phenotype simulation program HAPNEST to estimate the power to pick up gene-age interactions using single-nucleotide polymorphism (SNP)-by-age interaction (GxA) GWAS methods based on models by Werme et al. (2021) and Zhong et al. (2022). Third, we run a GxA GWAS on BMI in UKB participants. The results of the GxA GWAS are subsequently used to compute age-by-SNP interaction-based polygenic risk scores (PRSs) and age-dependent genetic correlations. Overall, results of the age-stratified GWAS were very homogenous with high effect size correlations for SNPs that were significant in at least one of the age intervals (͞r = .837), high global genetic correlations (͞r = .963) and moderate local genetic correlations (͞r = .492). Simulations are currently running. Using global genetic methods like LDSC did not reveal age-varying genetic effects for BMI in the UKB. Therefore, we propose adopting a SNP-level approach to model and detect varying genetic effects over age, allowing us to gain deeper insights into the complex interplay between genetics and age.