Age-varying effects in genetic epidemiological studies

Abstract

There is growing evidence to suggest that genetic effects on some phenotypes vary with age. Estimating this fluctuation in genetic effects with age is crucial for accurately mapping the genetic architecture of complex traits and diseases. This thesis aims to develop and evaluate methods for combining summary-level data from Genome-wide Association Studies (GWAS) to estimate both main genetic effects (i.e., SNP-phenotype associations) and age-varying genetic effects (i.e., SNP by age interactions).

Fixed-effects meta-analysis is a commonly used method to combine GWAS summary-level data across multiple contributing studies, assuming a common underlying genetic effect. In Chapter 3, I extend meta-regression as a method to
estimate both the main and age-varying genetic effects using GWAS summary-level data. I use simulations to assess its performance and compare it to fixed-effect and random-effects meta-analysis when genetic effects vary with age. I demonstrate the use of meta-regression in an empirical example, estimating the age-varying association between the genetic variant rs9939609 (at the FTO locus) and body mass index (BMI) during childhood using publicly available GWAS summary-level data.

In Chapter 4, I investigate how GWAS summary-level data from studies that provide estimates only on main genetic effects can be combined with summary-level data from studies that provide estimates on age-varying genetic effects to estimate main and age-varying genetic effects. I modify an existing method for meta-analysis of non-linear effects and assess its performance. Additionally, I assess the performance of random-effects meta-analysis and multivariate meta-analysis in combining these types of GWAS summary-level data. In Chapter 5, I apply the developed methods to estimate the age-varying genetic effects of variants rs9939609 (at the FTO locus) and rs9436303 (at the LEPR locus) on BMI during childhood.

Mendelian randomization (MR) is a method that uses genetic data to estimate the causal effect of an exposure on an outcome of interest. In Chapter 6, I show how estimating age-varying genetic effects enhances the interpretation of MR studies when the exposure-outcome relationship of interest varies over time. This discussion is based on an empirical example investigating the causal effect of childhood BMI (using rs9939609 and rs9436303 as instruments) on adult diastolic
blood pressure.

Overall, this thesis contributes to the field of genetic epidemiology by proposing and validating methods to combine GWAS summary-level data to estimate age-varying genetic effects, with important implications for the interpretation of MR studies and our understanding of the genetic architecture of complex traits.

Date of Award	1 Oct 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	Debbie A Lawlor (Supervisor), Evangelia Stergiakouli (Supervisor), Kate M Tilling (Supervisor) & Nicole M Warrington (Supervisor)