The contribution of non-additive and longitudinal genetic effects to health and cognitive traits

Abstract

This thesis aims to investigate two important yet understudied aspects of human trait genetic architectures: non-additive genetic effects and longitudinal genetic effects, with a focus on health and cognitive traits. Most genetic studies often focus on additive genetic effects estimated in unrelated individuals at single time points. In the first project, I developed a method to address confounding in studying non-additive genetic effects and applied this method to study health traits. In the second project, I characterised the trajectory of genetic influences on cognitive performance from early childhood to adolescence. The first project investigates the effect of elevated genome-wide autozygosity (genomic regions identical by descent) on disease risk. Initially, using data from the Genes & Health cohort and UK Biobank, I developed a method to control for social and environmental confounding in autozygosity-trait associations by including indicators of parental relatedness in regression models. Using this approach, I found associations between autozygosity and increased risk for multiple poor health outcomes, including infectious, respiratory, and congenital disorders. When granted access to the 23andMe cohort of over 8 million individuals, I determined that this approach was insufficient for very large cohorts and refined the methodology by restricting analyses to highly consanguineous individuals. Using this method, I identified significant associations between elevated autozygosity and risk of multiple disorders, including type 2 diabetes and post-traumatic stress disorder. I then leveraged the large number of siblings in the 23andMe dataset to replicate these findings in a within-sibship analysis. Population attributable risk calculations demonstrated that autozygosity resulting from consanguinity accounts for 8-15% of type 2 diabetes cases among British Pakistanis. Simulation studies indicated these associations likely stem from non-additive genetic architectures (e.g., recessive effects captured by autozygosity) rather than increased additive genetic variance. In the second project, I investigated how common and rare genetic variants differentially influence cognitive performance across childhood and adolescence, using data primarily from the Avon Longitudinal Study of Parents and Children. By analysing both common variant polygenic indices and rare exonic variant burden, I demonstrated that these genetic factors have distinct patterns of contribution across development: the effect of common variants increased with age, while the effect of rare, damaging variants attenuated. Using trio analyses, I showed these patterns reflect direct genetic effects rather than parental genetic nurture or other sources of confounding. Furthermore, these genetic effects differed across the cognitive ability distribution, with rare variants having displayed stronger effects on the lower tail that diminished with age, while common variants had increasing effects primarily in the upper half of the distribution. This work highlights the importance of methodological innovation in addressing confounding in genetic studies, the widespread nature of non-additive genetic effects across the phenotypic spectrum, and the dynamic relationship between genetic variation and cognitive development.