Investigating the genetic interplay between adult height and cardio-metabolic traits

Abstract

Adult height is an easily and accurately measured anthropometric trait, that is highly heritable and genetically complex. The recent GWAS meta-analysis for height performed by the Genetic Investigation of ANthropometric Traits (GIANT) consortium, using up to 5.4 million cross-ancestry individuals, explained 40% of the phenotypic variability in Europeans and up to 20% in other ancestries. Several studies have reported associations of genetically predicted height with a plethora of other traits including disease. Firstly, this thesis aimed to perform a deeper investigation of the role of genetically predicted height with health-related outcomes in diverse ancestries. To achieve this, a phenome-wide association study (PheWAS) in individuals of diverse ancestry from six study populations was performed by myself and other analysts from the participating cohorts, using my cross-ancestry polygenic score for height, which is based on the latest GIANT GWAS meta-analysis. Afterwards, I meta-analysed the PheWAS results from the six cohorts to gain higher statistical power, leading to discoveries of conditions not previously associated with height, such as hyperpotassaemia and autism. Based on the PheWAS meta-analysis, I decided to further explore the causal role and pleiotropic effects of adult height on the risk of hypothyroidism and thyroid cancer, through Mendelian Randomisation (MR) analysis. The reverse association of hypothyroidism and attained height were also explored. Besides the MR investigations, I expanded my research in the area of disease prediction, taking advantage of the rapid development of computational tools and the utilisation of large-scale datasets. To this end, I employed machine and deep learning approaches for the prediction of atrial fibrillation (AF), and secondly for the prediction of ischemic stroke occurrence in AF patients, using UK Biobank’s prospective health, genomic and clinical data. Furthermore, the feature importance ranking and contribution to the prediction of the disease were explored.