Polygenic scores, a measure of genome-wide allelic contribution for a trait, have gained attention in the medical research community in recent years and have led to polarised opinions in terms of their clinical importance. A growth in the number and size of genome-wide association studies, enabled by the assembly of large consortia of case-control and cohort studies, and the advent of national biobanks, has led to the discovery of millions of DNA sequence variants associated with thousands of continuous traits of biomedical relevance (e.g. blood pressure) and disease endpoints (e.g. coronary artery disease). This has contributed to the development of thousands of polygenic scores and a heightened interest in their use in disease prediction and screening.

This thesis evaluates the clinical utility of polygenic scores mainly in the context of cardiovascular disease prediction and screening. The poor performance of polygenic scores in disease prediction is first demonstrated by analysing the Polygenic Score Catalog that aggregates many published polygenic scores for various disease endpoints. The incremental predictive utility of polygenic scores to currently used cardiovascular risk prediction tools in the UK, based on non-genetic risk factors (e.g. QScores) is then evaluated for various cardiovascular disease endpoints using the appropriate metrics. The thesis also explores the potential application of polygenic scores for the discovery of individuals more likely to carry rare genetic variants, using the example of familial hypercholesterolaemia (FH), the most common monogenic disease, which is still currently highly underdiagnosed worldwide. This section begins by modelling a two-stage population screen for the systematic identification of FH cases in the general adult population, followed by an evaluation of the improvement in FH case detection by the inclusion of environmental predictors and a polygenic score for low-density lipoprotein cholesterol.

In conclusion, this thesis puts into perspective the incremental utility of polygenic scores in cardiovascular disease prediction, questioning the claims made on the performance of polygenic scores in prediction. The thesis also explores a potential new avenue of their utility as a tool for aiding with rare variant discovery.