Polygenic risk scores (PRS) combine the estimated effects of multiple genetic variants to quantify individual’s susceptibility to a trait or disease. PRS can also have other applications, as demonstrated in this thesis. In the first study, PRS is used to investigate shared genetic aetiology between epidemiologically associated T2D and pancreatic, colorectal, prostate and breast cancers. However, PRS does not reveal the mechanisms underlying the association. To link the associations to mechanisms, PRSs are partitioned based on patterns of genetic effects. The results of the study suggest that genetic predisposition to T2D does not explain the increased risk of breast, colorectal and pancreatic cancers, but partially explains the decreased risk of prostate cancer. Furthermore, the results add to the body of evidence that the link between T2D and some types of cancer is driven by metabolic features, such as higher adiposity and insulin resistance.

In the second study, methods partitioning genetic associations into independent genetic components are assessed on simulated data. Such methods can be used to construct partitioned-PRSs that could be used to better understand shared mechanisms. Simulation study shows that proposed method is able to estimate latent components better than the existing method, laying ground for application to real world data and further research.

In the third study, the interplay between genetic predisposition, captured by PRS, and long-term exposure to air pollution in cardiometabolic disease is investigated. The results show that long-term exposure to ambient air pollutants is associated with a small increase in the risks of T2D and CHD, and that that BMI PRS modifies the effects of air pollutants on BMI, suggesting increased susceptibility in individuals with genetic predisposition to higher BMI.

Lastly, in the fourth study, post-GWAS analyses of signals associated with random glucose levels are conducted to identify novel loci, demonstrate multi-phenotype associations and heterogeneity of effects, and highlight the mechanisms through which genetic variation affects random glucose levels using transcriptome-wide association study.

This work highlights the importance of considering cross-phenotype effects and investigating genetic associations to understand the underlying mechanisms.