A large part of older population have multiple conditions such as cardiovascular disease and Diabetes, and all of them have a whole variety of exposures that may interact in various ways. The biological mechanisms underlying this are highly complex, even with the vast amounts of data currently available, we have limited approaches to comprehending it in its totality. This project aims to establish novel methodologies for understanding and discovering comorbidity and genetic risks using causal networks and a variety of statistical techniques, utilising data on stroke, hypertension and diabetes, to prove these methods. This may lead to new insights into disease mechanisms. We will start by looking into polygenic risk scores (PRSs), looking into how they are constructed and methods to determine suitable numbers of genetic variants (in particular, SNPs) to include, and methods for removing highly correlated SNPs. We will then look at how using PRSs in Mendelian randomisation (MR) compares to individual SNPs. Developing the theories of causal networks alongside their use in Mendelian randomisation will form the bulk of the project. This will align nicely with the application of machine learning algorithms to causal network analysis. These techniques will be applied to stroke, hypertension and diabetes using the UK Biobank. The predominant outcome will be methodological, focused around our two focuses of PRSs and causal networks (based on MR and Bayesian networks), and how they apply to research into comorbidities. This is a technique that can be applied to many diseases, especially with the current expansion in genome-wide association data, and it provides an ample movement forward and application in many fields. In our application for stroke, hypertension and diabetes, we will hopefully be able to develop more accurate diagnostic and prognostic tools, discover new causal pathways, leading to new interventions.
