Stroke remains a leading cause of death and disability worldwide, yet current risk prediction tools often rely solely on clinical and lifestyle factors and do not fully capture genetic susceptibility. This project aims to develop and evaluate statistical and machine learning models to predict first-time stroke risk by integrating genetic (polygenic risk scores), clinical, and lifestyle data available within UK Biobank. Research questions: To what extent does incorporating genetic information improve prediction accuracy for first-time stroke compared with models based on traditional risk factors alone? How do genetic, clinical, and lifestyle factors interact to influence stroke risk? Can integrated risk models better identify high-risk individuals for targeted prevention?
Objectives: Construct and validate predictive models (statistical and ML-based) for first-time stroke using UK Biobank data. Quantify the contribution of polygenic risk scores beyond established clinical and behavioural risk factors. Explore potential interactions and nonlinear effects among genetic, clinical, and lifestyle variables. Evaluate model performance, calibration, and interpretability to inform potential clinical translation.
Scientific rationale: Existing stroke risk models such as Framingham scores are limited in predictive power and generalisability. Leveraging the depth of UK Biobank data-including genotyping, clinical measures, and lifestyle information-provides a unique opportunity to develop more accurate, data-driven models. Improved prediction of first-time stroke risk could facilitate personalised prevention strategies and contribute to precision public health initiatives.
