MRC DTP in Precision Medicine

Synergising ‘worldwide’ health data for personalised risk predictions

Supervisors

Professor Honghan Wu, School of Health and Wellbeing, University of Glasgow 

Professor David McAllister, School of Health and Wellbeing, University of Glasgow 

 

Summary:

This project aims to improve personalized risk prediction for cardiovascular diseases (CVDs) using machine learning. Existing CVD risk prediction models often have limitations due to the representativeness of their training data. 

To address this, the project proposes a novel approach that combines published models with their corresponding derivation cohort descriptions. This allows for ‘proxied access’ to a more diverse and global dataset. 

The key steps involve: 

1. Systematic review: Extracting information from a large corpus of CVD risk prediction models. 

1. Synthetic data generation: Creating realistic health records that match the characteristics of the derivation cohorts and model performances. 

1. Personalized model derivation: Developing individualized risk prediction models using the synthetic data. 

1. Evaluation: Assessing the performance of these models using real-world health records. 

This approach aims to provide more accurate and personalized risk predictions for individuals with CVDs.