• Queen Mary University of London
  • Barts Health NHS
  • Bradford NHS
  • Manchester Uni

S00027: Measuring the performance of published and machine learning-based polygenic scores

Investigator: Prof Mike Inouye

Insitution: University of Cambridge

Please provide information on the aims of the proposed research including the research question(s) that you are aiming to answer and the health condition(s) under investigation

Many diseases are preventable if they are detected early, which is easier if doctors know who is most at risk. For diseases like coronary artery disease, a common type of heart disease caused by narrow arteries, risk predictions are usually made based on risk factors like age, sex, family history, high cholesterol, weight and smoking habits. However, there is also a strong genetic component to the chance a person will develop coronary artery disease. The same is true for other diseases like diabetes, obesity and stroke.

Researchers have developed a new method to predict a person’s genetic risk for different diseases, called ‘Polygenic Scores’ (PGS). These PGS work by adding up the risk of a disease that come from small changes in lots of different genes. Predictions from these PGS can be combined with traditional risk factors like age and weight, to make a tool that predicts a person’s risk of developing a disease. Current PGS are often developed using genetic information from people of European heritage. Because there are genetic differences between people of different heritages, individuals whose ancestors come from Africa or Asia might not benefit from using existing PGS.

We aim to test how well PGS predict disease in South Asian people and develop new PGS that are better at predicting the risk of diseases in these communities.


How will your research improve health?

By using information from Genes & Health we will be able to identify and develop risk prediction tools that work for people of South Asian heritage. Tools that accurately predict the risk of diseases allows doctors to detect and treat diseases earlier.


Please give a non-technical description of how the research will be undertaken

We will use genetic and health information from Genes & Health to test how well a catalogue of published PGS predict the health of Genes & Health volunteers across a range of disease areas like coronary artery disease, diabetes and stroke. We will also use advanced machine learning techniques to develop PGS that work for people of different ancestries.


How does your research meet the other purposes of Genes & Health?                  

This will help us better understand the genetic basis for disease in South Asian individuals.