Leveraging available data to enable prevention and early detection of non-communicable diseases in under-represented populations: novel risk prediction methods and image-based risk visualisation approaches

Risk stratification underpins effective screening and prevention of non-communicable diseases (NCDs). However, as current research has largely focused on Europeans where large-scale genetic and prospective studies are available, risk assessment is less accurate in non-Europeans, where the genetic architecture and risk factor distribution differ significantly. Additionally, studies have shown that providing risk score alone does not motivate behaviour change, so a new approach is needed. Therefore, I propose to develop and apply advanced statistical techniques (1) to harness the value of available, less powerful datasets to build NCDs risk prediction models for under-represented populations; and (2) to extract disease risk-associated medical imaging features to enable visualisation of risk. Starting with breast cancer in diverse Asian populations, my aims are to develop and apply (1) trans-ancestry methods for improving genetic risk profiling, (2) methods for bias adjustment in risk factors estimated from non-prospective studies, (3) methods to quantify information on mammograms that are important for risk prediction in Asians, and (4) methods to determine which mammogram features differ between women at different level of disease risk to enable risk visualisation. Together, these approaches take a pragmatic approach of harnessing available data in under-represented populations to enable equitable access to precision medicine.