Conventional cardiovascular risk measures rely on age, blood pressure, smoking and similar variables, which often miss early biological change. Polygenic risk scores capture inherited predisposition but not current physiological state. Researchers in the Department of Pharmacology and Pharmacy in the LKS Faculty of Medicine at HKU, with the Musketeers Foundation Institute of Data Science, built a tool intended to reflect what is happening in a patient's body now.
CardiOmicScore draws on a single blood test to forecast risk across six conditions: coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease and venous thromboembolism. It was developed using UK Biobank data, analysing 2,920 circulating proteins and 168 metabolites, and combines genomic, metabolomic and proteomic information through deep learning rather than treating each layer separately.
The reported performance is the ability to flag elevated risk up to fifteen years before clinical onset, particularly in high-risk groups. Paired with basic clinical information such as age and sex, it improved prediction accuracy over conventional polygenic risk scores. The work was published in Nature Communications. Cardiovascular disease caused around 19.8 million deaths worldwide in 2022, so the potential population impact of earlier identification is considerable.
Development in a single large cohort is the obvious limitation, and UK Biobank participants are not representative of global populations in ancestry or in health behaviour. Validation across diverse cohorts will determine whether the multi-omic signal generalises. The conceptual advance, measuring present biological state rather than fixed predisposition, is nonetheless a meaningful departure from how cardiovascular risk has usually been estimated.