AI Tool CardiOmicScore Predicts Cardiovascular Disease Risk Up to 15 Years Early
Researchers at the LKS Faculty of Medicine of the University of Hong Kong (HKUMed) have introduced an innovative artificial intelligence tool named CardiOmicScore, designed to predict serious cardiovascular problems well in advance of any symptoms. This tool utilizes data from a single blood test to estimate the future risk of six major cardiovascular diseases (CVDs), including coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism. Remarkably, the model is capable of detecting warning signals as far as 15 years prior to the clinical onset of these conditions.
The findings of this research were published in the journal Nature Communications, highlighting the urgent need for improved methods to assess cardiovascular risk. Cardiovascular diseases are the leading cause of death globally, with approximately 19.8 million fatalities reported in 2022 alone. Traditionally, doctors evaluate cardiovascular risk using factors such as age, blood pressure, and smoking history. While these indicators are helpful, they may not capture the earliest biological changes that occur within the body before a disease becomes apparent.
Genetic risk tests, such as polygenic risk scores, offer another means of estimating disease likelihood. However, these scores are largely fixed at birth and do not account for dynamic changes caused by lifestyle factors, aging, or environmental exposures. CardiOmicScore aims to provide a more current and comprehensive picture of an individual's health by analyzing proteins and metabolites that reflect immediate biological conditions.
The HKUMed team employed deep learning techniques to develop this tool, integrating various biological data layers in a process known as multiomics. This approach combines information from genomics, metabolomics, and proteomics. Genomics focuses on genetic information, proteomics examines proteins, and metabolomics studies small molecules called metabolites that are produced as the body processes food and responds to diseases. By analyzing large-scale population data from the UK Biobank, the researchers examined 2,920 circulating proteins and 168 metabolites in blood samples, creating a detailed snapshot of an individual's biological state.
Professor Zhang Qingpeng, an Associate Professor at HKUMed, explained that while genes set a baseline health risk, proteins and metabolites provide insights into current physical health. The AI tool is designed to decode these complex molecular signals, enabling earlier identification of risks, which could lead to timely lifestyle modifications and preventive measures.
The results indicated that CardiOmicScore could generate personalized cardiovascular risk estimates from complex molecular measurements, outperforming conventional polygenic risk scores. Its accuracy improved further when clinical information such as age and gender were included. The model is specifically designed to assess risks for coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism.
The research reflects a significant shift toward precision medicine, moving from traditional genetic assessments to more dynamic evaluations of health risks. In the future, a small blood sample could be used to create a comprehensive risk profile for multiple cardiovascular diseases, providing patients and healthcare providers with valuable time to implement lifestyle changes or preventive strategies. Professor Zhang emphasized the goal of leveraging technology to identify and prevent diseases before they develop, aiming to shift health management from reactive treatment to proactive prediction and intervention, thereby creating a lasting impact on public health and individual patient care.




