AI blood test model flags cardiovascular risk years before symptoms
University of Hong Kong researchers say CardiOmicScore uses proteins and metabolites to estimate risk across six cardiovascular diseases.
By Lucas Ferreira · Science & Environment Writer
3 min read
Researchers at the University of Hong Kong have developed an artificial intelligence tool that uses a single blood test to estimate long-term risk for major cardiovascular diseases. The team says the approach could help identify people at elevated risk years before symptoms appear, giving doctors and patients more time to consider prevention.
The tool, called CardiOmicScore, was described in a study published in Nature Communications, according to the University of Hong Kong. HKUMed researchers said the model assessed risk for six conditions: coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease and venous thromboembolism.
According to the university, CardiOmicScore was able to detect elevated risk signals as much as 15 years before clinical onset among people identified as high risk. The system was built using large-scale data from the UK Biobank, the university said.
The research team used deep learning to combine several types of biological data, an approach known as multiomics. According to HKUMed, the model analyzed 2,920 circulating proteins and 168 metabolites measured in blood samples, along with genomic information.
Proteins carry out many functions in the body, while metabolites are small molecules linked to processes such as energy use, diet and disease response. The university said those blood-based signals can offer a current biological snapshot that reflects immune activity, metabolism and vascular health.
That differs from genetic risk scores, which estimate inherited vulnerability based on a person’s DNA. HKUMed said genetic scores remain largely fixed across life, while proteins and metabolites can shift with aging, illness, lifestyle and environmental exposure.
Professor Zhang Qingpeng, associate professor in HKUMed’s Department of Pharmacology and Pharmacy, said genes define a person’s starting risk, while proteins and metabolites reflect current physical health. He said the AI system is intended to interpret complex molecular patterns so risks can be recognized earlier and addressed through prevention.
The university said the model outperformed conventional polygenic risk scores in the study. Its accuracy improved further when researchers added clinical information such as age and gender, according to HKUMed.
The six conditions covered by CardiOmicScore include several common and serious cardiovascular problems. Atrial fibrillation is an irregular heartbeat associated with stroke risk, peripheral artery disease involves reduced blood flow to the limbs, and venous thromboembolism refers to blood clots that form in veins and can travel to the lungs, according to the university’s summary.
Cardiovascular disease remains the world’s leading cause of death, accounting for about 19.8 million deaths in 2022, according to figures cited by the University of Hong Kong. Doctors typically estimate risk using factors such as age, blood pressure and smoking history, but the university said those measures may miss early biological changes before disease becomes visible.
The study was led by Zhang and researchers at HKUMed and the HKU Musketeers Foundation Institute of Data Science. The first author was Luo Yan from the HKU Institute of Data Science, according to the university.
Zhang said the team aims to use technology to identify and prevent disease before it develops. According to HKUMed, the broader goal is to support earlier intervention through lifestyle changes, closer monitoring or other preventive care when risk appears high.
This story draws on original reporting from ScienceDaily.