A Blood Test That Sees 15 Years Ahead
Most heart disease warnings show up late. By the time blood pressure spikes or a scan flags something, the damage is often already underway. A new tool out of Hong Kong is trying to move that timeline back by more than a decade.
A research team from the Department of Pharmacology and Pharmacy at the LKS Faculty of Medicine of the University of Hong Kong has developed an AI-based cardiovascular risk prediction tool called CardiOmicScore, which uses a single blood test to forecast the future risk of six major cardiovascular diseases: coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism. The findings were published in Nature Communications.
Why genes alone weren’t enough
Cardiovascular diseases remain the leading cause of death worldwide, responsible for roughly 19.8 million deaths in 2022. Doctors typically assess risk using age, blood pressure, and smoking history, but those measures often miss the subtle biological shifts that happen before a disease becomes clinically obvious. Genetic risk scores have helped fill part of that gap, but they come with a limit. Genetic predisposition is largely fixed at birth and doesn’t change over time, so it can’t reflect what’s actually happening in someone’s body as their lifestyle or environment shifts.
What the AI is actually reading
To build CardiOmicScore, the HKUMed team used deep learning to combine multiomics data, including genomics, metabolomics, and proteomics, drawing on large scale population data from the UK Biobank and analyzing 2,920 circulating proteins and 168 metabolites from blood samples. Those molecular signals work like a running log of the body. They sensitively reflect subtle changes in the immune system, metabolism, and vascular health, offering a far more current picture than a genetic score frozen since birth.
Professor Zhang Qingpeng, who led the study, put the distinction simply: genes decide where a person starts, but proteins and metabolites reflect present-day physical health. Decoding those signals, he said, gives doctors and patients a much earlier chance to catch risk and act on it through lifestyle changes and early prevention.
The 15-year window
When combined with basic clinical information like age and gender, the model significantly improved prediction accuracy for all six conditions and could flag elevated risk up to 15 years before symptoms appeared. That’s the headline number, and it’s a meaningful one. A test that can point to risk a decade and a half out gives people time to actually change course, not just react once something’s already gone wrong.
The researchers frame this as part of a broader shift in precision medicine, moving from a static, gene-centric model toward a more dynamic, multiomics-based approach where a single blood sample could eventually generate a full cardiovascular risk profile. It’s still research, not a product on a clinic menu yet. But the direction is clear: earlier warnings, built from what’s happening in the body right now.
