Researchers at the University of Hong Kong have built an artificial intelligence tool that reads a single blood sample and estimates a person's risk of six major cardiovascular diseases up to 15 years before symptoms appear.
The tool, called CardiOmicScore, was described in Nature Communications in 2026 and screens for coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism.
The system uses deep learning to scan 2,920 circulating proteins and 168 metabolites in the blood.
In analyses of UK Biobank data, the model detected warning signals in high-risk individuals as far as 15 years before clinical onset, and performed substantially better than polygenic risk scores, which estimate risk based only on inherited genetic variants.
Also read: Microplastics are turning up in human brain tissue, and no one yet knows what they do
Why a 15-year window matters
Cardiovascular disease remains the country's largest killer. Heart disease alone accounted for 683,491 deaths in 2024, keeping it the leading cause of death in the United States, according to CDC mortality data. Stroke has moved up to fourth, and together the two conditions kill more Americans than cancer and accidents combined.
"The fact remains that heart disease and stroke continue to take the lives of too many of our loved ones," said Dr. Stacey E. Rosen, volunteer president of the American Heart Association, in a January 2026 statement on the group's latest statistics update.
A longer warning window, in theory, gives a doctor and patient more time to act on things that actually change risk, blood pressure control, cholesterol treatment, weight, sleep, exercise, smoking, and diabetes management.
Standard tools today, cholesterol panels, blood pressure readings, and family history questionnaires, often flag risk once metabolic damage is already underway.
Also read: Nearly half of dementia cases could be prevented, and most risk factors are in your control
How the tool reads the blood
Existing genetic risk scores work off DNA a person is born with. CardiOmicScore instead reads proteins and metabolites, molecules that shift with lifestyle, environment, and current disease processes, and combines those readings with basic clinical inputs such as age and sex.
Professor Zhang Qingpeng, associate professor in the Department of Pharmacology and Pharmacy at the LKS Faculty of Medicine of the University of Hong Kong, has framed the distinction as genetic information setting a person's baseline risk while proteins and metabolites act as a more immediate readout of current health, according to the HKU research team's summary.
Genetic risk is largely fixed at birth; protein and metabolite patterns move over time and can catch trouble that DNA alone cannot see.
Not yet a bedside test
CardiOmicScore is a research tool, not a product a primary care doctor can order.
Also read: One bedtime drink added 84 minutes of sleep a night in early research
The published study relied on UK Biobank participants, a population that is largely of European ancestry, and the model's performance still needs to be tested in more diverse groups and prospective clinical trials before it could enter routine care.
There is also no shortcut around the standard advice that anyone with chest pain, breathlessness, palpitations, or a strong family history of heart disease should talk to a doctor now, rather than wait for a blood-based algorithm to reach the clinic.
What the Hong Kong work suggests is where prevention could be heading, from static, one-shot risk calculators to dynamic readings that update as a person's biology changes.
Also read: Adding just half a tablespoon of olive oil a day may protect your heart and brain
Also read: Your brain may be warning you about heart disease 3 to 8 years before it strikes