AI Blood Test Spots Heart Disease Risk 15 Years Before Symptoms
Cardiovascular disease remains the leading cause of death worldwide, claiming approximately 19.8 million lives in 2022 alone. Standard risk assessments, age, blood pressure, smoking history, cholesterol, catch many at-risk patients, but they can miss the earliest biological changes that develop years before a heart attack or stroke.
Now, researchers at the University of Hong Kong have developed an artificial intelligence tool called CardiOmicScore that can predict an individual’s risk of six major cardiovascular diseases up to 15 years before symptoms appear, using nothing more than a single blood sample.
The tool, described in a paper published in Nature Communications, combines three layers of molecular data, genomics, proteomics, and metabolomics, into a personalized risk score that the researchers say substantially outperforms conventional polygenic risk scores.
Beyond Static Genetics
Traditional genetic risk scores have a fundamental limitation: they are fixed at birth. A person’s DNA does not change, but their current health status does. Two people with identical genetic risk profiles can have dramatically different cardiovascular futures depending on their lifestyle, environment, and metabolic health.
CardiOmicScore addresses this by measuring what is happening in the body right now. From a single blood draw, the AI analyzes 2,920 proteins and 168 metabolites, the molecules that reflect immune activity, metabolism, and vascular health in real time. It then integrates those measurements with genomic data using a deep learning model to produce a dynamic, personalized risk estimate.
“Genes determine where we start, they define our baseline health risk,” said Professor Zhang Qingpeng, associate professor in the Department of Pharmacology and Pharmacy at HKUMed and lead researcher on the project. “However, proteins and metabolites reflect our current physical health. Our AI tool is designed to decode these complex molecular signals, enabling doctors and patients to identify risks much earlier.”
Six Diseases, One Blood Test
The tool covers the full spectrum of major cardiovascular conditions: coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism. For each condition, CardiOmicScore generates a risk estimate that, according to the researchers, becomes even more accurate when combined with basic clinical information such as age and gender.
The team trained and validated the model using data from the UK Biobank, a large-scale biomedical database containing genetic, proteomic, and metabolomic data from hundreds of thousands of participants. The ability to predict risk up to 15 years in advance means the tool could identify individuals in an early, pre-symptomatic window when lifestyle modifications and preventive treatments have the greatest impact.
From Reactive to Proactive
The shift from waiting for symptoms to predicting them is a central goal of precision medicine, but it has remained largely aspirational for common diseases. Most risk prediction tools improve only incrementally over traditional methods. CardiOmicScore’s multiomics approach represents a different strategy: instead of asking whether a single biomarker can predict disease, it asks whether the entire molecular state of a person’s body can reveal its trajectory.
“We aim to leverage technology to identify and prevent diseases before they develop,” Professor Zhang said. “By shifting health management from reactive treatment to proactive prediction and intervention, we aim to create a lasting impact for both public health and individual patient care.”
The researchers caution that the tool requires validation in more diverse populations before it can be deployed clinically. The UK Biobank, while large, is predominantly composed of participants of European ancestry, and the model may not perform equally well in other ethnic groups. Broader validation studies, including in Asian populations, are already underway.
The first author of the study is Luo Yan of the HKU Musketeers Foundation Institute of Data Science. Co-authors include researchers from HKU, the University of Liverpool, and the Karolinska Institute.
Sources
[1] University of Hong Kong. “New AI blood test predicts heart disease 15 years early.” ScienceDaily, 19 July 2026. https://www.sciencedaily.com/releases/2026/07/260716023603.htm
[2] Luo, Y., Zhang, N., Yang, J., et al. “CardiOmicScore: Multiomics deep learning for cardiovascular risk prediction.” Nature Communications, 2026. DOI: 10.1038/s41467-026-68956-6