University of Hong Kong researchers developed CardiOmicScore, an AI-based system that analyzes thousands of proteins and metabolites in a blood sample to estimate a person’s future risk of six cardiovascular diseases potentially up to 15 years before symptoms appear.
Editorial Note
This article provides independent health and science reporting for educational purposes. It does not provide medical advice, diagnose cardiovascular disease, or recommend that readers request a particular test, medication, or treatment.
New To Education is an independent publication. It is not affiliated with, sponsored by, endorsed by, or acting on behalf of the University of Hong Kong, HKUMed, UK Biobank, Nature Communications, CardiOmicScore’s developers, healthcare providers, technology companies, or other organizations discussed in this article.
CardiOmicScore remains a research tool rather than a routine clinical blood test available through ordinary medical appointments. Its ability to predict risk in research data does not prove that using it in everyday healthcare will prevent heart attacks, strokes, or deaths. Additional validation, cost analysis, regulatory review, and clinical trials may be necessary before widespread adoption.
A Blood Sample May Reveal Cardiovascular Risk Years Before Symptoms
Researchers at the University of Hong Kong have developed an artificial intelligence system that may identify elevated cardiovascular risk as much as 15 years before symptoms appear.
The system, called CardiOmicScore, analyzes thousands of proteins and metabolites measured in a blood sample. It then combines those molecular signals with artificial intelligence to estimate a person’s future risk of six major cardiovascular conditions.
Those conditions include coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism.
The research was originally announced by the University of Hong Kong in March 2026 after publication in Nature Communications. It received renewed international health-news attention on July 19, 2026. The date distinction matters because July 19 marked wider coverage of the study rather than the original publication of the scientific findings.
The researchers believe the approach could eventually help doctors move beyond reacting to established disease and instead identify biological warning signs while prevention may still be possible.
What CardiOmicScore Actually Measures
Traditional cardiovascular-risk assessments usually rely on information such as age, sex, blood pressure, cholesterol, smoking status, diabetes, family history, and existing medical conditions.
These factors remain important, but they may not capture every early biological change taking place inside the body.
CardiOmicScore was designed to examine a much deeper layer of information. Researchers used blood measurements covering 2,920 circulating proteins and 168 metabolites.
Proteins perform countless functions throughout the body, including regulating inflammation, immunity, blood clotting, tissue repair, and communication between cells.
Metabolites are smaller molecules produced as the body processes food, generates energy, responds to stress, and carries out other biological activities.
Together, these measurements can provide a detailed picture of a person’s current biological condition. Subtle patterns in the blood may indicate changes in metabolism, immune activity, circulation, or vascular health long before a person experiences chest pain, shortness of breath, weakness, or another recognizable symptom.
How Artificial Intelligence Helps Interpret the Blood
The challenge is not simply measuring thousands of molecules. It is understanding what their combined patterns mean.
A doctor could not realistically examine more than 3,000 molecular measurements individually and determine how they interact to influence several different diseases over many years.
CardiOmicScore uses deep-learning methods to identify patterns across those large datasets. The system integrates information from several biological fields, an approach commonly known as multiomics.
Genomics examines genetic information. Proteomics studies proteins. Metabolomics analyzes metabolites and the chemical processes occurring throughout the body.
By combining these layers, the model attempts to detect relationships that may be too complicated for conventional risk calculations.
The AI does not physically examine the patient or make an independent medical diagnosis. It converts molecular data into a personalized estimate of future cardiovascular risk.
The Test Assesses Six Different Cardiovascular Conditions
One notable feature of CardiOmicScore is that it was designed to estimate risk across several diseases rather than focusing on only heart attacks.
Coronary artery disease develops when the vessels supplying blood to the heart become narrowed or blocked. It can contribute to chest pain and heart attacks.
Stroke occurs when blood flow to part of the brain is interrupted or when a blood vessel in the brain ruptures.
Heart failure develops when the heart cannot pump blood effectively enough to meet the body’s needs.
Atrial fibrillation is an irregular heart rhythm that can increase the likelihood of stroke, heart failure, and other complications.
Peripheral artery disease occurs when narrowed arteries reduce blood circulation, commonly affecting the legs.
Venous thromboembolism includes blood clots that form in veins, such as deep-vein thrombosis, and clots that travel to the lungs, known as pulmonary embolism.
These conditions differ in their causes, symptoms, and treatment. A system capable of estimating risk across all six could potentially provide a broader cardiovascular profile from one blood sample.
Why the 15-Year Warning Is Important
Cardiovascular disease often develops quietly.
A person may feel healthy while high blood pressure, inflammation, abnormal cholesterol, damaged blood vessels, irregular heart rhythms, or metabolic changes gradually increase risk.
By the time symptoms become noticeable, the underlying disease may already be advanced.
A warning 10 or 15 years earlier could theoretically give patients and clinicians more time to address modifiable risk factors. That might include controlling blood pressure, improving cholesterol levels, stopping smoking, increasing physical activity, managing diabetes, improving sleep, or using preventive medication when medically appropriate.
The phrase “15 years early” does not mean the model can predict the exact date when someone will experience a heart attack or stroke.
It means that researchers found molecular patterns associated with elevated risk among people who later developed cardiovascular disease during a follow-up period extending as far as 15 years.
Risk is not destiny. A person identified as high risk may never develop the predicted condition, while someone classified as lower risk could still become ill.
How This Differs From Genetic Risk Testing
Genetic risk scores estimate how inherited variations may affect a person’s likelihood of developing a disease.
That information can be useful because genetics may help explain why some people face elevated risk even when they appear healthy.
However, inherited DNA remains largely unchanged throughout life. A genetic score does not automatically show how recent changes in diet, physical activity, weight, medication, stress, pollution exposure, infection, or another health condition are affecting the body today.
Proteins and metabolites are more dynamic.
Their levels may change as the body ages or responds to lifestyle and environmental influences. CardiOmicScore therefore aims to provide a biological snapshot that is more responsive to a person’s present condition.
The researchers reported that the multiomics model performed substantially better than conventional polygenic risk scores. Prediction improved further when molecular information was combined with clinical factors such as age and sex.
This does not mean genetic testing has become useless. Genetic and molecular information may eventually complement each other, with inherited risk describing a person’s starting position and blood-based signals reflecting how health is changing over time.
The Research Used UK Biobank Data
The researchers developed and evaluated CardiOmicScore using information from UK Biobank, a major health-research database containing biological, medical, genetic, and lifestyle information from a large population of participants.
Large databases are valuable because they allow researchers to connect blood measurements with health outcomes recorded over many years.
However, performance within a research database does not guarantee identical results in every hospital, country, or population.
UK Biobank participants may differ from the broader public in age, ethnicity, socioeconomic background, health behavior, and willingness to participate in research.
Before CardiOmicScore could be used widely, researchers would need to confirm that it performs reliably across more diverse populations, including people of different racial and ethnic backgrounds, younger adults, older adults, and patients with existing medical conditions.
Independent research teams should also be able to test the model rather than relying only on results produced by its original developers.
This Is Not Yet a Routine Doctor’s-Office Test
The phrase “single blood test” may make CardiOmicScore sound as simple as an ordinary cholesterol panel.
The blood collection itself may eventually be straightforward, but the analysis is considerably more complicated.
Measuring thousands of proteins and metabolites generally requires specialized laboratory equipment, technical expertise, data-processing systems, and quality controls.
Cost will be a major consideration. A test that predicts disease accurately but is too expensive for routine use may remain limited to specialized hospitals, research programs, or higher-risk patients.
Researchers may eventually reduce the number of molecular markers needed. If a smaller group of proteins and metabolites provides nearly the same predictive accuracy, the test could become easier and more affordable to deploy.
For now, CardiOmicScore should be understood as a promising research framework rather than a blood test people can expect to receive during their next annual examination.
Prediction Does Not Automatically Improve Health
An accurate forecast is valuable only when healthcare providers know what to do with it.
A patient may learn that they face elevated risk 15 years before symptoms appear, but the next steps must still be defined.
Doctors would need evidence showing which preventive interventions work best for people identified through the test. Researchers would also need to determine how often testing should be repeated and whether changes in the score reflect genuine improvements or worsening health.
A successful clinical trial would ideally show more than predictive accuracy. It would show that using CardiOmicScore changes medical decisions and leads to fewer heart attacks, strokes, blood clots, hospitalizations, or deaths.
Without that evidence, the test could identify risk without proving that it improves outcomes.
This is a common challenge in medical screening. Detecting more abnormalities is not always the same as helping more patients.
False Alarms Could Create Harm
No risk-prediction model is perfect.
Some people identified as high risk may never develop cardiovascular disease. They could experience unnecessary anxiety, repeated testing, medical appointments, or treatment side effects.
Other people may receive a low-risk result and become falsely reassured. A favorable score should not encourage someone to ignore high blood pressure, smoking, diabetes, chest pain, or other established warning signs.
Researchers and healthcare providers would need to determine acceptable thresholds for deciding when a score warrants intervention.
The balance may differ depending on the disease. Missing a person at high risk of a dangerous blood clot may carry different consequences from incorrectly identifying someone who would never develop the condition.
Medical professionals would also need to explain uncertainty clearly. A probability is not a diagnosis, and a prediction should not be presented as an unavoidable future.
AI Bias Must Be Carefully Examined
AI systems learn from the data used to build them.
When certain populations are underrepresented, predictions may be less accurate for those groups.
A model could perform well overall while producing weaker results for women, particular ethnic communities, people with disabilities, or individuals from regions with different diets and healthcare systems.
Bias may also enter through the biological measurements themselves. Protein and metabolite levels can be affected by medication, pregnancy, diet, infections, exercise, kidney function, liver disease, and other factors.
Clinical adoption would therefore require transparent testing across different groups and health conditions.
Patients and clinicians should be able to understand how reliable the score is for people with similar characteristics rather than being shown only one overall accuracy figure.
Health Data Privacy Will Matter
Multiomics data can reveal highly sensitive information.
A blood sample may provide clues about disease risk, genetic traits, medication use, metabolism, inflammation, and other aspects of health.
Healthcare systems would need strong protections governing where the information is stored, who can access it, how long it is retained, and whether it may be used to train commercial AI systems.
Patients should know whether insurers, employers, technology companies, or research partners can receive any part of the data.
A person may consent to cardiovascular-risk analysis without realizing that the same molecular information could potentially reveal other health conditions.
The benefits of early prediction should not require patients to surrender control over deeply personal biological information.
Early Detection Could Widen Health Inequality
Advanced testing often reaches wealthier patients and well-funded hospitals first.
People with private insurance, specialist access, or the ability to pay directly may benefit before those in rural, low-income, or underserved communities.
That could widen existing cardiovascular inequalities.
Heart disease already disproportionately affects populations facing limited access to preventive care, nutritious food, safe places to exercise, stable housing, medication, and regular health monitoring.
A sophisticated test cannot solve those structural problems.
There is little value in identifying elevated risk years early when patients cannot afford follow-up appointments, blood-pressure medication, healthy food, transportation, or treatment.
The technology’s public-health value will depend partly on whether it becomes accessible and whether healthcare systems can provide meaningful prevention after risk is discovered.
Current Heart-Health Screening Still Matters
People should not wait for CardiOmicScore to become available before thinking about cardiovascular health.
Blood-pressure monitoring, cholesterol testing, diabetes screening, smoking cessation, physical activity, healthy sleep, and medical evaluation of concerning symptoms remain important.
Traditional risk assessments may appear less technologically impressive, but they are supported by decades of clinical evidence and can identify many people who would benefit from prevention.
Anyone experiencing chest pressure, sudden weakness, difficulty speaking, severe shortness of breath, fainting, or other possible emergency symptoms should seek immediate medical care rather than relying on a future-risk test.
CardiOmicScore may eventually add information to existing medical care. It is not intended to replace emergency diagnosis, clinical judgment, or established preventive screening.
Why the Research Still Represents an Important Advance
Despite the limitations, CardiOmicScore demonstrates how artificial intelligence may change preventive medicine.
The system does not rely on AI to invent a diagnosis from nothing. It uses AI to interpret an amount of biological information that would otherwise be extremely difficult to evaluate.
That approach may eventually allow doctors to detect disease processes before they become visible through symptoms or conventional tests.
It also moves precision medicine toward a more dynamic model. Instead of estimating risk largely from fixed genetics and broad population averages, future systems may track how a person’s biology changes over time.
Repeated testing could potentially show whether preventive measures are improving a person’s molecular risk profile.
That possibility remains to be proven, but it points toward healthcare that is more proactive, personalized, and focused on preventing illness before major damage occurs.
Key Takeaways
University of Hong Kong researchers developed CardiOmicScore, an AI-based framework that analyzes blood-based molecular information to estimate future cardiovascular risk.
The model examined 2,920 circulating proteins and 168 metabolites using large-scale UK Biobank data.
It was designed to predict coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism.
Researchers reported that the model could identify elevated risk up to 15 years before clinical onset and performed better than conventional polygenic risk scores.
The study does not establish that CardiOmicScore is ready for routine clinical use or that using it will prevent cardiovascular events.
Additional independent validation, diverse population testing, clinical-outcome research, regulatory review, affordability analysis, and privacy protections will be necessary.
Frequently Asked Questions
Was the Study Published on July 19, 2026?
No. The underlying research and University of Hong Kong announcement were published in March 2026. The study received renewed public attention through health-news coverage dated July 19.
What Is CardiOmicScore?
CardiOmicScore is an AI-based research system that analyzes molecular signals found in blood and converts them into personalized estimates of future cardiovascular risk.
Which Diseases Can It Predict?
The model assesses risk for coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism.
How Can It Predict Disease 15 Years Early?
The model identifies patterns among proteins and metabolites associated with people who later developed cardiovascular disease. It estimates risk rather than predicting an exact event or date.
Is the Test Available From Doctors Now?
It is not currently a standard blood test offered through routine medical appointments. The technology remains in the research and development stage.
Does a High Score Mean Someone Will Definitely Develop Heart Disease?
No. A high score would indicate elevated statistical risk, not certainty. Lifestyle changes, preventive care, treatment, and other factors may influence the outcome.
Does It Replace Cholesterol or Blood-Pressure Testing?
No. CardiOmicScore was studied as an additional risk-prediction approach. Established screening and clinical evaluation remain essential.
Why Is AI Necessary?
The blood analysis includes thousands of molecular signals. AI can identify complex relationships among those measurements that would be difficult to evaluate through conventional methods.
What Are the Main Concerns?
Important concerns include cost, false-positive results, unequal access, performance across diverse populations, patient anxiety, data privacy, and whether the test actually improves health outcomes.
Support New To Education
New To Education provides independent reporting and analysis about health research, artificial intelligence, science, education, wellness, and the developments affecting individuals, families, professionals, and communities.
Readers can support our work by sharing this article, exploring our educational and professional services, joining the New To Education community, or contributing through the support options available on our website.
Every visit, share, booking, and contribution helps us continue producing accessible reporting that explains promising medical research without overstating what the available evidence can prove.
Related Articles
New Health Research From July 6, 2026: Cholesterol Testing, Exercise, Pregnancy Risks and Brain Health
New Blood Test Research May Help Predict Cognitive Decline Years Before Symptoms Appear
Final Thoughts
CardiOmicScore offers a glimpse of what preventive medicine may eventually become.
Instead of waiting for symptoms, doctors could use molecular information to identify the earliest signs that a person is moving toward heart disease, stroke, heart failure, an irregular rhythm, poor circulation, or a dangerous blood clot.
A warning delivered 15 years early could create an enormous opportunity for prevention.
It could also create uncertainty.
Patients may be told that they are at elevated risk long before anyone knows whether disease will actually develop. Healthcare systems would need to ensure that the information leads to useful action rather than anxiety, unnecessary treatment, or expensive testing without proven benefit.
The research is promising because it combines several types of biological information rather than relying on one biomarker or a fixed genetic score. Proteins and metabolites may reveal what is happening inside the body now, including changes influenced by aging, lifestyle, disease, and the environment.
Artificial intelligence makes it possible to interpret those complicated signals together.
Still, prediction should not be confused with prevention.
The next major test for CardiOmicScore will not be whether it can produce an impressive risk number. It will be whether doctors can use that number to make better decisions, reach underserved patients, protect sensitive health data, and prevent real cardiovascular events.
Until then, the system remains an exciting research development rather than a replacement for established heart-health care.
The possibility is powerful: one blood sample could someday reveal dangers that might otherwise remain hidden for more than a decade.
The responsibility will be equally significant making sure that knowing the future risk genuinely helps people change it.
Sources
University of Hong Kong Faculty of Medicine — HKUMed Develops Innovative AI Tool: A Single Blood Test Can Predict Heart Diseases up to 15 Years Before Onset
https://www.med.hku.hk/en/news/press/20260312-hkumed-develops-innovative-ai-tool
University of Hong Kong — Cardiovascular Risk Prediction Tool
https://www.hku.hk/press/press-releases/detail/28986.html
Nature Communications — AI-Based Multiomics Profiling Reveals Complementary Omics Contributions to Personalized Prediction of Cardiovascular Disease
https://www.nature.com/articles/s41467-026-68956-6
ScienceDaily — New AI Blood Test Predicts Heart Disease 15 Years Early
https://www.sciencedaily.com/releases/2026/07/260716023603.htm
