Ok so this one is kind of wild: a new AI blood test can help predict your risk of heart disease up to 15 years before symptoms appear, if you pair it with a few basic health details like your age. A research team at the University of Hong Kong’s medical school, known as HKUMed, published the study in the journal Nature Communications on February 2, 2026. A science news site called ScienceDaily featured it again on July 19, 2026, which is how it landed back in the news.
The tool is called CardiOmicScore. It reads one blood sample and turns it into a personal risk score.
What the AI blood test does
Let me explain the basics first, because this touches almost everyone. Cardiovascular disease is just a big phrase for problems with your heart and blood vessels, the tubes that carry your blood. It is the top cause of death in the world. The researchers say it caused about 19.8 million deaths in 2022.
Here is how doctors usually check your risk today. They look at simple things like your age, your blood pressure, and whether you smoke. That helps. But the researchers say these basic checks often miss small, quiet changes inside your body. Those changes can start long before you feel anything.
This is where CardiOmicScore comes in. From a single blood sample, it estimates your future risk of six major heart and blood vessel diseases. Those six are coronary artery disease, stroke, heart failure, atrial fibrillation, which is an irregular heartbeat, peripheral artery disease, and venous thromboembolism, which is a dangerous blood clot. Pair the tool with basic details like age and gender, and the researchers say it can flag high risk up to 15 years early.
Fifteen years.
How can one blood test see so far ahead
Good question. The AI reads thousands of tiny substances in your blood. To be exact, it looks at 2,920 proteins and 168 metabolites. These are small chemical messengers your body makes. The team calls them real time recorders, because they quietly show what is happening in your immune system, your metabolism, and your blood vessels right now.
The team trained the tool on the UK Biobank. That is a huge store of health data from a large number of people. They used deep learning, which just means an AI that learns by spotting patterns across huge amounts of information. It also pulls together three types of data at once: your genes, your proteins, and your metabolites. Scientists call that mix multiomics.
The lead researcher, Professor Zhang Qingpeng of HKUMed, puts it simply. He says your genes set your starting point, but your proteins and metabolites show your current health. So the team built the AI to read those live signals and spot trouble earlier than usual.
Now here is a thing I didn’t know. The older style of test, called a genetic risk score, stays mostly fixed at birth. Your genes do not really change, so that score barely moves across your life. The researchers say CardiOmicScore works differently, because it reads those live chemical signals instead. So it can reflect your current health, your lifestyle, and your surroundings. And it can shift as you do.
A couple of honest notes
It is easy to get carried away with a headline like this, so let me be fair.
First, this is not brand new, even though it is back in the news. The team published the paper in February 2026, and the university announced it in March 2026. ScienceDaily simply featured it again on July 19, 2026.
Second, this is a research study, not a product. You cannot buy it today. The sources do not say regulators have approved it. They also do not say it is available in hospitals or to the public yet.
Third, the 15 year number comes with conditions. The researchers say the early warning works when you combine the blood signals with basic clinical details. They also describe it mainly for people already in higher risk groups. So it is not a magic reading from blood alone.
Finally, the team built the tool mostly on UK Biobank data. The researchers themselves call for more outside testing. This suggests the tool may still need that testing before it can help people everywhere.
So treat this as a promising early step, not a finished tool.
Why this could matter
If we analyze, we can see the trade the researchers made. They chose to build the tool around your body’s live, changing signals, the proteins and metabolites, instead of leaning only on your fixed genes. That choice reportedly buys something valuable. It gives a warning that can arrive much earlier and that fits you personally, because the test can react to how your body is doing right now. But the choice has a cost. Live signals are messy and complex. So reading them needs heavy AI, a lot of data, and extra clinical details to work. And a tool trained mostly on one database still has to prove it works for everyone else. In simple words, they swapped simple and fixed for early and personal. That swap is the whole reason this could catch heart trouble sooner.
So here is the real question for you. Would you want to know your heart risk 15 years in advance? And if you did know, would it push you to live healthier, or would it just sit in the back of your mind and worry you?
If you like tech and science news explained in plain simple words like this, subscribe to the VulpisLab newsletter for a small daily dose.
Quick questions people are asking
What is CardiOmicScore?
It is an AI tool from a research team at the University of Hong Kong’s medical school, HKUMed. It uses one blood test to estimate your future risk of major heart and blood vessel diseases.
How early can the AI blood test warn you?
The researchers say it can flag high risk up to 15 years before symptoms appear, when you combine the blood signals with basic details like age and gender.
Which diseases can it predict?
Six major ones: coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease, and venous thromboembolism.
How does it work?
It reads 2,920 proteins and 168 metabolites in your blood, which act like live signals from inside your body. The team trained it on UK Biobank health data using deep learning.
Who made it and when?
A research team at HKUMed, led by Professor Zhang Qingpeng. They published the study in Nature Communications on February 2, 2026. ScienceDaily featured it again on July 19, 2026.
Can I get this test today?
No. It is a research study, not a product. The sources do not say it is approved or available to the public yet.
Is it proven to work for everyone?
Not yet. The team built it mostly on UK Biobank data, and the researchers say it needs more outside testing before people can trust it widely.
Source: The University of Hong Kong (HKUMed), “HKUMed develops innovative AI tool,” March 12, 2026; study published in Nature Communications, February 2, 2026 (DOI: 10.1038/s41467-026-68956-6); featured by ScienceDaily, July 19, 2026. https://www.sciencedaily.com/releases/2026/07/260716023603.htm