AI Blood Test Estimates Risk of Six Cardiovascular Diseases but Remains Unvalidated

Researchers at The University of Hong Kong built CardiOmicScore through deep learning applied to multiomics data from the U.K. Biobank. The model incorporates measurements of 2,920 circulating proteins and 168 metabolites to assess an individual's likelihood of developing coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease or venous thromboembolism.

In high-risk individuals the tool identified biological warning signs as much as 15 years prior to disease onset and showed greater accuracy than conventional polygenic risk scores, with performance improving when age and gender were included.

Cardiovascular disease caused an estimated 19.8 million deaths worldwide in 2022. Professor Zhang Qingpeng, the study's lead author, said genes set a baseline risk while proteins and metabolites reflect current health. The AI decodes those signals to allow earlier identification of risks that could be altered through lifestyle changes or prevention, he stated.

Rick Snyder, a board-certified interventional cardiologist and president of HeartPlace Cardiology, said the approach may help physicians spot biological signals years before symptoms appear. He cautioned that the model was developed primarily from U.K. Biobank data, has not been independently validated and should not replace existing risk assessments or physician judgment.

The study appeared in Nature Communications in 2026 under DOI 10.1038/s41467-026-68956-6.