CardiOmicScore
#4 v AI ve zdravotnictvíUniversity Of Hong Kong · 4× · naposledy 22. 7. 2026
CardiOmicScore is an AI model developed by the LKS Faculty of Medicine at the University of Hong Kong (HKUMed) for cardiovascular risk prediction. It uses a multitask deep learning framework (two neural networks, MetNet and ProNet) that analyzes 2,920 circulating proteins and 168 metabolites from a single blood sample to predict the risk of six cardiovascular diseases (coronary heart disease, stroke, heart failure, atrial fibrillation, peripheral arterial disease, and venous thromboembolism). The model is based on UK Biobank data and can provide early warning signals up to 15 years before clinical manifestation, according to the study. It is an academic research project published in Nature Communications with accompanying Open Source code.
Vlastnosti
| Deployment Model | Research prototype; based on UK Biobank data, code openly available on GitHub, not yet independently clinically validated |
| Use Case Scope | Prediction of risk for 6 cardiovascular diseases (CAD, stroke, heart failure, atrial fibrillation, PAD, venous thromboembolism) from a single blood test |
| Integrations | Combines proteomic and metabolomic scores with polygenic risk scores (PRS) and clinical predictors in Cox proportional hazards models |
| License | Companion code released on GitHub under MIT license |
| Platform | Multitask deep learning framework (MetNet & ProNet), code in Python and R |
| Release Date | Nature Communications publication 2026 (vol. 17); HKUMed press release March 12, 2026 |
