Article Impact Level: HIGH Data Quality: STRONG Summary of Nature Medicine. https://doi.org/10.1038/s41591-025-03648-0 Dr. Shang-Fu Chen et al.
Points
- This study developed a machine learning model to predict coronary artery disease (CAD) risk using data from the UK Biobank. The model incorporates traditional risk factors, genetic predispositions, lifestyle factors, and medical history.
- The model identified 53 predictive risk factors, including physical measurements, blood biomarkers, and genetic variants, and accurately predicted CAD risk, with 62.9% of high-risk individuals developing CAD in 10 years.
- The machine learning model outperformed traditional clinical methods, improving risk classification for about one in four individuals and accurately identifying high-risk patients, especially in younger individuals and women.
- Validation using the NIH All of Us dataset showed that the model effectively predicts CAD risk across diverse populations, including European, African, and Hispanic groups, demonstrating its broader applicability.
- Genetic predisposition emerged as the strongest predictor of CAD risk, influencing how individuals responded to interventions. This highlights the importance of personalized medicine in CAD prevention and improving risk stratification.
Summary
This study developed a machine learning model to predict coronary artery disease (CAD) risk by incorporating traditional risk factors like age, genetic predispositions, lifestyle factors, and medical history. Using data from the UK Biobank, the model initially considered over 2,000 features, including physical measurements, blood biomarkers, and genetic variants, which were then narrowed down to 53 predictive risk factors. The model identified individuals at high risk of CAD, with 62.9% of those identified as high-risk developing CAD after 10 years, compared to only 0.3% in the lowest-risk group.
The model could predict CAD significantly better than traditional clinical methods. It improved risk classification for approximately one in four individuals, allowing for more accurate identification of high-risk individuals, especially among typically low-risk groups such as younger patients and women. The model’s accuracy was further validated by its performance on a separate cohort from the National Institutes of Health All of Us dataset, which includes more diverse populations. This showed that the model could effectively predict CAD risk across various ethnicities, including European, African, and Hispanic populations.
The study also found that genetic predisposition was the strongest predictor of CAD risk, surpassing other factors like lifestyle. Genetic risk for CAD-related conditions such as high cholesterol, high blood pressure, and diabetes influenced how much benefit individuals gained from interventions targeting these conditions. The results from this study underscore the importance of personalized medicine in CAD prevention and suggest that incorporating genetic risk factors into predictive models can significantly improve risk stratification and guide personalized treatment strategies.
Link to the article: https://www.nature.com/articles/s41591-025-03648-0
References Chen, S.-F., Lee, S. E., Sadaei, H. J., Park, J.-B., Khattab, A., Chen, J.-F., Henegar, C., Wineinger, N. E., Muse, E. D., & Torkamani, A. (2025). Meta-prediction of coronary artery disease risk. Nature Medicine. https://doi.org/10.1038/s41591-025-03648-0
