Machine Learning Unlocks 'Rare Genetic Variants' in Coronary Artery Disease - Towards Personalized Prevention with High-Precision Risk Prediction

A research group led by Professor Kaoru Ito of the Graduate School of Medicine, Chiba University (also a Visiting Senior Scientist at the RIKEN Center for Integrative Medical Sciences) and Research Trainee Hiromasa Ieki (at the time of research) at the RIKEN Center for Integrative Medical Sciences, conducted a whole-genome analysis using machine learning to clarify the contribution of rare genetic variants to the onset of coronary artery disease (Note 1) in the Japanese population. As a result, they significantly improved the accuracy of disease risk stratification for rare genetic variants, which had been difficult with conventional methods. This achievement is expected to enable precise assessments tailored to individual genetic characteristics, leading to the realization of future personalized prevention and optimal medical strategies.

This research outcome was published on June 4, 2026, in the academic journal Circulation: Genomic and Precision Medicine.
(Paper here: 10.1161/CIRCGEN.125.005341)

Figure: Machine Learning Framework for Whole-Genome Analysis of Coronary Artery Disease

■ Background of the Research

Coronary artery disease, such as myocardial infarction, is one of the leading causes of death worldwide. It is estimated that genetic factors are involved in over 50% of its onset, in addition to lifestyle habits (References). To prevent this disease, it is necessary to accurately predict an individual's "susceptibility." However, conventional methods have not comprehensively elucidated how rare genetic variants possessed by each individual influence the disease. Therefore, this study aimed to establish a new method for predicting future disease risk with higher accuracy by utilizing machine learning to examine the genetic information of the Japanese population in detail and capture previously overlooked aspects.

■ Key Research Findings

1. Broke through the limitations of conventional analysis methods and identified 59 associated genes. This revealed that not only lipid metabolism but also diverse mechanisms such as immune function and platelet function are involved in the disease.

2. Based on the data of identified rare genetic variants, a new "Rare Variant Score (RVS)" was developed to quantify an individual's disease risk (Figure upper right). RVS not only distinguishes between the presence or absence of disease but also predicts long-term cardiovascular death, suggesting its potential utility in predicting disease severity in clinical settings.

3. A comparison between RVS and the currently mainstream Polygenic Risk Score (PRS) (Note 3) revealed that RVS reflects unique genetic risks not captured by PRS. Furthermore, constructing an "Integrated Risk Score" by combining the two scores demonstrated a significant improvement in disease prediction accuracy (Figure lower right).

■ Future Prospects

This research outcome lays the foundation for highly accurate risk prediction based on rare genetic variants, which have been overlooked until now, and for precision personalized medicine tailored to individual constitutions. In the future, we aim to elucidate a more comprehensive disease system through multi-omics analysis.

■ Glossary of Terms

Note 1) Coronary Artery Disease: A general term for diseases where the blood vessels (coronary arteries) that supply blood and oxygen to the heart muscle become narrowed or blocked due to arteriosclerosis, etc. Typical examples include myocardial infarction and angina pectoris.

Note 2) Rare Genetic Variant: A "genetic variant" is a slight difference in an individual's genetic information (genome) that is linked to disease. Among these, a rare genetic difference found in only a very small number of people is called a "rare genetic variant."

Note 3) Polygenic Risk Score: An index that quantifies an individual's risk of developing a specific disease in the future by summing up individual differences across tens of thousands to millions of genomic (genetic information) sites.

■ Publication Information

Title: Machine Learning Reveals the Contribution of Rare Genetic Variants and Enhances Risk Prediction for Coronary Artery Disease in the Japanese Population

Authors: Hirotaka Ieki, Sai Zhang, Satoshi Koyama, Martin Kjellberg, Hiroki Yoshida, Ryo Kurosawa, Hiroshi Matsunaga, Kazuo Miyazawa, Nobuyuki Enzan, Changhoon Kim, Jeong-Sun Seo, Koichiro Higasa, Kouichi Ozaki, Yoshihiro Onouchi, The Biobank Japan Project, Koichi Matsuda, Yoichiro Kamatani, Chikashi Terao, Fumihiko Matsuda, Michael Snyder, Issei Komuro, Kaoru Ito

Journal: Circulation: Genomic and Precision Medicine

DOI: 10.1161/CIRCGEN.125.005341

■ References

Title: Heritability of death from coronary heart disease: a 36-year follow-up of 20 966 Swedish twins

Journal: Journal of Internal Medicine

DOI: 10.1046/j.1365-2796.2002.01029.x.

■ About the Research Project

This research was supported by the Japan Agency for Medical Research and Development (AMED) under the PRIME program (Creating Predictive Technologies for Understanding Sex and Individual Differences) for the research project "Elucidation of Multifactorial Circulatory Disease Systems Driven by Static and Dynamic Multi-omics Fusion and Realization of Preventive Medicine."
Keywords: