The Role of Artificial Intelligence in Genetic Disease Risk Assessment

In the era of modern technology, artificial intelligence plays a crucial role in various fields, particularly in medicine. This article reviews a recent study conducted by a team of researchers at the Icahn School of Medicine at Mount Sinai, where AI was used to explore genes and assess disease risks.

Introduction to the Study and Its Objective

The study aims to address the challenge of accurately assessing the risk of genetic diseases. By using artificial intelligence along with routine laboratory tests such as cholesterol levels, blood analyses, and kidney functions, the researchers strive to provide a more precise, data-driven insight into genetic disease risks.

Details of this study were published in the online edition of Science on August 28. The new approach seeks to go beyond the traditional limitations of genetic studies that rely on a simple yes-or-no diagnosis to classify patients.

Using AI to Understand the Genetic Spectrum of Diseases

The study indicates that many diseases, such as hypertension, diabetes, and cancer, cannot be easily classified into binary categories. Therefore, researchers trained AI models to measure disease on a spectrum, offering a more accurate understanding of how disease risks develop in real life.

Dr. Ron Do, the study’s lead author, explains that using AI with routine medical data can provide a better estimate of the likelihood of disease development in individuals carrying certain genetic variants.

Developing Models and Applying Them to Common Diseases

Utilizing over a million electronic health records, the researchers developed AI models for ten common diseases. These models were applied to individuals with rare genetic variants, producing a score between 0 and 1 that reflects the likelihood of disease development.

High scores indicate that the variant may significantly contribute to disease development, while low scores suggest minimal or no risk.

Surprising Results and Future Applications

Some results were surprising, as variants previously considered uncertain showed clear pathological signals, while others thought to cause diseases had minimal impact in real-world data.

Dr. Ian Forrest, a co-author of the study, emphasizes that the model is not intended to replace clinical judgment but can serve as a valuable guide, especially when test results are unclear.

Conclusion

This study points to a potential future where AI and routine clinical data work together to provide more personalized and actionable insights for patients and their families when dealing with genetic test results. It is hoped that this approach will become a scalable means to support better decision-making, clearer communication, and greater confidence in the true meaning of genetic information.