By Erin Jester
A new dual-purpose AI model can predict the risk of complex diseases such as Alzheimer’s and Type 2 diabetes while also helping scientists understand how those diseases develop in the body, according to a UF College of Public Health and Health Professions faculty member-authored paper published Friday in the journal Nature Biotechnology.
In a study conceptualized and designed by Sai Zhang, Ph.D., an assistant professor in the Department of Epidemiology, the model, called single-cell polygenic risk score or scPRS, showed superior accuracy in disease prediction over existing models. The breakthrough development could lead to earlier interventions and more personalized treatments for some of the most challenging health conditions.
scPRS analyzes how tiny changes in a person’s DNA affect cell function, allowing researchers the ability to predict disease risk with high accuracy — 74%, in the case of Alzheimer’s disease — potentially years before symptoms appear.
The model also helps identify which cells are driving diseases and how, providing a deeper understanding of genetic factors influence disease.
For example, scientists using scPRS discovered that in addition to beta cells, lesser-known alpha cells in the pancreas play a role in the development of type 2 diabetes. In Alzheimer’s disease, the model linked a specific genetic change to problems in microglia — immune cells in the brain — leading to issues with waste removal that may contribute to disease progression.
The study also applied scPRS to hypertrophic cardiomyopathy and severe COVID-19.
Zhang said scPRS can be applied to all complex diseases with a genetic basis, and that the research team will extend the framework to incorporate rare variants and other single-cell omic modalities.