UTS AI Model Analyzes DNA to Predict Ischemic Stroke Risk

Researchers at the University of Technology Sydney have developed an artificial intelligence model that analyzes human DNA to improve prediction of ischemic stroke risk, a step toward more personalized preventive care. Stroke remains the world's second leading cause of death, and genetic susceptibility is an important contributor alongside lifestyle and clinical factors. The new transformer-based tool goes beyond standard genetic risk scores by accounting for the position of genetic variants in the genome and their potential interactions. The team trained the model on thousands of genetic variations using data from the UK Biobank and the U.S. All of Us database, then evaluated it against health records. Published in Computer Methods and Programs in Biomedicine, the study identified 18 genetic variants associated with ischemic stroke, the most common form of stroke. Individuals with the highest genetic risk score had a 22% greater likelihood of stroke than those in the lower-risk group. The analysis highlighted EDEM2 in brain cortex tissue as a key susceptibility gene, offering a potential target for biological research. The project was led by Dr. Yue Eric Yang, who completed a Ph.D. in artificial intelligence at UTS, and guided by Distinguished Professor Jie Lu, director of the Australian Artificial Intelligence Institute. Corresponding author Dr. Kairui Guo has been collaborating with the stroke unit at Liverpool Hospital in Sydney, with clinical input from neurologist Dennis Cordato. The researchers say the approach could help families understand inherited risk, especially after a stroke occurs at a young age. Looking ahead, the group is proposing a preventive clinical trial that would sequence genomes of an initial cohort and analyze stroke risk, with interest in sex-based differences. It is also developing a multimodal AI system that combines genetic scores with health surveys, scans, imaging and established risk factors such as blood pressure, age, atrial fibrillation, type 2 diabetes and cholesterol to support monitoring and prevention for stroke and other diseases.