AI Model Identifies 18 Genetic Variants Linked to Stroke Risk
A groundbreaking AI model developed at UTS reveals genetic factors that could enhance personalized stroke assessments.
The AI model identifies 18 genetic variants associated with ischaemic stroke.
Individuals with high genetic risk scores face a 22% increased chance of stroke.
The research aims to integrate genetic data with clinical risk factors for better stroke prevention.
Researchers at the University of Technology Sydney (UTS) have developed an innovative AI model that identifies 18 genetic variants linked to the risk of ischaemic stroke, the most prevalent type of stroke caused by blocked blood supply to the brain. This advancement offers a promising avenue for enhancing personalized stroke assessments, potentially leading to improved prevention strategies.
The study highlights that individuals with the highest genetic risk scores are 22% more likely to experience a stroke compared to those in lower-risk categories. The model utilizes a transformer-based AI technology, similar to that used in large language models like ChatGPT, to analyze extensive genetic variations and their interactions within the genome. This method has shown superior performance compared to existing genetic risk assessment tools when evaluated against health datasets.
Yue (Eric) Yang, the lead researcher and a recent PhD graduate in artificial intelligence from UTS, emphasized that their model significantly outperformed previous tools in distinguishing stroke patients from non-patients. The research utilized genetic and health data from the UK Biobank and the All-of-us database in the United States, further validating the model's effectiveness.
The findings also pinpointed the EDEM2 gene in brain cortex tissue as a significant contributor to stroke susceptibility, suggesting a potential target for future research into the biological mechanisms underlying strokes. Stroke remains the second leading cause of death globally, with approximately 46,000 Australians affected annually, and one-third of survivors facing moderate to severe disabilities.
The research team is now focused on developing a multimodal AI system that integrates genetic assessments with clinical and lifestyle data. Jie Lu, director of the Australian Artificial Intelligence Institute, noted that combining genetic information with established clinical risk factors, such as blood pressure and diabetes, could lead to more comprehensive risk assessments and better prevention strategies. The upcoming tool aims to unify various data types, including health surveys and imaging, to enhance stroke risk evaluation and address other health conditions like endometriosis.
