Scientists Develop ChatGPT-Like AI That Predicts Stroke Risk from DNA

A new artificial intelligence tool that identifies genetic risk factors for stroke could pave the way for more personalized prevention and treatment.

Researchers have developed an AI model that analyzes a person’s DNA to identify genetic risk factors for stroke, the second leading cause of death worldwide. A team at the University of Technology Sydney (UTS) identified specific genes linked to a higher susceptibility to ischemic stroke, the most common type of stroke, in which a blood clot blocks blood flow to the brain. The study’s findings have been published on the ScienceDirect platform.

The study’s lead author, Kairui Guo, said that people who have a stroke, particularly at a young age, often want to know whether other family members are also at risk. According to him, the new tools could help families better understand this risk.

According to earlier studies, genetic factors account for about 38 percent of a person’s stroke risk. Traditionally, genetics was not considered a major factor in stroke risk assessment. Guo noted that the new work shows how significant the impact of genetic factors is. In his view, processing such a large volume of data at the individual level would not have been possible without AI.

The team developed a new model based on transformer technology, which underlies large language models such as ChatGPT. The model was trained on genetic variations to identify links to ischemic stroke.

The model identified 18 genetic variants associated with ischemic stroke. The EDEM2 gene, which is active in the cerebral cortex, and the KANK2 gene, found in the blood, were highlighted as key determinants of stroke susceptibility.

The tool was tested on two large medical datasets covering more than 40,000 people. The AI successfully classified individuals into low, medium and high genetic risk groups. Over time, people in the high-risk group were 21 percent more likely to have a first ischemic stroke than those in the low-risk group.

Until now, scientists have calculated genetic risk using a statistical tool called polygenic risk scores (PRS). This method estimates a person’s risk of developing a particular disease by looking at the combined effect of numerous genetic variants. However, it does not take into account how different variants interact with each other or how their positions in the DNA sequence affect overall risk.

Jie Lu, director of the Australian Artificial Intelligence Institute, said that AI can be used to uncover links between human genes and diseases and to produce individual risk assessments. This could help prevent a life-changing or potentially fatal event.

According to Lu, the next stage will be to create a multimodal AI tool that can combine genetic risk assessment with other data, such as health and lifestyle questionnaires, examinations and imaging.

“Genetics is only one part of stroke risk. Combining genetic scores with established risk factors such as blood pressure, age, atrial fibrillation, type 2 diabetes and cholesterol could strengthen monitoring and prevention,” she said.