Researchers have had limited success developing a genomic risk score (GRS) that can predict stroke. New findings published in Nature Communications, however, suggest machine learning could be the answer.
The study’s authors developed a machine learning algorithm capable of detecting patients at an increased risk of ischemic stroke, all from a single blood draw or blood sample. The algorithm was tested on data from 420,000 patients from the UK Biobank dataset.
Overall, the team’s GRS outperformed previous genetic scores and achieved a performance comparable to “other well-known risk factors for stroke, such as smoking status or body mass index.” Approximately one in 400 individuals are at a threefold increased risk of ischemic stroke, the authors observed, and this GRS could help identify those patients so that they can manage risk factors and receive necessary treatments as early as possible.
“The sequencing of the human genome has revealed many insights,” Michael Inouye, PhD, Baker Heart and Diabetes Institute in Australia, said in a prepared statement. “For common diseases, such as stroke, it is clear that genetics is not destiny; however, each person does have their own innate risk for any particular disease. The challenge is now how we best incorporate this risk information into clinical practice so that the public can live healthier and longer.”
The authors did note that their research had certain limitations. The GRS still needs to be independently validated with other patient cohorts, for instance, and the effects of some risk factors may be underestimated.
Overall, however, Inouye and colleagues said their study “lays the groundwork for larger genome-wide association studies of stroke and its multiple subtypes as well as analyses which leverage the totality of information available for stroke genomic risk prediction.”