AI Powered Blood Test Predicts Suicide Risk in Bipolar Patients
Posted on 29 Oct 2025
Suicide remains one of the gravest risks for individuals with bipolar disorder, with nearly 40% attempting suicide and up to 10% dying by it. Despite these alarming statistics, clinicians have lacked reliable tools to identify patients most at risk. Researchers have now developed an artificial intelligence (AI)-powered blood test that can predict suicide risk in bipolar patients with over 95% accuracy, offering a potentially life-saving diagnostic breakthrough.
In collaborative research led by the University of Haifa (Haifa, Israel), the team developed an AI model based on genetic changes in white blood cells from patients with bipolar disorder. Using blood samples from six patients who died by suicide and 14 others with differing risk profiles, researchers isolated and “immortalized” white blood cells with the Epstein-Barr virus, allowing them to live indefinitely outside the body.
They then extracted RNA to identify differences in gene expression between high- and low-risk individuals. Machine-learning algorithms were trained on this genetic data to detect molecular patterns linked to suicide risk. In statistical analyses equivalent to 1,700 repeated cross-validations, the model consistently achieved over 95% accuracy in distinguishing patients who later died by suicide from those who did not.
Remarkably, the findings published in Translational Psychiatry showed that genetic signatures associated with brain function and psychiatric disorders could be detected in blood cells. The researchers were surprised to discover that they had a very good prediction of which patients would commit suicide by finding markers in white blood cells.
The results suggest that white blood cells mirror some of the same genetic processes occurring in the brain, offering an accessible window into neurological and psychiatric disorders. These white blood cells also express many neuronal genes, suggesting that this unexpected link may help bridge biological and behavioral understanding of suicide risk.
The research also challenges previous assumptions that suicide-related biomarkers could only be found in neurons or postmortem brain tissue. Future work will expand the study to larger, more diverse patient populations to ensure the genetic signatures are stable and reproducible. The team is now using induced pluripotent stem cell (iPSC) technology to derive neurons from blood samples and further investigate molecular mechanisms underlying suicidal behavior.
Experts say this approach represents a major advance for psychiatry. The team hopes the technology could one day become a clinical screening tool, helping psychiatrists adjust treatment plans and monitor patients proactively.
