Dr. Jingchun Chen (Nevada Institute of Personalized Medicine) and her team used deep learning AI techniques to develop genetic models based on polygenic risk scores (PRS). A PRS functions like a genetic risk calculator, combining the effects of thousands of small genetic variations to estimate the likelihood of developing specific diseases. Instead of relying on a few significant variants, the genome-wide variations collectively contribute to the risk.
The study, “,” was published in the journal Schizophrenia on February 5, 2025. Co-authored by Dr. Mira Han (School of Life Sciences), alongside faculty from the University of Texas Health Science Center at Houston, Johns Hopkins School of Medicine, and the University of Pittsburgh.
The study utilized PRS data from 42 related health conditions. These conditions include other mental health disorders (e.g., ADHD, anxiety), brain-related issues (e.g., epilepsy, migraines), personality traits (e.g., risk-taking, neuroticism), substance use (e.g., smoking, alcohol), and physical health problems (e.g., diabetes, heart disease). Additional contributions were made by Yimei Lu and Joan Manuel Cue (a mathematics undergraduate student) from Chen’s group at 8kbet.
The team found that schizophrenia, bipolar disorder, and depression share significant genetic risks with a variety of other health conditions. By incorporating these shared genetic risk scores, they significantly improved the accuracy of classifying these psychiatric disorders. Remarkably, their AI models were able to accurately classify these disorders using only the genetic risks from related conditions, without including disorder-specific genetic information.
These findings suggest that shared genetic variants may contribute to both psychiatric and other health conditions, underscoring the biological connections between mental and physical health. Leveraging AI-driven genetic analysis, this approach could revolutionize the early diagnosis, risk prediction, and personalized treatments—not just for psychiatric disorders, but for a broad range of human complex diseases in the future.
