Gene-Environment Interactions in Predicting Self-Harm: A Machine Learning Approach Using Explainable Artificial...

Area of Science:

• Psychiatry
• Genetics
• Public Health

Background:

• Self-harm is a significant public health issue.
• The interaction between genetic and environmental factors in self-harm is not well understood.

Purpose of the Study:

• To investigate the interaction between polygenic risk scores (PRSs) for psychiatric disorders and environmental risk factors in predicting self-harm.
• To identify key genetic and environmental predictors and their interactions in self-harm.

Main Methods:

• Utilized UK Biobank data from 156,873 participants.
• Employed polygenic risk scores (PRSs) for 15 psychiatric disorders/traits.
• Applied automated machine learning and explainable artificial intelligence for model building and analysis.

Main Results:

• Environmental factors, particularly interpersonal trauma (e.g., partner belittlement, sexual assault), were stronger predictors of self-harm than genetic factors alone.
• Gene-environment interactions explained approximately 12% of the variance in self-harm risk.
• PRSs for major depression, cannabis use disorder, and anorexia nervosa showed the most significant interaction effects.

Conclusions:

• Individuals with genetic vulnerabilities may be more susceptible to the harmful effects of interpersonal trauma.
• Findings underscore the importance of considering combined genetic and environmental risks for personalized self-harm prevention.
• Highlights the need for tailored interventions addressing specific gene-environment interactions.