Identifying patients with major depressive disorder based on tryptophan hydroxylase-2 methylation using machine...

Area of Science:

• Computational psychiatry
• Epigenetics and gene regulation
• Machine learning in clinical diagnostics

Background:

• Major Depressive Disorder (MDD) diagnosis relies on clinical assessment, with a need for objective biomarkers.
• Tryptophan hydroxylase-2 (TPH2) gene methylation and environmental stressors are implicated in MDD pathophysiology.
• Machine learning (ML) offers potential for integrating complex biological and environmental data for diagnostic support.

Purpose of the Study:

• To develop and evaluate ML models for identifying MDD patients.
• To investigate the predictive power of TPH2 methylation and environmental stress markers in MDD.
• To assess the performance of Support Vector Machine (SVM), Back Propagation Neural Network (BPNN), and Random Forest (RF) algorithms.

Main Methods:

• Collected data from 291 MDD patients and 100 healthy controls, including demographic information, Negative Life Events Scale (NLES), Childhood Trauma Questionnaire (CTQ) scores, and TPH2 methylation levels at 38 CpG sites.
• Utilized information gain for feature selection and employed SVM, BPNN, and RF algorithms for model development.
• Evaluated model performance using 10-fold cross-validation and assessed feature importance with SHapley Additive exPlanations (SHAP).

Main Results:

• Gender, NLES scores, CTQ scores, and 13 TPH2 CpG sites were identified as key predictors.
• All three ML algorithms demonstrated satisfactory performance in predicting MDD.
• The BPNN model achieved the highest prediction accuracy.

Conclusions:

• ML models integrating TPH2 methylation and environmental stress show high efficacy in identifying MDD patients.
• These findings highlight the potential of AI to augment traditional diagnostic methods for MDD.
• The study provides a foundation for developing data-driven diagnostic tools in psychiatry.