Predicting depression severity using machine learning models: Insights from mitochondrial peptides and clinical...

Area of Science:

• Neuroscience
• Computational Biology
• Genetics

Background:

• Depression is a major global mental health issue, often linked to oxidative stress.
• The role of mitochondrial pathways in depression is not fully understood.
• Machine learning offers new methods to explore complex biological data.

Purpose of the Study:

• To classify major depressive disorders using machine learning.
• To investigate the predictive value of clinical indicators and mitochondrial oxidative stress markers.
• To assess the performance of various machine learning algorithms.

Main Methods:

• Employed six machine learning algorithms, including Random Forest.
• Classified depression based on clinical data and mitochondrial oxidative stress markers.
• Evaluated model performance on balanced and unbalanced datasets for binary and multiclass scenarios.

Main Results:

• Random Forest achieved high accuracy (e.g., 92.7% for binary classification) on balanced data.
• Including mitochondrial peptides significantly improved predictive accuracy compared to oxidative stress markers alone.
• Models demonstrated strong performance in classifying depression severity across multiple classes.

Conclusions:

• Machine learning models, particularly Random Forest, show significant potential for classifying depression.
• Mitochondrial oxidative stress markers, especially peptides, are valuable predictors of depression severity.
• These findings support machine learning's role in clinical assessment for depression, including in patients with comorbidities.