Models for depression recognition and efficacy assessment based on clinical and sequencing data

Area of Science:

• Psychiatric Disorders
• Computational Biology
• Machine Learning in Healthcare

Background:

• Major depression is a complex disorder influenced by genetic, neurological, and cognitive factors.
• Current diagnostic methods are often subjective and prone to misdiagnosis.
• Early detection and intervention are crucial for managing depression and optimizing treatment selection.

Purpose of the Study:

• To develop objective prediction models for diagnosing major depression using clinical assessments and sequencing data.
• To construct a model for predicting patient prognosis and treatment response.
• To identify potential biomarkers for depression diagnosis and efficacy prediction.

Main Methods:

• Data preprocessing, including normalization.
• Feature engineering to curate relevant data subsets.
• Implementation of machine learning and deep learning algorithms for model construction.

Main Results:

• A deep learning model achieved 87.26% accuracy and 91.56% AUC for depression recognition.
• A prognostic model, using 33 features, reached 75.94% accuracy and 83.33% AUC for predicting treatment effects.
• Identified 18 differential features for diagnosis and 33 for prognosis.

Conclusions:

• Deep learning models utilizing clinical and sequencing data offer accurate and objective methods for depression diagnosis and pharmacodynamic prediction.
• Selected differential features show promise as candidate biomarkers for clinical utility.
• This approach enhances the objectivity and accuracy of depression assessment and treatment monitoring.