Liquid Biopsy-Based Detection and Response Prediction for Depression

Area of Science:

• Biochemistry
• Artificial Intelligence
• Psychiatry

Background:

• Predicting antidepressant treatment response is vital for effective mental healthcare.
• Current diagnostic methods rely on subjective indicators, necessitating objective approaches.
• Personalized therapeutic strategies require reliable biomarkers for treatment success.

Purpose of the Study:

• To develop a deep learning-based method for depression detection and antidepressant treatment response prediction.
• To utilize spectroscopic analysis of extracellular vesicles (EVs) for objective diagnostic indicators.
• To enhance personalized medicine through a novel liquid biopsy approach.

Main Methods:

• Extracellular vesicles (EVs) were isolated from plasma samples of depressed and non-depressed individuals.
• Raman spectroscopy was employed to acquire spectral data from EVs.
• A deep learning algorithm was developed and validated for depression diagnosis and treatment response prediction.

Main Results:

• The algorithm achieved an Area Under the Curve (AUC) of 0.95 in distinguishing depression patients from healthy individuals and those with panic disorder.
• The model demonstrated high accuracy in identifying depression patients likely to respond to antidepressants, with an AUC of 0.91 for classifying responders and non-responders.
• Explainable AI (XAI) was applied to provide a diagnostic foundation and support personalized medicine.

Conclusions:

• Deep learning-based spectroscopic analysis of plasma EVs offers a highly accurate and objective method for depression diagnosis.
• This approach can predict antidepressant treatment response, paving the way for personalized psychiatric care.
• The study highlights the potential of liquid biopsy using EVs for mental disorder diagnosis and companion diagnostics.