Pathway and network embedding methods for prioritizing psychiatric drugs.

Area of Science:

• Genomics
• Computational Biology
• Psychiatry

Background:

• Mental illnesses affect many Americans, with current treatments often failing due to complex genetic and signaling disruptions.
• Transcriptional changes are key to understanding psychiatric disease pathophysiology, highlighting the need for personalized diagnostic and treatment approaches.
• Shared genetic modules across psychiatric disorders suggest an integrated, multi-disease diagnostic strategy is feasible.

Purpose of the Study:

• To classify psychiatric diseases using patient gene expression data.
• To create augmented, disease-specific gene sets by integrating diverse biological data.
• To recommend targeted drugs for psychiatric conditions based on molecular insights.

Main Methods:

• Extracted gene expression data from schizophrenia, bipolar disorder, and major depressive disorder patient cohorts.
• Employed pathway-based approaches and a random forest model for disease diagnosis.
• Utilized protein-protein interaction networks and embedding methods to prioritize treatments.

Main Results:

• Achieved 78% accuracy in predicting psychiatric disease diagnosis using gene expression data.
• Derived key features to enhance existing drug and disease signatures.
• Developed a treatment prioritization pipeline showing a 3.4-fold improvement over baseline.

Conclusions:

• Gene expression data and pathway features can effectively diagnose psychiatric diseases.
• Molecular insights from disease classification can guide more effective treatment recommendations.
• This integrated approach offers a promising avenue for improving psychiatric care and drug discovery.