Mining mouse behavior for patterns predicting psychiatric drug classification

Area of Science:

• Neuroscience
• Pharmacology
• Computational Biology

Background:

• Predicting psychopharmacological effects is crucial for psychiatric drug discovery.
• Current methods rely on multiple, labor-intensive, disorder-specific behavioral assays.

Purpose of the Study:

• To investigate the feasibility of a single high-throughput behavioral assay for classifying psychiatric drugs.
• To develop a model for classifying compounds into multiple psychopharmacological classes.

Main Methods:

• Utilized Pattern Array data mining to analyze ~100,000 mouse exploratory behaviors.
• Developed a classification model integrating key behavioral patterns to predict drug class and dose.
• Classified compounds into six clinically relevant categories: antipsychotic, antidepressant, opioids, psychotomimetic, psychomotor stimulant, and α-adrenergic.

Main Results:

• A small subset of behaviors, including a 'universal drug detector', accurately predicted drug class.
• The model demonstrated dose-dependent effects and correctly classified 9/11 unknown compounds in blind validation.
• Misclassifications suggested potential drug repurposing opportunities.

Conclusions:

• The developed classification model offers a more efficient alternative to standard animal models for drug screening.
• The model is systematically updatable, enhancing predictive power and accommodating new therapeutic classes.
• Highlights the utility of data mining for behavioral phenotyping and drug discovery.