Integrated One-Against-One Classifiers as Tools for Virtual Screening of Compound Databases: A Case Study with CNS...

Area of Science:

• Chemoinformatics
• Computational Chemistry
• Drug Discovery

Background:

• Central nervous system (CNS) inhibitors are crucial drug candidates.
• Efficient methods for classifying and screening these molecules are needed.
• Existing methods may require extensive datasets or complex parameterization.

Purpose of the Study:

• To develop and validate a novel classification tool for CNS inhibitors.
• To enhance the efficiency of early-stage drug discovery through virtual screening.
• To facilitate compound retrieval and analysis in chemoinformatics.

Main Methods:

• Collected 21,833 CNS inhibitors from the Binding Database.
• Employed discriminant analysis (DA) techniques, specifically a combination of genetic algorithm and quadratic discriminant analysis (GA-QDA).
• Utilized one-against-one (OAO) QDA classifiers for molecule classification and virtual screening of PUBCHEM and ZINC databases.

Main Results:

• GA-QDA classifiers accurately separated molecules based on therapeutic targets, performing comparably to support vector machines.
• Classification models served as effective virtual filters, demonstrating high enrichment factors and AUC values.
• OAO classifiers enabled sorting of compound databases using relative distances, circumventing the need for inactive sets in virtual screening.

Conclusions:

• The GA-QDA approach provides a robust tool for classifying CNS inhibitors and charting chemical space.
• These classification models significantly accelerate early-stage drug discovery by improving virtual screening efficiency.
• The developed methods offer a valuable approach for compound retrieval and analysis in chemoinformatics, particularly for multiclass classification scenarios.