High-throughput ligand screening via preclustering and evolved neural networks

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• The vast chemical space of small molecules presents a significant challenge in identifying novel drug leads.
• Existing methods for lead compound identification often lack efficiency and scalability.
• Artificial neural networks (ANNs) optimized by evolutionary computation (EC) offer a promising approach to address these limitations.

Purpose of the Study:

• To investigate the utility of preclustering small molecules before ANN optimization for quantitative structure-activity relationship (QSAR) modeling.
• To enhance the efficiency and accuracy of lead drug discovery by reducing the search space and improving predictive models.
• To identify interpretable features from QSAR models for the rational design of new drug candidates.

Main Methods:

• Application of artificial neural networks (ANNs) optimized via evolutionary computation (EC).
• Implementation of a preclustering strategy for small molecules prior to ANN optimization.
• Development of quantitative structure-activity relationship (QSAR) models for a set of HIV inhibitors.

Main Results:

• Preclustering significantly improved the efficiency of ANN optimization for QSAR model generation.
• The developed methods achieved high predictive accuracy in prescreening compounds, distinguishing active from inactive molecules.
• The approach successfully reduced the feature space while maintaining high predictive performance.

Conclusions:

• Preclustering small molecules prior to ANN optimization is a valuable strategy for efficient QSAR modeling in drug discovery.
• This integrated approach accelerates the identification of potent and specific lead compounds, including HIV inhibitors.
• The method facilitates the discovery of interpretable molecular features crucial for the rational design of novel therapeutics.