An evaluation of structural descriptors and clustering methods for use in diversity selection

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• Diversity selection is crucial for identifying novel drug candidates.
• Traditional methods often rely on non-hierarchical clustering and less informative descriptors.
• Understanding ligand-receptor binding is key to predicting biological activity.

Purpose of the Study:

• To evaluate the effectiveness of various structural descriptors and clustering methods for diversity selection.
• To compare the performance of these methods in simulated biological activity prediction tasks.
• To identify optimal strategies for enhancing the diversity of chemical libraries.

Main Methods:

• Evaluation of multiple structural descriptors, focusing on 2D representations.
• Comparison of hierarchical and non-hierarchical clustering algorithms.
• Simulated biological activity prediction using diverse chemical datasets.
• Analysis of descriptor information content related to ligand-receptor binding forces.

Main Results:

• Simple 2D structural descriptors demonstrated high effectiveness in diversity selection.
• Hierarchical clustering methods significantly outperformed non-hierarchical approaches.
• The utility of descriptors correlated with their ability to encode ligand-receptor binding information.
• Improved accuracy in simulated biological activity predictions was observed.

Conclusions:

• Hierarchical clustering and 2D structural descriptors offer a superior approach to diversity selection.
• These methods enhance the efficiency of identifying potential drug candidates.
• Encoding ligand-receptor binding information is critical for successful diversity selection and activity prediction.