Virtual Screening Using Pharmacophore Models Retrieved from Molecular Dynamic Simulations.

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Pharmacophore models are crucial for identifying potential drug candidates from large compound libraries.
• Pharmacophores derived from molecular dynamics simulations show superior performance compared to those from static crystal structures.
• The large number of pharmacophores generated computationally poses a challenge for efficient virtual screening.

Purpose of the Study:

• To develop a method for selecting distinct representative pharmacophores to improve computational efficiency.
• To introduce a novel conformer coverage approach for ranking compounds using representative pharmacophores.
• To enhance the accuracy of virtual screening by optimizing pharmacophore selection and compound ranking.

Main Methods:

• Implemented a method to remove redundant pharmacophores using three-dimensional (3D) pharmacophore hashing.
• Developed a conformer coverage approach for ranking compounds against selected representative pharmacophores.
• Applied and validated the methods using four cyclin-dependent kinase 2 (CDK2) complexes with various ligands.

Main Results:

• The proposed selection and ranking methods demonstrated superior performance over the common hits approach for CDK2 targets.
• Ranking based on averaged scores from multiple complexes outperformed ranking based on individual complex scores.
• The developed methods were integrated into the open-source software pharmd.

Conclusions:

• The novel pharmacophore selection and ranking strategies significantly improve the efficiency and accuracy of virtual screening.
• Averaging scores across multiple protein-ligand complexes enhances the reliability of compound ranking.
• The pharmd software provides a valuable tool for computational drug discovery and lead identification.