An Effective Approach for Clustering InhA Molecular Dynamics Trajectory Using Substrate-Binding Cavity Features.

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Molecular dynamics (MD) simulations provide crucial protein receptor flexibility data for drug discovery.
• Incorporating full MD ensembles in docking is computationally infeasible for large compound libraries.
• Existing clustering methods using Root-Mean Square Deviation (RMSD) may not accurately represent plastic active sites.

Purpose of the Study:

• To develop and evaluate a novel clustering strategy for reducing MD ensembles.
• To compare the effectiveness of partitioning and hierarchical clustering methods for receptor conformation analysis.
• To assess the performance of the new method in drug candidate binding evaluation using cross-docking experiments.

Main Methods:

• Applied k-means, k-medoids, and four agglomerative hierarchical clustering methods.
• Generated ensembles of representative MD conformations using medoids from identified clusters.
• Evaluated the method using cross-docking experiments on the InhA enzyme with 20 ligands and a 20 ns MD trajectory.

Main Results:

• The novel ensemble, representing only 0.48% of MD conformations, reproduced 75% of dynamic behaviors in binding cavity docking.
• The new approach significantly outperformed two RMSD-based clustering solutions.
• The method successfully distilled biologically relevant information from MD trajectories for docking.

Conclusions:

• The developed clustering strategy offers a computationally efficient way to represent protein receptor flexibility from MD simulations.
• This method enhances the accuracy and feasibility of molecular docking for drug discovery.
• The approach shows promise for analyzing large MD datasets and identifying potential drug candidates.