Improving protein-ligand docking results using the Semiempirical quantum mechanics: testing on the PDBbind 2016 core...

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Molecular docking is crucial for predicting ligand binding in drug design.
• Accurate pose prediction and affinity ranking are essential for virtual screening.
• Refinement of docked poses is necessary to remove steric clashes and improve accuracy.

Purpose of the Study:

• To evaluate the effectiveness of semi-empirical quantum mechanics (SQM) method PM7 for optimizing molecular docking poses.
• To compare the sampling and ranking capabilities of AutoDock4 and AutoDock Vina.
• To propose an improved protocol for structure-based virtual screening by combining both docking programs.

Main Methods:

• Geometry optimization of top-ranked docking poses using the PM7 SQM method.
• Utilizing the PDBbind core set (2016) for high-quality protein-ligand complexes.
• Comparative analysis of AutoDock4 and AutoDock Vina performance in pose generation and ranking.

Main Results:

• Docking pose optimization using PM7 enhances prediction accuracy and refines docked complexes by eliminating clashes.
• AutoDock Vina demonstrates superior sampling power for accurate ligand pose generation (RMSD ≤ 2.0 Å).
• AutoDock4 shows better ranking power compared to AutoDock Vina.

Conclusions:

• Geometry optimization with SQM methods like PM7 is beneficial for improving docking pose accuracy.
• A hybrid approach combining AutoDock Vina's sampling and AutoDock4's ranking provides a robust protocol for virtual screening.
• The proposed protocol enhances the reliability of structure-based virtual screening in drug design.