The Performance of Several Docking Programs at Reproducing Protein-Macrolide-Like Crystal Structures

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Molecular docking is crucial for predicting ligand-receptor interactions.
• Accurate pose prediction is essential for virtual screening and drug design.
• Evaluating docking program performance is vital for optimizing computational drug discovery workflows.

Purpose of the Study:

• To assess the accuracy of five docking programs (Glide, AutoDock Vina, DOCK, AutoDock 4.2.6, AutoDock 3.0.5) in reproducing crystallographic structures of macrolide and macrocyclic complexes.
• To compare the speed and accuracy of different docking algorithms.
• To evaluate the effectiveness of re-scoring methods in improving pose prediction.

Main Methods:

• Self-docking calculations were performed to assess program performance.
• Standard docking calculations were conducted using lowest-energy conformers and nearby conformers.
• Root-mean-square deviation (RMSD) values were used to measure accuracy.
• Re-scoring methods including Amber Score and MM-GBSA were applied.

Main Results:

• Self-docking showed excellent performance (mean RMSD ≤ 1.0) with AutoDock Vina being notably fast.
• Glide, AutoDock Vina, and DOCK demonstrated superior accuracy in predicting high-affinity poses.
• AutoDock Vina outperformed other methods in selecting poses closest to crystal structures.
• Re-scoring methods (AutoDock 4.2.6//AutoDock Vina, Amber Score, MM-GBSA) significantly improved agreement between calculated and experimental data.

Conclusions:

• Glide, AutoDock Vina, and DOCK are highly reliable for predicting accurate ligand poses in molecular docking studies.
• Re-scoring techniques enhance the predictive power of docking simulations.
• The choice of docking program and re-scoring strategy impacts the reliability of computational drug discovery efforts.