EDock: blind protein-ligand docking by replica-exchange monte carlo simulation

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Drug Discovery

Background:

• Protein-ligand docking is crucial for virtual screening and annotating protein functions.
• Existing methods often require high-resolution protein structures and predefined binding sites, which are unavailable for many targets.
• Developing blind docking methods for low-resolution models is essential for exploring unknown proteins.

Purpose of the Study:

• To introduce EDock, a novel Monte Carlo-based method for blind protein-ligand docking.
• To enable docking without predefined binding sites and with low-resolution protein models.
• To improve the accuracy and robustness of protein-ligand interaction predictions.

Main Methods:

• Binding sites are predicted using sequence-profile and substructure searches with graph-matched initial ligand poses.
• Replica-exchange Monte Carlo (REMC) simulations refine ligand conformation using a physical force field and binding-site constraints.
• The method was validated on large datasets of protein-ligand pairs.

Main Results:

• EDock achieved an average ligand RMSD of 2.03 Å on experimental structures, outperforming DOCK6 (2.68 Å) and AutoDock Vina (3.92 Å).
• Using predicted protein models (I-TASSER), EDock showed significantly higher success rates than DOCK6 (159%) and AutoDock Vina (67%).
• Key advantages include accurate binding site prediction and extensive REMC sampling with adaptable van der Waals weightings.

Conclusions:

• EDock provides a robust and accurate solution for blind protein-ligand docking, especially with low-resolution or predicted protein structures.
• The method's ability to predict binding sites and handle structural uncertainties makes it valuable for drug discovery targeting unknown proteins.
• EDock advances computational approaches for virtual screening and protein function annotation where experimental structural data is limited.