Comparison of Molecular Recognition in Docking Versus Experimental CSD and PDB Data

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Molecular docking is crucial for structure-based drug design, generating protein-ligand poses for compound prioritization.
• The accuracy of drug design prioritization relies heavily on the quality of docked poses.
• Existing pose assessment methods have limitations, impacting the reliability of computational drug discovery.

Purpose of the Study:

• To assess the accuracy of the Vina docking algorithm by comparing its generated poses with crystallographic data.
• To identify specific deficiencies in molecular docking algorithms, such as electrostatic interactions and ligand conformations.
• To develop an improved pose scoring approach for better identification of experimental protein-ligand poses.

Main Methods:

• Applied statistical approaches to quantify atomic interaction preferences and torsional ligand strain.
• Compared Vina-generated poses against crystallographic data from the Protein Data Bank (PDB) and Cambridge Structural Database (CSD).
• Developed and tested a novel pose scoring method to enhance experimental pose retrieval.

Main Results:

• Identified potential deficiencies in the Vina docking algorithm, including underestimated electrostatic repulsion and high-energy hydroxyl conformations.
• Statistical analysis revealed specific areas where docking pose accuracy can be improved.
• The proposed pose scoring approach significantly enhanced the ability to retrieve the correct experimental pose from docked poses.

Conclusions:

• The accuracy of molecular docking poses directly impacts the success of structure-based drug design.
• Identifying and addressing specific algorithmic deficiencies is key to improving docking accuracy.
• The developed pose scoring method offers a promising advancement for prioritizing compounds in drug discovery pipelines.