Ligand-Binding-Site Refinement to Generate Reliable Holo Protein Structure Conformations from Apo Structures

Area of Science:

• Computational Biology
• Drug Discovery
• Structural Bioinformatics

Background:

• Structure-based virtual screening relies heavily on accurate receptor protein structures.
• The absence of holo (ligand-bound) structures often leads to reduced virtual screening efficacy.
• Apo (unbound) structures are frequently used when holo structures are unavailable, posing a challenge for screening accuracy.

Purpose of the Study:

• To develop a robust computational method for generating reliable holo protein conformations from apo structures.
• To enhance the performance of virtual screening using refined protein structures.
• To validate the method's effectiveness using established benchmark datasets.

Main Methods:

• Employing molecular dynamics (MD) simulations with restraints derived from holo structure binding-site templates.
• Refining apo protein structures to achieve holo conformations.
• Benchmarking the method on the Directory of Useful Decoy-Enhanced (DUD-E) and Gunasekaran datasets.
• Evaluating virtual screening enrichment factors (EF) and ligand root mean square deviation (RMSD).

Main Results:

• Successfully refined apo binding-site structures towards holo conformations in 82% of tested cases.
• Significantly improved virtual screening performance on the DUD-E dataset, with average EF values increasing from 3.5 (apo) to 6.2 (MD-refined).
• Achieved ligand RMSD values comparable to self-docking when docking native ligands to refined structures, outperforming docking to apo structures.

Conclusions:

• The developed MD-based method effectively generates reliable holo protein conformations from apo structures.
• This approach robustly enhances virtual screening performance and improves ligand docking accuracy.
• The method offers a valuable tool for structure-based drug discovery, particularly when holo structures are initially unavailable.