Highly Flexible Ligand Docking: Benchmarking of the DockThor Program on the LEADS-PEP Protein-Peptide Data Set

Area of Science:

• Computational biology
• Structural biology
• Drug discovery

Background:

• Protein-peptide interactions are vital for cellular functions and drug development.
• Predicting binding modes for flexible peptides in protein docking remains a significant computational challenge.

Purpose of the Study:

• To evaluate the performance of the DockThor program for protein-peptide docking.
• To compare DockThor's pose prediction accuracy against other leading docking software using the LEADS-PEP dataset.

Main Methods:

• DockThor was tested on the LEADS-PEP dataset, comprising 53 protein-peptide complexes.
• Pose prediction accuracy was assessed using backbone Root Mean Square Deviation (RMSD) compared to crystal structures.
• DockThor's performance was benchmarked against AutoDock, AutoDock Vina, Surflex, GOLD, Glide, rDock, DINC, and HPepDock.

Main Results:

• DockThor successfully docked 40% of cases within 2.5 Å backbone RMSD for the top-scored pose, comparable to Glide.
• For the best-RMSD pose, DockThor achieved a 60% success rate in pose prediction.
• While rDock and HPepDock showed superior performance, DockThor demonstrated suitability for docking highly flexible ligands.

Conclusions:

• DockThor is a valuable tool for protein-peptide docking, particularly for flexible ligands.
• Its performance suggests applicability in virtual screening for peptide-based drug discovery.
• The DockThor web server is freely available for research purposes.