Extensive benchmark of rDock as a peptide-protein docking tool.

Area of Science:

• Computational biology
• Structural biology
• Biochemistry

Background:

• Peptide-protein interactions are crucial for cellular functions and disease.
• Accurate prediction of peptide-protein binding modes is a significant computational challenge.
• Existing docking programs often struggle with peptide targets.

Purpose of the Study:

• To evaluate the performance of the rDock program for peptide-protein docking.
• To compare rDock's accuracy against other small molecule and peptide docking tools.
• To assess rDock's utility for predicting binding modes of medium-sized peptides.

Main Methods:

• Benchmarking rDock using 100 peptide-protein systems from AutoDockVina and LEADSPEP datasets.
• Comparing rDock performance against AutoDockVina, Surflex, GOLD, Glide, PIPER-FlexPepDock, and HPepDock.
• Analyzing docking accuracy based on backbone RMSD for peptides of varying lengths.

Main Results:

• rDock achieved an overall backbone RMSD below 2.5 Å in 58.5% of cases.
• Performance was higher on the AutoDockVina set (76%) compared to the LEADSPEP set (43%).
• rDock outperformed most small molecule docking programs for 6-10 residue peptides, especially in extended binding modes.

Conclusions:

• rDock demonstrates competitive performance in peptide-protein docking, particularly for medium-sized peptides.
• Its accuracy is comparable to specialized peptide docking tools under certain conditions.
• rDock is a viable option for predicting binding modes when protein targets bind peptides in an extended conformation.