GM-Dock_Zn: a geometry matching-based docking algorithm for zinc proteins

Motivation: Molecular docking is a widely used technique for large-scale virtual screening of the interactions between small-molecule ligands and their target proteins. However, docking methods often perform poorly for metalloproteins due to additional complexity from the three-way interactions among amino-acid residues, metal ions and ligands. This is a significant problem because zinc proteins alone comprise about 10% of all available protein structures in the protein databank. Here, we developed GM-Dock(Zn) that is dedicated for ligand docking to zinc proteins. Unlike the existing docking methods developed specifically for zinc proteins, GM-Dock(Zn) samples ligand conformations directly using a geometric grid around the ideal zinc-coordination positions of seven discovered coordination motifs, which were found from the survey of known zinc proteins complexed with a single ligand. Results: GM-Dock(Zn) has the best performance in sampling near-native poses with correct coordination atoms and numbers within the top 50 and top 10 predictions when compared to several state-of-the-art techniques. This is true not only for a non-redundant dataset of zinc proteins but also for a homolog set of different ligand and zinccoordination systems for the same zinc proteins. Similar superior performance of GM-Dock(Zn) for near-native-pose sampling was also observed for docking to apo-structures and cross-docking between different ligand complex structures of the same protein. The highest success rate for sampling nearest near-native poses within top 5 and top 1 was achieved by combining GM-Dock(Zn) for conformational sampling with GOLD for ranking. The proposed geometry-based sampling technique will be useful for ligand docking to other metalloproteins.