An efficient searching algorithm based on local shape complementarity of hydrogen bonds and relaxation labeling for protein-ligand docking

A new search algorithm based on the hydrogen bond shape complementarity and relaxation labeling for fast protein-ligand docking is proposed. Given that the protein-ligand binding sites are demonstrated to relate with the cavities in the protein structure, Fpocket was used to identify the pockets on the proteins. Then based on the hydrogen bond model, the protein pockets as well as the ligand can be transformed to a series of objects and labels representing potential hydrogen bond donors and acceptors. In order to find the potential hydrogen bonds made up of those donors and acceptors, a relaxation labeling is carried out. The local patch histogram is introduced for the estimation of initial probability and the coexistence hydrogen bond criterion according to the geometric constraint is used to compute the compatible coefficient. As a result, the matching pairs can be identified after the relaxation labeling. Then the ligand position can be located based quaternion framework for each matching. 92 flexible complexes with at least one intermolecular hydrogen bond are taken into consideration of our experiment and the result shows a high successful docking rate of the proposed algorithm.