Modeling Ligand Binding Site Water Networks with Site-Identification by Ligand Competitive Saturation: Impact on Ligand Binding Orientations and Relative Binding Affinities

Water molecules play a pivotal role in modulating protein-ligand interactions, wielding a substantial influence on ligand binding affinity and orientation in many systems. Water molecules can facilitate the interaction between a protein and ligand through hydrogen-bond networks. Conversely, water molecules occupying a ligand binding site may need to be displaced at an energetic cost prior to ligand binding. As a result, appropriate treatment of water contributions to protein-ligand interactions is important for ligand design. However, this is a very challenging problem in the context of adequately determining the number of waters to consider and undertaking the conformational sampling of the ligands, the waters, and the surrounding protein. In the present study, an extension of the Site Identification by Ligand Competitive Saturation-Monte Carlo (SILCS-MC) docking approach is presented that enables determination of the location of water molecules in the binding pocket and their impact on the predicted ligand binding orientation and affinities. The approach, termed SILCS-WATER, involves initial exhaustive MC docking of the ligand and 15 water molecules in a binding site followed by selection of a subset of waters for additional MC sampling. In the approach, the Metropolis criteria for the MC is based on the overlap of the ligand and water with the SILCS FragMaps and the interaction energy between the waters and ligand. Results show that the SILCS-WATER methodology is able to capture important waters and improve ligand positions and orientation. For 6 of 10 multiple-ligand protein systems the method improved the relative binding affinity prediction against experimental results to varying degrees when compared to standard SILCS-MC. The method identifies waters interacting with ligands that occupy unfavorable locations with respect to the protein whose displacement through the appropriate ligands modifications should improve ligands binding affinity. Results are consistent with the binding affinity being modeled as a ligand-water complex interacting with the protein. The presented approach offers new possibilities in revealing water networks and their contributions to the binding affinity of a ligand to a protein and is anticipated to be of utility for computer-aided drug design.