WATGEN: An algorithm for modeling water networks at protein–protein interfaces

Abstract

Water molecules at protein-protein interfaces contribute to the close packing of atoms and ensure complementarity between the protein surfaces, as well as mediating polar interactions. Therefore, modeling of interface water is of importance in understanding the structural basis of biomolecular association. We present an algorithm, WATGEN, which predicts locations for water molecules at a protein-protein or protein-peptide interface, given the atomic coordinates of the protein and peptide. A key element of the WATGEN algorithm is the prediction of water sites that can form multiple hydrogen bonds that bridge the binding interface. Trial calculations were performed on water networks predicted by WATGEN at 126 protein-peptide interfaces (X-ray resolutions <or= 2.0 A), using different criteria for water placement. The energies of the predicted water networks were evaluated in AMBER8 and used in the choice of parameters for WATGEN. The 126 interfaces include 1264 experimentally determined bridging water sites, and the WATGEN algorithm predicts 72 and 88% of these sites within 1.5 and 2.0 A, respectively. The predicted number of water molecules at each interface was much higher than the number of water molecules identified experimentally. Therefore, random placement of the same number of water molecules as that predicted at each interface was performed as a control, and resulted in only 22 and 40% of water sites placed within 1.5 and 2.0 A of experimental sites, respectively. Based on these data, we conclude that WATGEN can accurately predict the location of water molecules at a protein-peptide interface, and this may be of value for understanding the energetics and specificity of biomolecular association.

Hydration of protein–RNA recognition sites

“…The remaining 1998 waters make 4625 H-bonds with a biased distribution: 58% with the RNA polar groups and 42% with the protein polar groups. This asymmetry is also observed in interface area, where the RNA side contributes more compared to the protein side (31). Of the 442 interface waters, 367 (83%) make at least one H-bond with the other interface waters, while the rest, 75 (only 3% of total interface water), do not participate in any H-bond.…”

Halogen–water–hydrogen bridges in biomolecules

“…Water bridges are also responsible for various protein-small ligand interactions, such as hexokinase-glucose (Lunin et al, 2004), HIV-1 protease-inhibitor (Dreyer et al, 1992), and cytochrome P450cam-substrate complexes (Helms and Wade, 1995). Owing to the important functionality of watermediated interactions in biological systems, a number of computational methods were proposed to model and predict the water sites at protein surfaces and interfaces Bui et al, 2007), and the Monte Carlo and molecular dynamics simulations have also been intensively performed for ascertaining the structural basis and energetic feature of water acting on proteins and nucleic acids (Goodfellow, 1982;Steinbach and Brooks, 1993).…”

A generalized framework for computational design and mutational scanning of T-cell receptor binding interfaces

“…Because we relied on crystallographically observed water, there is a corresponding demand on the resolution and quality of the original crystal structure. As TCR-pMHC interfaces may be poorly packed and crystallographic resolutions low, incorporation of approaches to predict the location of water molecules not observed crystallographically could lead to further improvements (Bui et al, 2007;Jiang et al, 2005).…”