Patterns for prediction of hydration around polar residues in proteins.

The atomic co-ordinates of seven very high resolution (< 1.4 A) proteins have been used to define a water-hydrogen bond template for five polar side-chains (arginine, aspartic acid, glutamic acid, asparagine and glutamine). The average water molecule positions determined were consistent with the hydrogen bonding stereochemistry expected for each side-chain. Hydrogen bonding geometry around nitrogen atoms was significantly better localized than around oxygen atoms, perhaps because of the proton on nitrogen. A prediction algorithm written to locate water molecule sites around these side-chains from the protein co-ordinates only was tested for crambin as well as for two high resolution protein structures not included in the hydrogen bond data base. The root-mean-square deviation of the predicted positions from the crystallographically determined ones for these structures was better than the resolution of these structures. The method also successfully predicted water positions for X-ray refinement of two proteins, indicating that predicted water molecules are within the radius of convergence of refinement. This method has utility for X-ray models as well as for analysis of enzyme hydration and function.