PREDICTING AND ANALYZING DETERMINANTS OF WATER-MEDIATED LIGAND RECOGNITION. Leslie A. Kuhn¹, Michael L. Raymer^1,2, William F. Punch², Paul C. Sanschagrin¹, and Erik D. Goodman³, Depts. of ¹Biochemistry, ²Computer Science, and ³Case Center for Computer-Aided Engineering and Manufacturing, Michigan State University, East Lansing, MI 48824, USA

Protein recognition of ligands, from nucleic acids to small molecule inhibitors, is usually mediated by bound water molecules bridging the protein-ligand interface. These water molecules influence both the shape and chemistry of interaction. A barrier to appropriately incorporating active-site bound water to improve molecular simulations and ligand design has been the absence of a method for determining which water sites are likely to be conserved upon ligand binding. Our Consolv technique, using a hybrid k-nearest-neighbor classifier/genetic algorithm, predicts which water molecules will mediate ligand binding by examining the structural and chemical environment of each water molecule in the free protein structure, without knowledge of the ligand. After training on 13 non-homologous proteins, Consolv correctly predicted conservation or displacement of 74.6% of the active-site water molecules in 7 new proteins. Moreover, water sites mispredicted to be conserved typically were displaced by a polar (oxygen or nitrogen) atom from the ligand. Overall accuracy for predicting conserved water or polar ligand atom binding was 89.6%. The ability to predict water-mediated interactions from the free protein structure implies that the majority of conserved active-site water binding is independent of the ligand, and that the protein micro-environment of each water molecule is the dominant influence. We are now using genetic algorithms with linear weighting and genetic programs allowing non-linear scaling to evaluate the relative importance of several features of the site - temperature factor, hydrogen-bonding capacity, local atomic packing, and atomic hydrophilicity - for determining conserved water binding.