1 European Molecular Biology Laboratory Meyerhofstrasse
1, 6900 Heidelberg, Germany
2 Noyes Laboratory, School of Chemical Sciences
3 Universities of lllinois, 505 South Mathews Avenue Urbane,
Illinois 61801
The ability to predict ligand binding sites on biological macromolecules is an important goal in biotechnology. Because water plays a crucial role in the binding of ligands to proteins, we focus here on the prediction of water binding sites on proteins. We describe neural networks trained using crystallographic data to predict water sites on the basis of amino acid sequence and secondary structure. These networks make predictions an the atomic scale and surprisingly produce results comparable to those from other known methods of predicting water sites, even though the latter use tertiary structure information. The networks may be used to analyze relationships between the positions of water sites and protein sequence and secondary structure.
J. Am. Chem. Soc. (1992), 114, 8284-8285.