Native and modeled disulfide bonds in proteins: Knowledge-based approaches toward structure prediction of disulfide-rich polypeptides

Structure prediction and three-dimensional modeling of disulfide-rich systems are challenging due to the limited number of such folds in the structural databank. We exploit the stereochemical compatibility of substructures in known protein structures to accomodate disulfide bonds in predicting the structures of disulfide-rich polypeptides directly from disulfide connectivity pattern and amino acid sequence in the absence of structural homologs and any other structural information. This knowledge-based approach is illustrated using structure prediction of 40 nonredundant bioactive disulfide-rich polypeptides such as toxins, growth factors, and endothelins available in the structural databank. The polypeptide conformation could be predicted in 35 out of 40 nonredundant entries (87%). Nonhomologous templates could be identified and models could be obtained within 2 Angstrom deviation from the query in 29 peptides (72%). This procedure can be accessed from the World Wide Web (http://www.ncbs.res.in/-faculty/mini/dsdbase/ dsdbase.html). (C) 2005 Wiley-Liss, Inc.