Version 1.0, Updated December 2008.    Visits:  106.

We have systematically collected almost all currently available data of 321 S-nitrosylation sites on 221 proteins from Swiss-Prot and published papers, and developed a novel algorithm to predict S-nitrosylation sites. The integrated systemic data of S-nitrosylation and the prediction algorithm of S-nitrosylation sites together form SnoPred web server. In SnoPred, the collected S-nitrosylation sites can be searched by protein description, uniprot identifier, or sequences; moreover, potential S-nitrosylation site can be predicted based on user submitted sequence by our prediction model. The training datasets were obtained from SnoPred collected data after homolog reduction with 40% sequence identity as cutoff. Non-redundant positive and negative dataset each was consisted of 250 peptides with length of 15-mer, and was encoded to the feature vectors based on peptide fragment composition, physicochemical and biochemical properties. Nearest Neighbor Algorithm (NNA) coupled with Feature Forward Selection (FFS) were used to select 55 discriminative features. Tested by 5-fold and jackknife cross-validation, SnoPred reached 76.8% sensitivity, 80.4% specitivity, and 79.7% precision. Furthermore, a set of 25 S-nitrosylation sites identified by Han et.al were employed as independent validation data. SnoPred correctly predicted 20 sites out of 25 (80%) as S-nitrosylation targets. SnoPred may help S-nitrosylation site identification, and contribute to the development of NO biology as well.