Introduction & Further Information
PSIVER (Protein-protein interaction SItes prediction seVER) a server for predicting protein-protein interation sites in protein seqeunces, using only sequence features (position specific scoring matrix and predicted accessibility) are using a Naïve Bayes classifier (NBC) and a kernel density estimation method (KDE). The leave-one out cross validation of PSIVER achieved a Matthews Correlation Coefficient (MCC) of 0.151 and an F-measure of 35.3% on a non-redundant set of 186 protein sequences extracted from 105 hetero dimers in the Protein Data Bank (consisting of 36,219 residues, of which 15.2% were known interface residues). Even though the dataset used for training was highly imbalanced, a randomization test demonstrated that the proposed method managed to avoid overfitting. PSIVER was also tested on 72 sequences not used in training (consisting of 18,140 residues, of which 10.6% were known interface residues), and achieved an MCC of 0.135 and an F-measure of 31.5%. PSIVER enables experimentalists to identify potential interface residues from protein sequences alone and to mutate targeted residues selectively in order to unravel protein functions.
Murakami, Y. and Mizuguchi, K. (2010). "Applying the Naive Bayes classifier with kernel density estimation to the prediction of protein-protein interaciton sites", Bioinformatics, doi:10.1093/bioinformatics/btq302, Abstract / PDF
© PSIVER Copyright
PSIVER is maintained by Yoichi MURAKAM @ Bioinformatics Project, NIBIO