In silico prediction of the new drug-target interactions from existing databases is of important value for the drug discovery
process. Currently, the amount of protein targets that have been identified experimentally is still very small compared with the entire human
proteins. In order to predict protein-ligand interactions in an accurate manner, we have developed a support vector machine (SVM)
model based on the chemical-protein interactions from STITCH. New features from ligand chemical space and interaction networks have
been selected and encoded as the feature vectors for SVM analysis. Both the 5-fold cross validation and independent test show high predictive
accuracy that outperforms the state-of-the-art method based on ligand similarity. Moreover, 91 distinct pairs of features have been
selected to rebuild a simplifier model, which still maintains the same performance as that based on all 332 features. Then, this refined
model is used to search for the potential D-amino acid oxidase inhibitors from STITCH database and the predicted results are finally
validated by our wet experiments. Out of 10 candidates obtained, seven D-amino acid oxidase inhibitors have been verified, in which
four are newly found in the present study, and one may have a new application in therapy of psychiatric disorders other than being an antineoplastic
agent. Clearly, our model is capable of predicting potential new drugs or targets on a large scale with high efficiency.
Keywords: Protein-ligand interactions, prediction, chemical preference features, SVM, feature selection, fingerprint, MACCS key, D-amino
acid oxidase inhibitors.
Rights & PermissionsPrintExport