Computational Discovery of Putative Leads for Drug Repositioning Through Drug-Target Interaction Prediction

The emergence of multi-resistant bacterial strains and the existing void in the discovery and development of new classes of antibiotics is a growing concern, as some bacterial strains are now resistant to last-line antibiotics and considered untreatable. A growing trend in drug screening for the past decade is drug repositioning, which consists in focusing on one of the undesired effects of an already commercialized drug in an attempt to make it the main effect. While this was formerly performed experimentally, computational methods speed and reduce the associated costs of drug and drug-target screening.
Thus, we present a computational pipeline that enables the discovery of putative leads for drug repositioning that can be applied to any microbial proteome. Putative drug-targets are inferred by calculating network metrics for the interactome of the bacterial organism. Prediction of drug-target interactions (DTI) is performed using a random forest trained with high-quality publicly available data. Classifier performance achieved an area under the ROC curve of 0.91 for classification of out-of-sampling data. A drug-target network was created by combining 3,081 unique ligands and the expected ten best drug targets. This network was used to predict new DTIs and to calculate the probability of the positive class, allowing the scoring of the predicted instances.

Available data (download ZIP)

– classifier.py (classifier code: uses Yamanishi’s data for training and DrugBank data for external validation)
– yamanishi_DTIs_REAL_NEGS.txt (Training data set: Protein ID; Drug ID; True Label; + 702 features)
– drugbank_DTIs_REAL_NEGS.txt (External validation data set: Protein ID; Drug ID; True Label; + 702 features)
– test_data_sc_and_bc.txt (Test data set: Protein ID; Drug ID; True Label; + 702 features)