DTI-SNNFRA: Drug-Target interaction prediction by shared nearest neighbors and fuzzy-rough approximation
share
In-silico prediction of repurposable drugs is an effective drug discovery strategy which supplements de-nevo drug discovery from scratch. Lack of severe side effects of repurposable drugs in other applications and most importantly the reduced time frame in development suits the urgency of the situation in general. Most recent, and most advanced artificial intelligence (AI) approaches have boosted drug repurposing in terms of throughput and accuracy enormously. But, with the growing number of drugs, targets and their huge interactions produce imbalance data which may not be suitable to directly feed into the classification model. Here, we proposed DTI-SNNFRA, a framework for identifying drug-target interaction (DTI), based on shared nearest neighbor (SNN) and fuzzy-rough approximation (FRA). It uses a sampling technique to collectively reduce the huge search space covering the available drugs, large number of drugs targets and millions of interactions between them. DTI-SNNFRA operates on two stages: first it uses SNN followed by a partitional clustering for sampling the search space. Next, it computes degree of fuzzy-rough approximations and with proper degree threshold selection for undersampling of the negative samples from all possible interaction pairs between drugs and targets obtained in the first stage. Finally, classification is performed using the positive and selected negative samples. We evaluate the efficacy of DTI-SNNFRA using AUC (Area under ROC Curve), Geometric Mean, and F1 Score. The model performs extremely well with high prediction score (around 0.95). The predicted drug-target interactions are validated through a existing drug-target database (Connectivity Map (Cmap)).
READ FULL TEXT