A Drug-Side Effect Context-Sensitive Network approach for drug target prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Pharmacology

Background:

• Computational drug target prediction is crucial for drug discovery.
• Network-based approaches are common for drug-target interaction (DTI) prediction.
• Existing methods struggle to capture contextual information in drug, gene, and disease connections.

Purpose of the Study:

• To propose a novel context-sensitive network (CSN) model for enhanced DTI prediction.
• To model contextual drug phenotypic relationships for improved accuracy.
• To evaluate the CSN model's performance against traditional methods.

Main Methods:

• Constructed a Drug-Side Effect Context-Sensitive Network (DSE-CSN) and a protein-protein interaction network (PPIN).
• Integrated DSE-CSN and PPIN into a heterogeneous network using known DTIs.
• Employed a network-based ranking algorithm for predicting and prioritizing genetic targets.

Main Results:

• The CSN-based model achieved a high area under the ROC curve (0.95) in de novo cross-validation.
• Known DTIs were ranked highly, with an average rank in the top 3.2% in leave-one-out cross-validation.
• The CSN model demonstrated superior performance (higher MAP) compared to similarity-based network models, with further improvements through differential weighting.

Conclusions:

• The CSN model effectively predicts drug targets by leveraging context-specific inter-relationships among drugs and side effects.
• This approach shows significant potential for advancing drug discovery and target identification.
• The CSN model offers a more robust and accurate alternative to traditional DTI prediction methods.