TargetNet: a web service for predicting potential drug-target interaction profiling via multi-target SAR models

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Drug-target interactions (DTIs) are crucial for drug discovery and public health.
• Accurate DTI profiling is essential for predicting drug indications, adverse effects, and mechanisms of action.
• Genome-scale DTI prediction is highly desirable.

Purpose of the Study:

• To develop TargetNet, a web server for real-time DTI predictions based on molecular structures.
• To provide a tool for predicting drug activity across 623 human proteins.
• To demonstrate the utility of DTI profiling in various applications.

Main Methods:

• Utilized Naïve Bayes models and molecular fingerprints for DTI prediction.
• Employed ensemble learning to enhance prediction accuracy.
• Developed 623 structure-activity relationship (SAR) models for human proteins.
• Validated models using multiple strategies, achieving AUC scores of 75-100%.

Main Results:

• TargetNet successfully predicts DTI profiles from molecular structures.
• The generated DTI profiles were applied to predict potential targets, toxicity, and drug-drug interactions.
• The web server provides real-time predictions across 623 human proteins.

Conclusions:

• TargetNet offers a valuable tool for predicting DTI profiles, aiding in drug discovery and development.
• The DTI profiling generated by TargetNet has broad applications in pharmacology and toxicology.
• The web server is freely accessible and built on a robust framework.