Drug-target interaction prediction via chemogenomic space: learning-based methods.

Area of Science:

• Genomic drug discovery
• Computational biology
• Pharmacology

Background:

• Identifying drug-target interactions is vital for genomic drug discovery.
• In silico prediction offers a cost-effective alternative to experimental methods.
• Computational approaches enable novel drug-target pair analysis.

Purpose of the Study:

• To review chemogenomic methods for predicting drug-target interactions.
• To categorize learning-based prediction frameworks.
• To examine supervised learning approaches in detail.

Main Methods:

• Focus on chemogenomic methods.
• Classification of learning-based methods into supervised and semi-supervised.
• Detailed study of supervised methods: similarity-based and feature-based.

Main Results:

• Learning-based methods provide a framework for drug-target interaction prediction.
• Supervised learning encompasses similarity and feature-based techniques.
• These methods contribute to advancements in pharmacology.

Conclusions:

• Learning-based methods have improved drug-target interaction prediction.
• Oversimplification in predictive models can lead to overly optimistic results.
• Further refinement of computational models is necessary for accurate predictions.