Comprehensive Review of Drug-Drug Interaction Prediction Based on Machine Learning: Current Status, Challenges, and...

Area of Science:

• Pharmacology
• Computational Biology
• Artificial Intelligence

Background:

• Drug-drug interactions (DDIs) are critical in drug development and administration.
• Experimental DDI detection methods have limitations, driving the need for alternative approaches.
• Machine learning (ML) has emerged as a powerful tool for DDI prediction.

Approach:

• Systematic review of ML-based DDI prediction studies.
• Analysis of classic DDI databases (drugs, side effects, interaction information).
• Evaluation of drug attributes (chemical, biological, phenotypic) for ML models.
• Categorization and comparison of ML approaches (shallow, deep, recommender systems, knowledge graphs).

Key Points:

• DDI databases provide foundational data for ML models.
• Diverse drug attributes enhance the accuracy of DDI prediction.
• Various ML methods offer different strengths for DDI detection.
• Knowledge graph-based and deep learning methods show significant promise.

Conclusions:

• ML-based DDI prediction is a rapidly advancing field.
• Future research should address data integration, model interpretability, and real-world validation.
• Opportunities exist in developing more robust and generalizable DDI prediction systems.