Deep learning for drug-drug interaction prediction: A comprehensive review

Area of Science:

• Pharmacology and Computational Chemistry
• Drug Safety and Bioinformatics

Background:

• Predicting drug-drug interactions (DDIs) is vital for drug safety and understanding combination therapy mechanisms.
• Experimental DDI prediction methods are slow, costly, and limited in scale.
• There is a critical need for efficient computational approaches to DDI detection.

Purpose of the Study:

• To review and categorize recent high-quality deep learning-based methods for DDI prediction.
• To analyze the strengths and limitations of current deep learning models in DDI prediction.
• To discuss future research directions and potential advancements in computational DDI prediction.

Main Methods:

• Categorization of deep learning methods into four groups: neural networks, graph neural networks, knowledge graphs, and multimodal approaches.
• Review of literature focusing on advanced computational models for DDI prediction.
• Comparative analysis of deep learning versus traditional machine learning for DDI prediction tasks.

Main Results:

• Deep learning models demonstrate significantly improved performance in DDI prediction compared to traditional machine learning techniques.
• Deep learning approaches offer enhanced scalability for large datasets.
• These models effectively integrate diverse data types, leading to more accurate and efficient DDI predictions.

Conclusions:

• Deep learning represents a substantial advancement in computational DDI prediction, offering superior efficiency and accuracy.
• The ability of deep learning to handle large-scale, multimodal data is key to its success in drug safety research.
• Further development in deep learning holds significant promise for revolutionizing DDI prediction and drug development.