Deep Learning for Drug-Target Interaction Prediction: A Comprehensive Review.

Area of Science:

• Computational Biology
• Pharmacology
• Artificial Intelligence

Background:

• Drug-target interaction (DTI) prediction is crucial for drug discovery but traditionally relies on slow experimental methods.
• Deep learning (DL) offers a powerful and efficient alternative for predicting DTIs, accelerating the discovery pipeline.

Purpose of the Study:

• To provide a structured overview of deep learning-based methods for drug-target interaction prediction.
• To synthesize current advances, challenges, and future research directions in DL for DTI prediction.

Main Methods:

• Review of feature representation strategies for drugs and proteins.
• Analysis of various DL architectures: DNNs, RNNs, CNNs, GNNs, and Transformers.
• Summary of common datasets and evaluation metrics for DTI prediction.

Main Results:

• DL models demonstrate significant efficiency and power in predicting DTIs compared to traditional methods.
• Exploration of DL applications in drug repositioning, drug design, and precision medicine.
• Identification of key challenges including data scarcity and model interpretability.

Conclusions:

• Deep learning is a transformative approach for DTI prediction, essential for modern drug discovery.
• Future research should focus on self-supervised learning and explainable AI to overcome current limitations.
• This review provides a comprehensive resource to guide future developments in DL-based DTI prediction.