Deep Learning in Drug Target Interaction Prediction: Current and Future Perspectives.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug-target interactions (DTIs) prediction is vital for efficient drug discovery.
• Experimental methods for DTIs are resource-intensive; computational approaches offer an alternative.
• Machine learning, particularly deep learning, shows significant promise for DTI prediction.

Purpose of the Study:

• To provide a comprehensive review of deep learning-based methods for DTI prediction.
• To analyze various deep network architectures used for feature extraction from drug and protein sequences.
• To compare the advantages and limitations of different DTI prediction approaches.

Main Methods:

• Systematic investigation of existing deep learning models for DTI prediction.
• Analysis of feature extraction techniques using deep neural networks on drug compounds and protein sequences.
• Exploration of feature descriptor combination strategies and commonly used DTI datasets.

Main Results:

• Identified and compared diverse deep network architectures employed in DTI prediction.
• Evaluated methods for integrating drug and protein features.
• Cataloged prevalent datasets utilized in the field.

Conclusions:

• Deep learning offers powerful tools for advancing DTI prediction in drug discovery.
• Understanding architectural choices and feature representations is key to improving prediction accuracy.
• Further research is needed to address current challenges and explore future directions in deep learning for DTI prediction.