EDC-DTI: An end-to-end deep collaborative learning model based on multiple information for drug-target interactions...

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Drug-target interactions (DTIs) are crucial for drug development.
• Deep learning models have significantly advanced DTIs prediction, offering cost and time savings.
• Existing methods face challenges with data complexity and computational efficiency.

Purpose of the Study:

• To develop an end-to-end deep collaborative learning model (EDC-DTI) for accurate DTIs prediction.
• To identify novel therapeutic targets for existing drugs by leveraging diverse drug-target information.
• To enhance the efficiency and reduce the cost of the drug discovery pipeline.

Main Methods:

• Developed an end-to-end deep collaborative learning framework (EDC-DTI).
• Employed a feature builder with algorithms for molecular and network topology properties.
• Utilized a classifier with an improved graph attention network-based feature encoder for heterogeneous information integration and a neural network-based feature decoder for prediction.

Main Results:

• EDC-DTI demonstrated superior predictive performance compared to six baseline models.
• The model achieved high accuracy with low computational requirements.
• Robustness tests confirmed strong performance on sparse datasets.
• Case studies successfully predicted validated drug-target interactions for Simvastatin, Nifedipine, and Afatinib.

Conclusions:

• EDC-DTI offers a powerful and efficient approach for predicting drug-target interactions.
• The model's ability to integrate heterogeneous information enhances prediction accuracy.
• EDC-DTI holds significant potential for accelerating drug discovery and development by identifying novel drug-target relationships.