SSGraphDTI: A Drug-Target Interaction Prediction Method Integrated Structural and Dynamic Systemic Biology...

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Drug-Target Interaction (DTI) prediction is vital for pharmaceutical development but faces challenges with expensive experiments and unreliable computational methods.
• Integrating drug molecular networks offers valuable information for improving DTI prediction accuracy.

Purpose of the Study:

• To propose and evaluate SSGraphDTI, a novel model that integrates structural and systemic biological attributes for enhanced DTI prediction.
• To address the limitations of current DTI prediction methods, including the
• cold-start problem
• and performance with limited data.

Main Methods:

• Extracted drug structural attributes using convolutional neural networks from Simplified Molecular Input Line Entry System (SMILES).
• Extracted target structural attributes from amino acid sequences using convolutional neural networks.
• Obtained systemic biological attributes via graph representation learning on a dynamically constructed heterogeneous drug-target interaction network.

Main Results:

• SSGraphDTI achieved approximately 1.0% improvement across five metrics on the Dataset_DrugBank compared to recent methods.
• The model effectively addresses the "cold-start problem" by utilizing solely structural data.
• Maintained strong predictive performance even with limited data by extracting systemic attributes from DTI networks.

Conclusions:

• Integrating both structural and systemic information significantly enhances DTI prediction accuracy.
• SSGraphDTI offers a promising approach for efficient and accurate drug discovery, particularly in scenarios with limited interaction data.