MvGraphDTA: multi-view-based graph deep model for drug-target affinity prediction by introducing the graphs and line...

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Accurate drug-target affinity (DTA) identification is crucial for efficient drug screening, design, and repurposing.
• Computational methods for DTA prediction can significantly reduce experimental costs and accelerate drug development.
• Current computational DTA identification methods face challenges in achieving high accuracy.

Purpose of the Study:

• To develop a novel multi-view graph deep learning model for enhanced drug-target affinity prediction.
• To improve the accuracy and reliability of computational drug-target interaction prediction.

Main Methods:

• Proposed MvGraphDTA, a multi-view graph deep model utilizing graph convolutional networks (GCNs).
• Extracted structural features from original drug and target graphs using GCNs.
• Constructed line graphs and applied GCNs to extract relationship features.
• Fused multi-view features from original and line graphs for enhanced complementarity.
• Utilized a fully connected network for final DTA prediction.

Main Results:

• MvGraphDTA demonstrated superior performance compared to state-of-the-art methods on benchmark DTA prediction datasets.
• Data augmentation was applied to training sets to improve model robustness.
• The model achieved high accuracy in predicting drug-target interactions.

Conclusions:

• MvGraphDTA exhibits excellent universality and generalization capabilities on additional datasets.
• The model proves to be a reliable tool for drug-target interaction prediction.
• The multi-view approach enhances the predictive power for DTA identification.