MotifGT-DTI: Pivotal Motif-Based Graph Transformer Model Improves Drug-Target Interaction Prediction.

Area of Science:

• Computational Biology
• Drug Discovery
• Bioinformatics

Background:

• Drug-target interactions (DTIs) are crucial for drug discovery and repurposing.
• Existing DTI prediction methods often lack interpretability and underutilize molecular structural information.

Purpose of the Study:

• To develop an interpretable DTI prediction model that effectively utilizes both drug and protein molecular structures.
• To improve the accuracy and generalization of DTI prediction, especially in cold-start scenarios.

Main Methods:

• Proposed MotifGT-DTI, a novel motif-based model employing a graph transformer (GT).
• Captured complex molecular patterns using drug molecular graph motifs and protein 3-D pocket subgraphs via GT.
• Fused 1-D sequence and 3-D structure protein features using cross-attention.
• Connected drug-protein structural associations with a bilinear attention network.

Main Results:

• MotifGT-DTI achieved superior accuracy compared to state-of-the-art baselines across four public datasets.
• Demonstrated competitive performance in accuracy, generalization, and stability across three cold-start scenarios.
• Successfully identified functional molecular motifs and provided interpretable predictions through visualization.

Conclusions:

• MotifGT-DTI represents a significant advancement in interpretable DTI prediction.
• The model shows strong potential for practical applications in drug discovery and repurposing.
• The approach effectively leverages molecular structure for enhanced DTI prediction accuracy and interpretability.