MVR-DTI: A Multimodal Molecular Visual Representation Learning for Drug-Target Interaction Prediction

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Drug-target interaction (DTI) prediction is crucial for drug discovery.
• Existing methods struggle with multimodal molecular representations.
• Molecular visual representation learning for DTI is underexplored.

Purpose of the Study:

• To develop a novel method for DTI prediction using multimodal molecular visual representations.
• To address limitations in current DTI prediction techniques regarding multimodal data integration.

Main Methods:

• Proposed Multimodal Molecular Visual Representation for DTI prediction (MVR-DTI).
• Utilized a vision transformer for structure-aware visual feature extraction.
• Integrated visual features with molecular descriptors, protein sequences, and knowledge graph embeddings via contrastive learning and attention mechanisms.

Main Results:

• MVR-DTI demonstrated superior performance compared to existing baseline methods.
• Consistent improvements were observed across multiple evaluation metrics.
• The method effectively captures spatial molecular information and integrates diverse data modalities.

Conclusions:

• Multimodal visual representation learning holds significant potential for enhancing DTI prediction.
• MVR-DTI offers a promising approach for advancing computational drug discovery.
• The integration of visual and other molecular data modalities is key to improving prediction accuracy.