TriDTI: tri-modal representation learning with cross-modal alignment for drug-target interaction prediction

Area of Science:

• Computational chemistry
• Bioinformatics
• Artificial intelligence in drug discovery

Background:

• Drug-target interaction (DTI) prediction is crucial for efficient drug screening and discovery.
• Existing methods struggle to integrate more than three modalities due to information loss.
• Pharmacological multimodal information offers potential for enhanced DTI prediction accuracy.

Purpose of the Study:

• To propose TriDTI, a novel framework for drug-target interaction prediction using three modalities.
• To overcome information loss in multimodal integration for improved DTI prediction.
• To enhance the accuracy and generalization capabilities of DTI prediction models.

Main Methods:

• TriDTI integrates structural, sequential, and relational modalities for both drugs and proteins.
• Projection and cross-modal contrastive learning are employed for modality alignment.
• A fusion strategy combining soft attention and cross-attention is used for multimodal representation integration.

Main Results:

• TriDTI significantly outperforms existing state-of-the-art approaches on three benchmark datasets.
• The framework demonstrates robust generalization in three challenging cold-start scenarios.
• Effective prediction of interactions involving novel drugs, targets, and bindings was achieved.

Conclusions:

• TriDTI presents a robust and practical framework for advancing drug discovery.
• The multimodal integration strategy effectively mitigates information loss.
• The proposed method shows significant potential for facilitating the identification of new therapeutics.