CM-MTL-DTI: Drug-Target Interaction Prediction via Cross-Modal Alignment and Multi-Task Learning

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Drug-target interaction (DTI) prediction is vital for identifying drug candidates and understanding their mechanisms.
• Existing DTI prediction methods struggle with heterogeneous data and fine-grained dependencies between drug structures and protein sequences.
• Current approaches often use unimodal representations or simplistic fusion techniques, limiting their ability to capture complex interactions.

Purpose of the Study:

• To develop an advanced framework for accurate drug-target interaction prediction.
• To address limitations in existing methods by modeling heterogeneous and fine-grained dependencies.
• To enhance drug discovery and repurposing through improved DTI prediction.

Main Methods:

• Proposed CM-MTL-DTI, a DTI-oriented collaborative alignment framework.
• Utilized independent 1D convolutional neural networks for drug and protein sequence encoding.
• Incorporated a GIN-based graph encoder for drug structural information.
• Implemented an asymmetric bidirectional cross-modal attention mechanism.
• Introduced three collaborative objectives: cross-modal masked reconstruction (XMR), graph-sequence consistency learning (GSC), and supervised contrastive learning (SupCon).

Main Results:

• CM-MTL-DTI demonstrated stable and competitive performance on three benchmark datasets.
• The framework effectively captured direction-sensitive dependencies between drug substructures and protein residues.
• Collaborative objectives enhanced local semantic recovery, multiview alignment, and interaction representation discrimination.

Conclusions:

• The proposed CM-MTL-DTI framework significantly improves drug-target interaction prediction.
• The DTI-oriented collaborative design effectively models complex drug-protein relationships.
• This approach offers a promising tool for accelerating drug discovery and repurposing efforts.