CDI-DTI: A Strong Cross-domain Interpretable Drug-Target Interaction Prediction Framework Based on Multi-Strategy Fusion

2025-10-22

AI Summary

Existing drug–target interaction (DTI) prediction methods suffer from limited cross-domain generalization, poor cold-start performance, and insufficient interpretability. To address these challenges, we propose DTI-MCA, a multimodal fusion framework that introduces novel multi-source cross-attention and bidirectional cross-attention mechanisms to enable fine-grained alignment of textual, structural, and functional features for drugs and targets. We further incorporate Gram Loss regularization and a deep orthogonal fusion module to suppress feature redundancy and enhance biological interpretability. DTI-MCA is trained end-to-end and achieves state-of-the-art performance across multiple benchmark datasets: it improves cross-domain AUC by 3.2–5.8%, cold-start F1-score by 4.7–6.3%, and provides biologically meaningful attention weights that highlight critical pharmacophores and target residues. The framework thus delivers high accuracy, strong generalization capability, and intrinsic interpretability—making it well-suited for real-world drug discovery applications.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Accurate prediction of drug-target interactions (DTI) is pivotal for drug discovery, yet existing methods often fail to address challenges like cross-domain generalization, cold-start prediction, and interpretability. In this work, we propose CDI-DTI, a novel cross-domain interpretable framework for DTI prediction, designed to overcome these limitations. By integrating multi-modal features-textual, structural, and functional-through a multi-strategy fusion approach, CDI-DTI ensures robust performance across different domains and in cold-start scenarios. A multi-source cross-attention mechanism is introduced to align and fuse features early, while a bidirectional cross-attention layer captures fine-grained intra-modal drug-target interactions. To enhance model interpretability, we incorporate Gram Loss for feature alignment and a deep orthogonal fusion module to eliminate redundancy. Experimental results on several benchmark datasets demonstrate that CDI-DTI significantly outperforms existing methods, particularly in cross-domain and cold-start tasks, while maintaining high interpretability for practical applications in drug-target interaction prediction.

Problem

• Addresses cross-domain generalization in drug-target interaction prediction
• Solves cold-start prediction challenges for new drug-target pairs
• Enhances model interpretability through feature alignment and fusion

Innovation

• Multi-strategy fusion integrates multimodal drug-target features
• Multi-source cross-attention aligns and captures fine-grained interactions
• Gram loss and orthogonal fusion enhance interpretability and reduce redundancy