SOC-DGL: Social Interaction Behavior Inspired Dual Graph Learning Framework for Drug-Target Interaction Identification

2025-06-02

AI Summary

Existing DTI prediction models predominantly rely on direct similarities within homogeneous graphs, overlooking rich relational information embedded in heterogeneous graphs and struggling with severe class imbalance. To address these limitations for drug repositioning, we propose a socially inspired dual-graph learning framework. First, we introduce a novel dual-module graph learning mechanism integrating affinity and balance theories. Second, we design an even-degree polynomial graph filter to efficiently capture high-order homogeneous similarities. Third, we propose a tunable imbalance-aware loss function combined with adaptive weighted negative sampling to mitigate bias from overwhelming negative instances. Evaluated on four benchmark datasets, our model achieves significant improvements in prediction accuracy—particularly under cold-start scenarios (novel drugs or targets) and highly imbalanced settings—demonstrating enhanced generalizability and robustness.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Social recommender systems and personalization for the social Web

Abstract

The identification of drug-target interactions (DTI) is crucial for drug discovery and repositioning, as it reveals potential uses of existing drugs, aiding in the acceleration of the drug development process and reducing associated costs. Despite the similarity information in DTI is important, most models are limited to mining direct similarity information within homogeneous graphs, overlooking the potential yet rich similarity information in heterogeneous graphs. Inspired by real-world social interaction behaviors, we propose SOC-DGL, which comprises two specialized modules: the Affinity-Driven Graph Learning (ADGL) module and the Equilibrium-Driven Graph Learning (EDGL) module. The ADGL module adopts a comprehensive social interaction strategy, leveraging an affinity-enhanced global drug-target graph to learn both global DTI and the individual similarity information of drugs and targets. In contrast, the EDGL module employs a higher-order social interaction strategy, amplifying the influence of even-hop neighbors through an even-polynomial graph filter grounded in balance theory, enabling the indirect mining of higher-order homogeneous information. This dual approach enables SOC-DGL to effectively and comprehensively capture similarity information across diverse interaction scales within the affinity matrices and drug-target association matrices, significantly enhancing the model's generalization capability and predictive accuracy in DTI tasks. To address the issue of imbalance in drug-target interaction datasets, this paper proposes an adjustable imbalance loss function that mitigates the impact of sample imbalance by adjusting the weight of negative samples and a parameter. Extensive experiments on four benchmark datasets demonstrate significant accuracy improvements achieved by SOC-DGL, particularly in scenarios involving data imbalance and unseen drugs or targets.

Problem

• Identifying drug-target interactions for drug discovery acceleration
• Mining similarity in heterogeneous graphs beyond homogeneous limits
• Addressing dataset imbalance with adjustable loss function

Innovation

• Dual graph learning for drug-target interaction
• Affinity and equilibrium driven modules
• Adjustable imbalance loss function