Graph Diffusion Network for Drug-Gene Prediction

2025-02-13

AI Summary

Drug–gene association prediction faces dual challenges of data sparsity and low efficiency in contrastive learning. To address these, we propose an end-to-end graph diffusion framework that introduces, for the first time, a meta-path-driven homogeneous graph learning mechanism—transforming the heterogeneous drug–gene–disease network into a semantically consistent homogeneous graph. We further design a parallel diffusion network to dynamically generate high-quality hard negative samples, eliminating inefficient exhaustive negative sampling. Our method integrates graph neural networks, self-supervised contrastive learning, and differentiable graph diffusion. Evaluated on DGIdb 4.0, it significantly outperforms state-of-the-art methods (achieving a +4.2% average AUC gain) and demonstrates strong generalization capability. It effectively models ternary heterogeneous relationships and supports cold-start prediction.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Predicting drug-gene associations is crucial for drug development and disease treatment. While graph neural networks (GNN) have shown effectiveness in this task, they face challenges with data sparsity and efficient contrastive learning implementation. We introduce a graph diffusion network for drug-gene prediction (GDNDGP), a framework that addresses these limitations through two key innovations. First, it employs meta-path-based homogeneous graph learning to capture drug-drug and gene-gene relationships, ensuring similar entities share embedding spaces. Second, it incorporates a parallel diffusion network that generates hard negative samples during training, eliminating the need for exhaustive negative sample retrieval. Our model achieves superior performance on the DGIdb 4.0 dataset and demonstrates strong generalization capability on tripartite drug-gene-disease networks. Results show significant improvements over existing methods in drug-gene prediction tasks, particularly in handling complex heterogeneous relationships. The source code is publicly available at https://github.com/csjywu1/GDNDGP.

Problem

• Predicts drug-gene associations for drug development
• Improves data sparsity in graph neural networks
• Enhances contrastive learning with hard negative samples

Innovation

• meta-path-based homogeneous graph learning
• parallel diffusion network for hard negatives
• improved drug-gene prediction accuracy