Hierarchical Graph Transformer With Contrastive Learning for Gene Regulatory Network Inference

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Gene regulatory networks (GRNs) are essential for understanding cellular processes and gene regulation.
• Computational models, particularly graph neural networks (GNNs), are increasingly used for GRN inference due to high-throughput sequencing data.
• Existing GNN methods struggle to capture long-distance structural semantic correlations in GRNs.

Purpose of the Study:

• To introduce a novel GNN-based model, HGTCGRN, for enhanced GRN inference.
• To address the limitations of existing GNNs in capturing long-distance interactions.
• To improve the accuracy and efficiency of computational GRN inference.

Main Methods:

• Developed Hierarchical Graph Transformer with Contrastive Learning for GRN (HGTCGRN) inference.
• Incorporated gene family nodes as virtual nodes to capture gene function semantics.
• Utilized gene ontology information for contrastive learning optimization of GRNs.

Main Results:

• HGTCGRN demonstrated superior performance in GRN inference compared to existing methods.
• The hierarchical graph Transformer effectively captured structural semantics and long-distance correlations.
• Contrastive learning enhanced the optimization and accuracy of the GRN inference.

Conclusions:

• HGTCGRN offers a significant advancement in computational GRN inference.
• The model's ability to capture hierarchical and semantic information improves biological insights.
• This approach provides a more accurate and cost-effective alternative to experimental methods for GRN discovery.