Motif-Based Hypergraph Representation Learning: Transductive and Inductive Inference for Gene Regulatory Networks.

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) control gene expression through complex interactions.
• Existing graph representation learning methods for GRNs often overlook higher-order regulatory patterns present in network motifs.
• This limitation hinders accurate inference of gene regulatory relationships.

Purpose of the Study:

• To develop a novel framework, Motif-GRN, for enhanced GRN modeling.
• To capture higher-order regulatory patterns using motif-based hypergraph representation learning.
• To improve the accuracy of gene regulatory inference.

Main Methods:

• Identification of statistically significant regulatory motifs to construct a multichannel motif-induced hypergraph.
• Design of a motif-aware hypergraph convolutional network for motif-centric feature extraction.
• Integration of cross-view contrastive learning to align representations and enhance gene embeddings.
• Development of an inductive extension for cross-dataset generalization.

Main Results:

• Motif-GRN effectively captures higher-order semantic structures in GRNs.
• The framework outperforms state-of-the-art methods in both transductive and inductive GRN inference tasks.
• Experiments on multiple datasets across different cell types validate the model's performance.

Conclusions:

• Motif-GRN provides a powerful approach for modeling higher-order regulatory patterns in gene networks.
• The proposed method enhances the accuracy and generalizability of GRN inference.
• This work offers significant potential for advancing our understanding of complex gene regulation.