Inferring gene regulatory networks from single-cell transcriptomics based on graph embedding.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) are crucial for understanding biological processes.
• Inferring GRNs from single-cell RNA sequencing (scRNA-seq) data presents challenges due to the dynamic and complex nature of gene regulation.
• Existing methods often overlook global structures and distal regulatory relationships.

Purpose of the Study:

• To develop a novel supervised deep learning framework, IGEGRNS, for accurate GRN inference from scRNA-seq data.
• To address limitations of existing methods by incorporating global network structure and distal regulatory information.
• To enhance the understanding of complex biological processes through improved GRN reconstruction.

Main Methods:

• Developed IGEGRNS, a supervised deep learning framework utilizing graph embedding.
• Employed GraphSAGE to capture contextual gene information by aggregating features and neighborhood structures into low-dimensional embeddings.
• Utilized Top-k pooling to identify influential nodes and Convolutional Neural Networks (CNNs) for predicting regulatory relationships.

Main Results:

• The IGEGRNS framework demonstrated superior performance compared to nine other supervised and unsupervised methods.
• Achieved enhanced accuracy in inferring gene regulatory networks across six time-series scRNA-seq datasets.
• Successfully captured global network structures and distal regulatory relationships, improving upon existing approaches.

Conclusions:

• IGEGRNS provides a robust and effective method for inferring gene regulatory networks from scRNA-seq data.
• The framework's ability to integrate global network topology and distal interactions offers significant advantages.
• This advancement facilitates a deeper understanding of gene regulation and complex biological systems.