Enhancing gene regulatory networks inference through hub-based data integration

Area of Science:

• Computational biology
• Bioinformatics
• Systems biology

Background:

• Gene Regulatory Network (GRN) reconstruction is crucial for understanding cellular responses.
• Current methods often rely solely on gene expression data, limiting accuracy.
• Integrating diverse data types is essential for more precise GRN models.

Purpose of the Study:

• To enhance GRN inference by integrating network data and structural prior knowledge.
• To develop a diffusion-based method that improves the accuracy and scalability of GRN reconstruction.
• To identify optimal centrality measures for detecting hub nodes in biological networks.

Main Methods:

• An enhanced diffusion-based method integrating network and structural data.
• Application of the Random Walk with Restart (RWR) algorithm with emphasis on hub nodes.
• Joint optimization of low-dimensional node feature vectors using diffusion component analysis.
• Evaluation of fourteen centrality measures for hub node detection.

Main Results:

• The proposed method achieved 0.02-0.08 AUROC improvement over gene expression data alone in yeast and E. coli.
• Successfully inferred the gene regulatory network for esophageal cancer data.
• Demonstrated substantial improvements in accuracy and scalability for GRN inference.

Conclusions:

• The enhanced diffusion-based framework significantly improves GRN inference accuracy and scalability.
• Fused features and identified centrality measures offer functional insights into genes and proteins.
• The method provides a general framework for integrating and analyzing network and structural data across scientific disciplines.