Inference of gene networks using gene expression data with applications.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Gene networks (GNs) model gene interactions using graphs, often featuring sparse structures and crucial hub genes.
• Inferring accurate GNs from large-scale gene expression data is computationally challenging.

Purpose of the Study:

• To propose NetARD, a novel method using Automatic Relevance Determination (ARD) for partial correlation estimation to infer GNs with hub genes.
• To integrate NetARD into GN Co-expression Extension (GNCE) for inferring co-expressed gene networks using predefined GNs as hubs.
• To apply these methods to identify biologically relevant gene interactions and pathways.

Main Methods:

• NetARD: Utilizes ARD for partial correlation estimation to infer gene networks.
• GNCE: Integrates NetARD with predefined gene sets (e.g., transcription factors) to extend gene networks.
• Validation: Tested on simulated and in silico GNs, E. coli transcription factor data, and colorectal cancer RNA-seq data.

Main Results:

• NetARD outperforms existing methods in inferring gene networks from simulated and in silico data.
• GNCE successfully extends known gene networks, as validated with E. coli transcription factor data.
• Analysis of colorectal cancer data identified significant biological process Gene Ontology (GO) terms involved in cancer progression.

Conclusions:

• NetARD provides an effective approach for inferring gene networks and identifying hub genes from gene expression data.
• GNCE enhances gene network inference by leveraging prior biological knowledge.
• These methods have practical applications in identifying disease-related pathways and biological processes.