Analysis of correlation-based biomolecular networks from different omics data by fitting stochastic block models

Area of Science:

• Systems Biology
• Bioinformatics
• Network Analysis

Background:

• Biomolecular networks comprise interacting genes, proteins, and metabolites crucial for biological functions.
• Network modules represent groups of nodes with similar topological properties, indicating key biological processes.
• Analyzing differential molecular mechanisms requires comparing molecular networks across conditions, but data noise complicates edge relevance.

Purpose of the Study:

• To investigate the utility of stochastic block models (SBMs) for analyzing biomolecular networks.
• To assess SBMs' capability in deriving both network modules and reliable edge confidence scores.
• To improve the analysis of noisy biological networks by incorporating global network characteristics.

Main Methods:

• Applied stochastic block models (SBMs) to correlation-based networks from breast cancer transcriptomics, proteomics, and metabolomics data.
• Utilized hierarchical SBMs to represent network structures and derive edge confidence scores.
• Networks were pre-processed using correlation significance thresholding and scale-freeness requirements.

Main Results:

• Hierarchical SBMs provided the best representation for the analyzed biomolecular networks.
• Identified modules showed significant biological and phenotypic functional annotations.
• Derived edge confidence scores generally aligned with existing biological evidence.

Conclusions:

• Stochastic block models are suitable for representing and analyzing biomolecular networks.
• SBM-derived edge confidence scores, based on global network topology and hierarchies, offer valuable complementary data for network comparisons.
• This method enhances the reliability of edge relevance estimation in noisy biological data.