SBM

SBM infers modular structure and edge confidence in correlation-based biomolecular networks using stochastic block models (SBMs) to address noise in high-throughput transcriptomics, proteomics, and metabolomics data.

Key Features:

• Module Identification: Identifies groups of nodes (modules) with similar topological properties in correlation-based biomolecular networks.
• Edge Confidence Scores: Computes edge confidence scores from global network characteristics and potential hierarchical structure to assess interaction reliability.
• Hierarchical SBM Representation: Implements hierarchical stochastic block models to capture nested and modular relationships within networks.
• Network Reduction: Reduces networks by thresholding correlations for significance or by enforcing scale-freeness to mitigate noise.
• Multi-omics Applicability: Applies SBMs independently to correlation-based networks derived from transcriptomics, proteomics, and metabolomics.

Scientific Applications:

• Comparative Network Analysis: Facilitates comparisons between networks under different conditions using modules and edge confidence scores to identify differential molecular mechanisms.
• Functional Annotation: Supports functional interpretation by relating predicted blocks/modules to biological and phenotypic annotations.
• Breast Cancer Multi-omics Analysis: Has been applied to correlation-based networks from breast cancer datasets across multiple omics layers.

Methodology:

Fits stochastic block models to correlation-based networks derived from high-throughput transcriptomics, proteomics, or metabolomics measurements; reduces networks by correlation-significance thresholding or enforcing scale-freeness; employs hierarchical SBMs and derives edge confidence scores from global network characteristics and hierarchical structure, with independent application to three types of correlation-based networks from breast cancer datasets.