Bayesian Structure Learning in Multi-layered Genomic Networks.

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Integrative network modeling of multi-platform genomic data offers a holistic view of biological systems.
• Understanding information flow across cancer types is crucial for identifying therapeutic targets.

Purpose of the Study:

• To propose a novel multi-layered Gaussian graphical model (mlGGM) for analyzing complex, multi-level genomic networks in human cancers.
• To develop and validate a Bayesian node-wise selection (BANS) approach for sparse and interpretable model selection in mlGGM.

Main Methods:

• Implemented a Bayesian node-wise selection (BANS) approach for variable selection in mlGGM.
• Utilized edge-specific prior knowledge for flexible and interpretable model construction.
• Validated the BANS approach using simulated data against existing multivariate regression methods.

Main Results:

• The BANS approach demonstrated superior performance compared to existing multivariate regression-based methodologies.
• Integrative genomic network analysis revealed commonalities and differences in p53 networks across multiple cancer types.
• Identified epigenetic effects of BRCA2 on p53 and its interaction with phosphorylated CHK2 (T68).

Conclusions:

• The proposed mlGGM and BANS approach provide a powerful framework for integrative genomic network analysis in cancer.
• Findings highlight potential translational utilities for identifying biomarkers and therapeutic targets, particularly concerning p53 and BRCA2 pathways.