A Bayesian Approach for Learning Gene Networks Underlying Disease Severity in COPD

Area of Science:

• Computational biology
• Systems biology
• Statistical genetics

Background:

• Understanding complex diseases like chronic obstructive pulmonary disease (COPD) requires analyzing gene interactions across different disease severities.
• Existing network inference methods may struggle to identify both shared and differential biological pathways across multiple patient groups.

Purpose of the Study:

• To develop a Bayesian hierarchical model for inferring network structures with shared and differential edges across multiple sample groups.
• To apply this model to understand gene pathway breakdown in chronic obstructive pulmonary disease (COPD) progression.
• To identify critical genes and disrupted connections associated with COPD severity.

Main Methods:

• A Bayesian hierarchical approach linking graphs via a Markov random field prior for network similarity.
• Incorporation of continuous shrinkage priors for computational efficiency and scalability in high-dimensional network estimation.
• Application to patient groups stratified by COPD severity.

Main Results:

• Identification of critical hub genes within four targeted pathways relevant to COPD.
• Detection of gene connections that are altered with increasing disease severity, characterizing disease evolution.
• Demonstration of superior performance compared to competing methods using simulated data.

Conclusions:

• The proposed Bayesian method effectively infers network structures and identifies key biological insights in multi-group settings.
• This approach advances the understanding of gene pathway dynamics in disease progression, exemplified by COPD.
• The model offers a scalable and efficient tool for network inference in complex biological systems.