NJGCG: A node-based joint Gaussian copula graphical model for gene networks inference across multiple states

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Gene interaction networks are crucial for understanding biological processes but change with environmental and state variations.
• Existing computational methods often infer single gene networks, neglecting similarities across related states and the role of hub genes.

Purpose of the Study:

• To develop a novel computational model for jointly inferring multiple gene networks from heterogeneous gene expression data.
• To improve the accuracy of gene network inference by considering similarities between networks and identifying common/specific hub genes.

Main Methods:

• Proposed a node-based joint Gaussian copula graphical (NJGCG) model for inferring multiple gene networks.
• Incorporated a tree-structured group lasso penalty to identify common and specific hub genes.
• The model handles gene expression data with missing values.

Main Results:

• Simulation studies demonstrated that the NJGCG model outperforms existing methods in inferring gene networks.
• Applied the NJGCG model to mouse embryonic stem cell differentiation and breast cancer subtypes, revealing dynamic network changes.
• Identified key common and specific hub genes relevant to stem cell differentiation and breast cancer heterogeneity.

Conclusions:

• The NJGCG model provides a robust framework for joint inference of multiple gene networks, enhancing biological insights.
• The identification of common and specific hub genes offers a deeper understanding of biological state transitions and disease heterogeneity.