Reassessing the modularity of gene co-expression networks using the Stochastic Block Model

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Gene co-expression networks are crucial for understanding biological insights.
• Existing community detection methods often assume a priori modular organization of genes.
• This assumption risks overlooking alternative gene interaction structures.

Purpose of the Study:

• To investigate if meaningful gene communities can be identified without pre-imposing a modular structure.
• To evaluate the modularity of discovered gene communities.
• To explore complex transcriptome organization beyond traditional modularity.

Main Methods:

• Utilized a novel community detection method: the weighted degree corrected stochastic block model (SBM).
• Applied the SBM to RNA-seq gene expression data from two Drosophila melanogaster tissues.
• The SBM does not assume assortative modules, instead organizing genes into hierarchically structured blocks.

Main Results:

• The SBM identified ten times more gene groups compared to competing methods.
• Several discovered gene groups exhibited non-modular characteristics.
• Functional enrichment analysis showed non-modular groups possess significant biological relevance, comparable to modular communities.

Conclusions:

• The transcriptome exhibits more complex organizational patterns than previously assumed.
• The assumption of modularity as the primary driver of gene co-expression network structure warrants re-evaluation.
• Novel community detection methods can uncover hidden biological structures in gene networks.