Disentangling factors affecting bacterial transcriptional regulatory network inference

Area of Science:

• Microbiology
• Computational Biology
• Systems Biology

Background:

• Bacterial gene expression databases have grown rapidly, enabling computational inference of transcriptional regulatory networks (TRNs).
• The success of simple mathematical models in capturing the complexity of TRNs is not fully understood.
• Understanding these principles is crucial for accurate TRN inference.

Purpose of the Study:

• To identify key principles of bacterial transcriptomes that facilitate accurate TRN inference.
• To investigate the reasons behind the effectiveness of simple models in TRN inference.
• To provide a catalog of dominantly regulated genes in *E. coli*.

Main Methods:

• Analysis of a 1035-sample *E. coli* gene expression database.
• Identification and quantification of regulon overlap and regulator dominance.
• Development of quantitative metrics to formalize observed transcriptome properties.

Main Results:

• Two transcriptome principles supporting TRN inference were identified: limited gene membership overlap in regulons and significant "regulator dominance" for 21% of genes.
• Regulator dominance, where gene expression strongly correlates with a single regulator, simplifies network structure.
• A catalog of 877 dominantly regulated *E. coli* genes was generated.

Conclusions:

• Regulator dominance explains discrepancies between expression-inferred and binding site-defined regulons.
• The presence of simply regulated gene subsets is central to effective TRN inference.
• These findings highlight the importance of regulator dominance in simplifying bacterial gene regulatory networks.