A statistical-based method for the construction and analysis of gene network：application to bacteria

Abstract

Bacteria play a crucial role in environmental conservation, human health, and medicine. Whether in the gut or the soil, bacterial genomes are rich repositories of resources, such as exploring potential drugs and biopesticides. However, our ability to develop new therapies and deepen our understanding of the bacterial world is hindered by the largely unknown functions of bacterial genes. In this study, we proposed a method of gene network construction and analysis based on a Gaussian Graphical Model (GGM) and random sampling strategy to infer direct interactions at the genomic level in bacteria. Using Vibrio cholerae and Staphylococcus aureus as examples, we integrated partial correlation-based gene co-expression data with gene regulatory and essentiality information extracted from public databases to construct more comprehensive gene networks. Networks built upon bacterial different phenotypes, such as biofilm formation, flagellar assembly, and stress response, demonstrate the effectiveness of this method in deciphering unknown gene functions, uncovering new phenotype-associated factors, and identifying their corresponding interactions, thus providing new targets for experimental validation by researchers. Additionally, we extended this method to 14 bacteria, including 13 pathogens, supporting the investigation of gene functions and pathways at the genomic level in these bacteria. More importantly, for other species, this method of gene network construction can be easily implemented, provided that sufficient transcriptome sequencing samples are available.