Bottlenecks and Hubs in Inferred Networks Are Important for Virulence in Salmonella typhimurium

Abstract

Recent advances in experimental methods have provided sufficient data to consider systems as large networks of interconnected components. High-throughput determination of protein-protein interaction networks has led to the observation that topological bottlenecks, proteins defined by high centrality in the network, are enriched in proteins with systems-level phenotypes such as essentiality. Global transcriptional profiling by microarray analysis has been used extensively to characterize systems, for example, examining cellular response to environmental conditions and effects of genetic mutations. These transcriptomic datasets have been used to infer regulatory and functional relationship networks based on co-regulation. We use the context likelihood of relatedness (CLR) method to infer networks from two datasets gathered from the pathogen Salmonella typhimurium: one under a range of environmental culture conditions and the other from deletions of 15 regulators found to be essential in virulence. Bottleneck and hub genes were identified from these inferred networks, and we show for the first time that these genes are significantly more likely to be essential for virulence than their non-bottleneck or non-hub counterparts. Networks generated using simple similarity metrics (correlation and mutual information) did not display this behavior. Overall, this study demonstrates that topology of networks inferred from global transcriptional profiles provides information about the systems-level roles of bottleneck genes. Analysis of the differences between the two CLR-derived networks suggests that the bottleneck nodes are either mediators of transitions between system states or sentinels that reflect the dynamics of these transitions.