Gene lethality detection across biological network domains: Hubs versus stochastic global topological analysis

In this paper, we investigate the properties of lethal genes in E. coli, our model organism. Topological analysis of networks of functional interactions among genes has shown that lethal genes share common local connectivity properties. In this paper, we analyze cellular networks across three domains. We show that a stochastic global topological analysis, via random walks, is more effective at predicting gene lethality than simply looking at local topology using the standard hub-based method. We also introduce the possibility of using metabolic pathways to understand lethal genes, as regulating these pathways is among one of the most important functions of the gene-encoded proteins. Additionally, we analyze lethal genes in terms of the Gene Ontology (GO) and find that the graph forms two highly connected clusters that are each GO enriched for specific terms. We also find that lethal metabolic regulators are extremely enriched. Finally, we provide applications of the work and avenues for future research.