Gene Lethality Detection and Characterization via Topological Analysis of Regulatory Networks

Recent engineering approaches to the analysis of biomolecular systems have shown a connection between the topological centrality of a protein in an interaction network and importance of that protein's underlying gene in the survival of the system. The recognized reason for this connection is the scale-free nature of these networks, which induces the presence of highly connected nodes (hubs) in the network. This approach would also be expected to identify these critical genes-called lethal because their removal would kill the cell- in the E. coli transcriptional regulation network, a prototypical scale-free network. However, this method does little better than simple random selection. Here we introduce a new approach, based on random walks, that identifies the critical nodes in the network on the basis of global properties of the network rather than the local connectivity of each node; we show that it significantly outperforms the standard local approach. We also find that lethal genes are more likely to belong to certain functional categories. The success of the our approach suggests that computational lethal gene detection can be effective and that scale free networks-and their biological instances-may enjoy more global properties than the one identified so far