Information processing in the transcriptional regulatory network of yeast: Functional robustness

In this paper we study the functional robustness of the transcriptional regulatory network of S. cerevisiae. We model the information processing in the network as a first order Markov chain and study the influence of single gene perturbations on the global, asymptotic communication among genes. Modification in the communication is measured by an information theoretic measure allowing to predict genes that are 'fragile' with respect to single gene knockouts. Our results demonstrate that the predicted set of fragile genes contains a statistically significant enrichment of so called essential genes that are experimentally found to be necessary to ensure vital yeast. Further, a structural analysis of the transcriptional regulatory network reveals that there are significant differences between fragile genes, hub genes and genes with a high betweenness centrality value.Our study does not only demonstrate that a combination of graph theoretical, information theoretical and statistical methods leads to meaningful biological results but also that such methods allow to study information processing in gene networks instead of just their structural properties.The advent of high-throughput technologies in molecular biology has initiated an avalanche of data that possess considerable challenges to quantitative sciences providing statistical analysis methods [1]. Due to the fundamental insight that biological processes should be studied holistically [2-4] instead of reductionistically, systems based approaches are of central importance in this respect [5]. For this reason, it is no surprise that network related studies experience an enormous interest starting with the investigation of small-world [6,7] and scale-free [8,9] networks in the mid 1990's followed by numerous studies devoted to the analysis of complex network topologies and their properties in general [8,10-14]. It is interesting to note that many apparently different networks have similar properties. Most prominent exa