Although systems biology helps to make sense of the complex interactions between genes, proteins and other biologically relevant molecules, most studies have provided only snapshots of how these networks operate in specific conditions. A recent paper in Nature describes the first genome-scale study of how biological networks are rewired according to the needs of the cell, revealing some important insights into network dynamics.

Luscombe and colleagues first defined a network of known interactions between 142 transcription factors and 3,420 genes in the yeast Saccharomyces cerevisieae. They then incorporated data from previous studies that had examined patterns of gene expression during the cell cycle, sporulation, diauxic shift (the switch from anaerobic to aerobic respiration), DNA damage and the stress response, and used an algorithm to identify sub-networks of interactions that are active in these different conditions.

To characterize these sub-networks, the authors devised a statistical method — SANDY (statistical analysis of network dynamics) — for the analysis of interactions both within and between conditions. Overall comparisons grouped the five sub-networks into two categories. The 'exogenous' diauxic shift, DNA-damage and stress-response sub-networks were characterized by the regulation of several genes by each transcription factor, and by shorter pathways leading from transcription factors to their target genes. This fits in with their ability to produce large and rapid responses to changes in the environment. By contrast, the 'endogenous' cell-cycle and sporulation sub-networks allow more precise, multi-stage control, with longer pathways to activation and more inter-regulation between transcription factors.

Although these results might not seem surprising, given the biological functions of the sub-networks, more unexpected patterns emerged when other characteristics were investigated. Static gene-regulatory networks are characterized by the existence of 'hubs' — individual transcription factors that regulate a disproportionately large number of genes. The importance of these hubs suggests that they are likely to function across a range of conditions, as they regulate key pathways, and this is supported by theoretical simulations. However, the authors found the reverse to be true: most hubs (78%) were important in only a single set of conditions and were therefore dubbed transient hubs. Another surprising result was seen when the interactions made by those hubs that do function across several conditions, known as permanent hubs, were examined. Rather than using a similar set of interactions in each condition, these hubs redefined their interactions just as frequently as transient hubs — further evidence that networks are more dynamic than was previously thought.

As Luscombe and colleagues point out, their study was limited to results that were available from previous experiments, although the robustness of the features they describe in response to random noise suggests that similar patterns are likely to emerge from direct studies of S. cereviseae network dynamics. The increasing availability of genome-wide data on regulatory interactions in cell types should allow future studies to determine whether these features apply on a wider scale.