There has often been a strong temptation to brush epistasis under the carpet, as thrashing out the details of how genes interact with one another was thought to be best left to the few hardy geneticists who had a way with complicated equations and patchy data. Now, however, the pervasiveness and the undeniable importance of genetic interactions in biology, disease suceptibility and evolution has seen epistasis enjoy a welcome comeback. No small part in this revival has been played by the availability of improved statistical tools and global gene-analysis methods. Daniel Segrè, Roy Kishony and collaborators have now applied such a systems-level approach to describe the spectrum of epistasic interactions that occur between metabolic genes in the yeast Saccharomyces cerevisiae, and in so doing have transported epistasis beyond the gene level.

For this work — the fruit of a collaboration between the laboratories of Roy Kishony and George Church — the authors took advantage of a computational method known as flux-balance analysis to assess the effects of gene interactions on the growth rate of yeast cells. The fitness of a single mutant is measured as the degree to which the biomass production of mutant yeast differs from that of the wild-type. The presence and type of epistasis could then be evaluated by comparing the fitness value of the double mutants to the product of the fitness values of the single mutants of the same gene pair. From this analysis, the authors concluded that epistasis could be classified as being either aggravating (where one mutant enhances the effect of the other) or buffering (where one mutant mitigates the effect of the other).

This rather neat classification opened the door for a new type of analysis, in which individual gene–gene relationships were used to paint a picture of the overall organization of the metabolic network. The clusters ('modules') of interacting genes that emerged on the basis of annotated gene function were almost always either aggravating or buffering with respect to other modules. This is pretty much what you would expect to find if the genes within a given module were involved in carrying out the same biological function: because they are made up of genes with the same or similar biological function, modules effectively behave as if they are individual genes. These 'monochromatic' relationships between modules largely held true even when the hierarchical clusters were built without previous knowledge of gene function, indicating that the clusters and how they interact are intrinsic properties of this and, by extension, every other gene network. Most of the observed interactions make biological sense; and those that were unexpected can now be used to generate testable hypotheses about the network.

This systematic survey of gene interactions brings epistasis to a new level — one that includes not only interactions between genes, but also between groups of genes. 'Monochromatic modularity', as it is called, is launching epistasis towards a more colourful future.