The advent of genome sequencing has generated a shift in the way in which researchers approach problems in biology – instead of analysing one gene at a time, studies are increasingly focused on whole pathways or gene networks. Joshua Stuart, Eran Segal and their colleagues bring us a prime example – by using microarray data from humans, flies, worms and yeast, they identify clusters of functionally related genes that have been conserved in evolution. The resulting gene co-expression network has uncovered new functional and evolutionary relationships.

Genes that have related functions often share expression patterns, and so microarray data can be used to investigate functional relationships between genes. Stuart, Segal and colleagues constructed a gene co-expression network using pre-existing microarray data from four species; the assumption being that evolutionarily conserved co-expression is an indicator of its functional significance.

To make the network, the authors first associated genes from one organism with their orthologues from the others, using all-against-all BLAST. Based on the highest BLAST scores, they identified 'meta-genes' that corresponded to sets of orthologues from the organisms used in the study. They came up with 6,307 meta-genes that correspond to 6,591 human genes, 5,180 worm genes and so on.

The next task was to identify pairs of meta-genes that were co-expressed in many organisms across different experimental conditions. The authors looked for correlation of expression in data from 3,182 DNA microarray experiments from humans, worms, flies and yeast. Using a probabilistic model, the authors selected interactions that did not occur by chance alone and combined them into a network of meta-gene co-expression.

So, what does this network of 3,416 meta-genes and 22,163 interactions tell us about biology? When the network is viewed as a three-dimensional map, the highly interconnected areas appear as peaks, which correspond to clusters of specific gene–gene interactions. As the authors show, most of the components of each peak are involved in similar biological processes. For example, peak 5 is made up of interactions between cell-cycle genes. Importantly, out of the 241 meta-genes in this peak, 131 have no known function. Thanks to the network, their candidate role in the cell cycle can now be tested.

The network can also tell us about the evolution and conservation of genetic interactions. For example, peak 1, which is enriched in meta-genes that are involved in signalling pathways, is enriched in animal-specific genes and has a lower degree of evolutionary conservation.

So, the co-expression network turns out to be an extremely useful tool with which to explore the functional relationships between already characterized genes, and also to implicate new players in given biological processes and inform us about the evolution of the underlying molecular complexity. No doubt we will hear from these authors with more interesting results of further analysis of this network.