As many captains of industry will tell you, people often work most effectively in closely interacting groups. The same goes for proteins, many of which function in multiprotein complexes — the workhorses of cellular life. However, such complexes are often large and dynamic, which makes them hard to study. Now, two teams report in Nature a high-throughput approach specifically tailored to analysing multiprotein complexes in Saccharomyces cerevisiae. Their findings provide a detailed view of the higher organization of this organism's proteome and provide valuable new information for annotating eukaryotic genomes.

Both teams took a similar approach to isolating protein complexes in S. cerevisiae. They began by adding an affinity tag to hundreds of yeast genes. When introduced into the yeast genome, these modified genes encode 'bait' proteins that can complex with other proteins under physiological conditions. Each bait can be captured via its tag, together with its associated proteins, which are then identified by mass spectrometry and computational analysis.

From a library of 1,548 tagged strains, Gavin et al. purified 589 bait proteins, from which they identified 1,440 distinct proteins that were present in 232 multiprotein complexes, 91% of which contained a novel component. Complexes were placed into functional categories based on literature- and database-derived information about the known proteins in each complex. Most complexes shared at least one component with another complex, forming a network of interactions. When organized by such connections, complexes from some functional categories grouped together, perhaps because their shared components reflect an underlying functional relationship. This analysis allowed Gavin et al. to propose roles for 231 of 304 proteins of unknown function. To assess whether the complexes that contain many highly conserved proteins are likely to be crucial to eukaryotic cells in general, Gavin et al. also purified proteins from three yeast multiprotein complexes — the Arp2–Arp3, Ccr4–Not1 and TRAPP complexes — from human cells, using the affinity tagging approach. They found the human complexes to be of a similar, if not identical, composition, indicating that this type of approach in yeast can inform studies of the human proteome.

Ho et al. used 725 proteins as baits, mostly kinases, phosphatases and components of the DNA damage response (DDR), from which they identified 1,578 proteins — 531 of unknown function — involved in 3,617 interactions. The authors analysed complexes associated with several important transcription factors and kinases, such as Fkh1, Kss1, Cdc28 and Dun1, a DDR component for which Ho et al. found many regulators and targets. Their use as bait of 86 proteins involved in the DDR allowed them to piece together the network of interacting proteins that controls this response, so uncovering many new interactions of probable biological significance.

Although this approach is clearly very powerful, it is not without limitations — for example, Gavin et al. could not purify proteins under 15 kDa in size. Both groups also report a significant number of false-positive interactions, while failing to detect some known interactions, perhaps because the tag can interefere with a protein's function or with its physical associations. Although there is still a long way to go before we fully understand how a proteome's functional networks respond to the ever changing life of a cell, these two studies provide a panoramic view of protein function and a wealth of new functional data for genome annotation.