For four days last July, Giulio Superti-Furga, Robert Russell and a few of their colleagues shut themselves away in a hotel room on the banks of the Danube. There they pored over thousands of images of yeast proteins. They were scouring the pictures to try to work out whether most cellular proteins exist as complexes and to see how they are organized.

The fruits of their labour appear on page 631 of this issue, where they confirm that the proteins do indeed form hundreds of ‘core’ complexes. But they also show that these core complexes can combine with other proteins to gain additional capabilities — much like extra tool modules can be attached to a machine.

The quest for the complexes began some six years ago at Cellzome, the biotechnology firm co-founded by Superti-Furga in Heidelberg, Germany. The initial plan was to characterize all of the detectable protein complexes in a cell — the team picked yeast as it is a relatively simple and well-studied model system.

Anne-Claude Gavin, the paper's joint first author, led the effort to express all of the known yeast proteins with a small tag that would allow each one to be extracted from the cell along with any other protein bound to it. The researchers then used mass spectrometry to identify the captured proteins and their partners.

A trial run for this technique, using a subset of yeast's proteins, was published a few years ago (A.-C. Gavin et al. Nature 415, 141–147; 2002). It revealed that protein complexes are indeed the rule rather than the exception. But in the latest results, the team also fished for different binding partners for each protein several times, eventually collecting data from 52,000 samples. “By doing the entire protein complement, there is nowhere a protein can hide,” says Superti-Furga. “We would have not seen the modularity otherwise.”

Gathering such a massive amount of data may seem like a daunting task, but the real challenge still lay ahead: converting the information into something biologically meaningful. Superti-Furga turned for help to Russell, a bioinformatician at the European Molecular Biology Laboratory in Heidelberg. “Getting first access to original data that are very exciting is a dream come true for a bioinformatician,” says Russell, who worked with Patrick Aloy on the analysis.

The process of getting from data to meaning involved much trial and error — and a lot of back-and-forth communication between the labs to reach a consensus. “An important element was the very intense discussions between biologists and bioinformaticians. That was the best part of the project, and where the magic happened,” says Superti-Furga, who recently left Cellzome to head the Research Center for Molecular Medicine at the Austrian Academy of Sciences in Vienna.

Apart from meetings in European hotels, these discussions took place on the phone and by e-mail over several months. Russell alone counts some 1,020 messages in his inbox relating to the project.

By explaining the data from each member's perspective, the group eventually arrived at a common interpretation. “You have to keep trying to describe what you think you are seeing and the other person keeps trying to see whether they understand,” says Superti-Furga.