...a fascinating snapshot of the yeast proteome...

Proteins rarely act alone in cells, but instead are components of larger 'molecular machines'. These dynamic complexes of proteins are the key to biological function, and understanding the networks of reactions that are mediated in these complexes is crucial for deciphering how a biological system responds to change. Now, reporting in Nature, Gavin et al. describe the first complete genome-wide screen for protein complexes, and provide a de novo characterization of the composition and organization of the cellular machinery of the yeast Saccharomyces cerevisiae.

Using tandem affinity purification coupled to mass spectrometry, the authors tagged all of the open reading frames of S. cerevisiae, thereby extracting complete protein complexes — constituting 73% of all known complexes — from the cells. The authors then derived a 'socio-affinity' index to quantify the likelihood of proteins to form partnerships — this represents the first attempt to recreate numbers that approximate physical measurements from proteomics data alone. A collection of 491 complexes was generated, and a comparison with the complete collection of known complexes showed that 257 were novel; interestingly, all but 20 of the previously known complexes were shown to contain novel components.

This purification procedure divided proteins into core components, which are present in most isoforms, and 'attachments', which are found only in some isoforms. Additionally, the authors noted several instances in which two or more proteins in the attachments were always together and were present in multiple complexes — they called these complexes 'modules'. Gavin et al. found that, overall, proteins that were within cores or modules showed the greatest degree of functional similarity and physical association, which strongly supports the view that core components represent functional units.

The architectural details of known complexes that were captured in the analysis provided new information — the data provided a dynamic view of cellular processes, hints of novel regulatory mechanisms, and allowed the specification of subtle differences in function among modules. In addition, by deriving a matrix that represented a global view of the connections between cores and modules, the authors were able to suggest functions for modules, and highlight many known connections.

Gavin and colleagues extrapolated that there might be a total of 800 core machines in yeast and ∼3,000 in humans. Yet, they noted that these numbers are small in comparison with the myriad cellular processes that these protein-complex cores mediate. The high modularity is therefore an efficient means to multiply functionality while simplifying temporal and spatial regulation. The next step will be to integrate the data that were generated in this study into rational models of entire systems. But in the meantime, the authors have provided a fascinating snapshot of the yeast proteome from which to start.