06 Nov 2003

According to the six degrees of separation theory, nobody is more than six personal contacts away from anyone else on the planet. In the human world, it may not always work like that, but how about the protein world, which is considerably smaller? In today’s ScienceExpress, John Chant and colleagues at CuraGen, New Haven, Connecticut, report that they have identified over 20,000 different interactions among over 7,000 proteins that are coded in the genome of the fruit fly Drosophila melanogaster. This has allowed them to create extensive and intricate protein-protein contact maps. “This work is very substantial and will have an impact on many aspects of fly biology, including human disease modeling,” said Mel Feany of Brigham and Women’s Hospital.

Joint first authors Loic Giot, JS Bader, C Brouwer, and A Chadhuri expanded on previous work from this group that used high-throughput two hybrid analysis to explore the yeast proteome (see Uetz et al., 2000). In the current study, the authors first made cDNAs representing each transcript on the Drosophila genome and cloned them into "bait” and “prey” vectors to be used in the classic two-hybrid screen. After eliminating any vectors that self-activated the reporter gene, they then generated a draft map by screening individual baits with prey libraries. This yielded 20,349 interactions among 7,048 proteins. The complete data set is available as supplementary material on the Science website.

Next, the authors refined the data to generate a higher-confidence map. Criteria included the number of times each interaction was observed, the number of interaction partners for each protein, and whether it was part of a local network of interactions. In addition, the scientists deemed protein-protein couples having orthologs in yeast more reliable, while assuming low confidence for those that seem unlikely in vivo, such as between nuclear and extracellular proteins. This analysis yielded a map of 4,780 unique interactions among 4,679 proteins.

The usefulness of this data comes with the mapping. For example, the authors show a global map of interactions based on proteins known to cause disease in humans. This reveals potential drug targets, for example, the calcium-dependent phosphatase that interacts with BCL6, a transcription factor which plays a key role in the pathogenesis of non-Hodgkin's lymphoma. The authors postulate from their data that BCL6 may require dephosphorylating before nuclear translocation. On a smaller scale, local maps—for example, one describing transcription of cell cycle regulatory proteins, which the authors used to identify novel Src-homology proteins and potential new pathways for ubiquitin-mediated protein recycling—may also provide useful information to those studying human diseases.

Does the company have an active program for AD drugs? Yes, but it’s in early stages and company rules prohibit divulging details, said Chant. Establishing the same maps for the mouse or human would require more computing power and sophisticated industrialization to handle the large data sets generated. Curagen has, however, established subsets of human protein-protein interaction data, Chant added.—Tom Fagan