Mike Boxem (1,2), Niels J. Klitgord (1), Na Li (1), Kristin C. Gunsalus (3), Fabio Piano (3), David E. Hill (1), Sander van den Heuvel (2,4), Mike Tipsworth (5), David Drechsel (5), Anthony A. Hyman (5) and Marc Vidal (1)

1: Dana Farber Cancer Institute, Boston, USA 2: MGH Cancer Center, Charlestown, USA 3: New York University, New York, USA 4: Utrecht University, Utrecht, The Netherlands 5: Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany

The embryo of the nematode C. elegans forms an excellent model system to study processes common to the development of multicellular organisms, such as cell division, cell-fate specification, differentiation and polarity establishment. Recent RNAi studies by the Hyman, Ahringer and Piano labs as well as Cenix Biosciences have collectively identified and characterized in detail the phenotype of 721 genes essential during the early embryonic cell divisions. These genes provide an ideal starting point for a systems biology approach aimed at enhancing our understanding of the molecular mechanisms underlying embryonic development.

We are mapping the domains mediating interactions between early embryogenesis proteins. We are systematically generating an average of 40 fragments for each of the 721 genes in the Gal4-AD Y2H vector. All fragments will be combined into a fragment library, which will be screened in the Y2H system with full length clones of all early embryonic genes. The resulting data will be combined with the phenotypic characterizations and expression data into an integrated map of early embryogenesis, which should provide a basis for gaining more detailed insight into the organization of the complex molecular systems involved in embryogenesis.

The predictive value of high-throughput approaches depends on the quality, level of detail, and completeness of the underlying data. Our approach is aimed at making advances on all three levels. First, the combination of several measurements of similarity has been shown to result in a more reliable network, containing fewer false positives. Second, knowledge of interaction domains provides a new level of detail not present in most current interaction datasets, and will allow us to generate more accurate models and predictions. For example, it will enable us to know whether two interactions can take place simultaneously, or are mutually exclusive because they are mediated by the same domain. Finally, current interaction datasets are still limited in their coverage. By testing multiple fragments of each protein for interaction we will lower the number of false-negatives in the Y2H system.