Protein interaction mapping: a Drosophila case study

Abstract

The Drosophila (fruit fly) model system has been instrumental in our current understanding of human biology, development, and diseases. Here, we used a high-throughput yeast two-hybrid (Y2H)-based technology to screen 102 bait proteins from Drosophila melanogaster, most of them orthologous to human cancer-related and/or signaling proteins, against high-complexity fly cDNA libraries. More than 2300 protein-protein interactions (PPI) were identified, of which 710 are of high confidence. The computation of a reliability score for each protein-protein interaction and the systematic identification of the interacting domain combined with a prediction of structural/functional motifs allow the elaboration of known complexes and the identification of new ones. The full data set can be visualized using a graphical Web interface, the PIMRider (http://pim.hybrigenics.com), and is also accessible in the PSI standard Molecular Interaction data format. Our fly Protein Interaction Map (PIM) is surprisingly different from the one recently proposed by Giot et al. with little overlap between the two data sets. Analysis of the differences in data sets and methods suggests alternative strategies to enhance the accuracy and comprehensiveness of the post-genomic generation of broad-scale protein interaction maps.

Figure 1.

A global view of the Drosophila embryo interaction map. This view of the Protein Interaction Map (PIM) resulting from our work is generated using the PIM Walker graphical interface, a tool developed for network display. The proteins in the map that bear an RA (Ras Association) or RBD (Raf-like Ras-binding) domain (colored in orange) define a discrete subnetwork around Ras-like GTPases (colored in yellow).