IntroductionInactivation of proteins that participate in more than one cellular process leads to a variety of apparently unconnected phenotypes. Understanding the molecular cause for each phenotype might reveal how seemingly independent cellular processes are regulated and coordinated in the cell. Genome-wide gene interaction data based on the simultaneous inactivation of more than one gene greatly facilitate this inherently complex analysis because genes with pleiotropic phenotypes often occupy central positions in the corresponding interaction networks (Costanzo et al., 2010;Tong et al., 2004). By assigning physical connections, protein-protein interaction maps provide the necessary complementary information. Interpretation of these maps is usually not straightforward. Genetic interactions can result from complex functional relationships between the investigated pairs of genes and protein interaction maps are generally projections of contacts that occur at different times and places in the cell. To transform protein interaction data into mechanistically meaningful models, it is necessary to resolve these projections into their different interaction planes. We define an interaction plane or state as the sum of all simultaneously occurring contacts. Ideally, these states should be defined by time-and space-resolved in vivo studies. However, these studies are technically demanding and usually not suited for measuring multiple contacts (Maeder et al., 2007). Using the protein pair Ptc1p-Nbp2p of the yeast Saccharomyces cerevisiae as an example and the split-ubiquitin method (SplitUb) as the experimental tool, we present an alternative approach for defining interaction states. The derived constraint interaction network reduces the number of possible states and thus provides a useful framework for model building and the initiation of more detailed studies.

The Slt2p-Slt2p interaction reported in Fig. 8E was mistakenly not included in Table 1. The authors apologise for this error.