Prediction of domain interactions in C. elegans
High throughput experiments and computational methods allow the elucidation of networks of protein-protein interactions (PPI) in several organisms. Since about 80% of all proteins consist of multiple self-folding units called domains, it is desirable to determine which domains are responsible for a given interaction. We use an integrative computational and experimental approach to predict domain-domain interactions (DDI) from the known PPI for the metazoan C. elegans. We use Maximum Likelihood Estimation and Maximum Specificity Set Cover algorithms to predict the DDI. We validate the results using a small and accurate yeast 2 hybrid (Y2H) screen on 8 protein-protein interactions involving 13 proteins in C. elegans. The computational predictions are consistent with Y2H in 7 of the 13 interaction domains, including the prediction of domain fusions. These results illustrate the predictive power of a combined computational and experimental approach to DDI determination.