Computational verification of protein-protein interactions by orthologous co-expression

We introduce a novel computational method for verification of protein-protein interactions based on the co-expression of orthologs of interacting partners. The performance of our method is analysed using known S. cerevisiae interactions, and is shown to overcome limitations of previous methods. We present specific examples of known and putative interactions that are detected by our method and not by previous methods, and suggest that they represent transient interactions that might have been conserved and stabilized in other species.Co-expression of orthologous protein-pairs can be used to increase the confidence of hypothetical protein-protein interactions in S. cerevisiae as well as in other species. This approach may be especially useful for species with no available expression profiles and for transient interactions.Protein-protein interactions (PPIs) have a central role in most biological processes, and identifying these interactions is an important goal of biological research. PPIs are the subject of extensive experimental studies, but the majority of them remain unknown. In the last few years, high-throughput techniques were developed for the identification of PPIs on a genomic scale. Yeast two-hybrid [1,2] and mass spectrometric analysis of protein complexes [3,4] were used to produce large sets of PPIs. However, these techniques are known to suffer from many false positives and the resulting PPIs are typically regarded as putative [5,6]. Thus, the development of computational methods for assessment and verification of putative PPIs is crucial [7-10]. Two such methods were proposed, that are based on the co-expression [11] and conservation [9] of PPIs, respectively. Here we propose to extend these methods by considering co-expression of orthologous protein pairs. We demonstrate the predictive power of our approach and discuss its advantages.It was previously shown that interacting pairs of proteins are often correlated in their expression profiles [11,12]. T