Poster #RP115
Evaluation of predicted human protein networks
Mario Albrecht*, Fidel Ramirez*, Andreas Schlicker*, Yassen Assenov*, Hagen Blankenburg*, Carola Huthmacher*
*Max Planck Institute for Informatics, Saarbruecken, Germany
Novel high-throughput techniques have generated enormous amounts of protein-protein interaction data for different species. This experimental data can now be mined for new information on the functions and interrelationships of proteins. In particular, different bioinformatics methods, mainly based on the homology of protein sequences, have supported the large-scale prediction of human protein networks. In addition, manually curated literature data and several yeast-2-hybrid maps of considerable size have recently become available for the human interactome. However, the experimental coverage of the human interactome is still low in contrast to predicted data.
Therefore, we used the recent experimental data to assess the reliability of previously predicted interactions. Such an evaluation and comparison of prediction methods is not only important for further methodological improvements, but also for gaining confidence into functional hypotheses derived from predictions, for instance, when studying disease-associated proteins and potential drug targets. Indeed, in our comprehensive analysis, we found significant differences between the experimental and predicted human networks regarding accuracy, functional consistency, information contents, and network topology. For this work, we built a sophisticated database to integrate diverse biological information on protein interactions and implemented a useful Cytoscape plugin named NetworkAnalyzer to compare networks and to compute topological parameters.