Prediction of Unannotated Proteins from a Protein Interaction Network Filtered by Using Localization and Domains in Yeast

Abstract

We have constructed and analyzed a protein-protein interaction network from the Saccharomyces cerevisiae genome by using public genomic sequences. This provides a starting point for the construction of protein network analysis, leading to a better understanding of the functions of uncharacterized proteins and of the cellular functions of modules the networks. We acquired a highly filtered reliable protein interaction database through filtering by using subcellular localization and by using protein domains. The protein-protein interaction network constructed from the highly filtered protein interaction database shows scale-free and hierarchical properties as most networks in complex systems. For the prediction of unannotated proteins, we employed a modified Chi-square method based on neighborhood counting within structural modules and we predicted and elucidated the functions of the 98 unannotated proteins with high certainty.