Novel hub protein classification and interaction rules in protein-protein interaction network in Saccharomyces cerevisia

The majority of cellular functions can rarely be attributed to a single protein, gene or enzyme. Protein-protein interactions are fundamental to most of cellular processes or functions. Hubs, proteins with high connectivity, mediate the organization and function of cellular protein interaction networks (PINs). As a first step, research on hub proteins starts with classification. Although there are a few methods to classify hubs, the connection between classification methods and property of protein-protein interaction is unclear. In this work, hub proteins in S. cerevisiae were classified into three categories: systemic hubs, component hubs and process hubs based on biological annotation information in Gene Ontology (GO) using X-means clustering. The sub-network graphs of hubs and non-hubs were constructed. The hub proteins in different categories in the PIN graphs distribute differently. The distributions of sub-networks of systemic hub and non-hub proteins are uniform, while those of process hub and component hub proteins are modular. A parameter as an indication of protein class interaction bias (PCIB) was introduced and calculated to describe the tendency of a protein to interact with other proteins. It was shown that interaction bias among hub proteins, and from nonhub proteins to hub proteins are strong, while interaction bias among non-hub proteins or from hub to non-hub proteins is weak. Our theoretical result is consistent with experimental findings that different groups of hubs function at specific times or in characteristic space in cellular system. For the first time, we show the interactions among different hub classes are biased, and the interaction between hubs is directional. Our result also suggests possible connection between hub classification and protein interaction bias.