Searching Quasi-bicliques in Proteomic Data

Finding a maximal biclique in a graph is an NPC problem. We developed a new, efficient method to search maximal quasi-bicliques from protein interaction network. At first, we divided the protein- protein interaction network into vertex's distance-2-subgraphs. Next, combining top down and branch and bound methods, we sought quasi-biclique from every vertex's Distance-2-Subgraph. At last, we merged the redundant ones in the quasi-bicliques. We successfully applied our method on the Saccharomyces cerevisiae dataset and obtained 2 754 distinct quasi-bicliques.