Protein Complex Detection Based on Semi-Supervised Matrix Factorization

The identification of protein complexes is important for understanding the organization and function within cells. Owing to the quick development of high-throughput experimental techniques, an increasing number of protein-protein interaction (PPI) datasets are accumulated, which makes it possible to detect protein complexes from PPI networks. However, most of the currently available protein complex detection methods are unsupervised learning methods and can not utilize the information of the known complexes. In this paper, we present a novel semi-supervised matrix factorization method, called Constrained Overlapping Complex Detection Model (COCDM), which incorporates the known complexes as additional constrains. Experimental results demonstrate that COCDM performs much better than most state-of-the-art complex detection techniques.