Improved spectral clustering for protein complex detection

Complexes of actually interacting proteins are one of the primary functional units responsible for driving key biological mechanisms within the cell. Finding the Protein complexes from protein-protein interaction networks is first and foremost step in understanding the cell function. Currently large amount of accessible protein-protein interaction (PPI) data are available, which makes the researcher to find an accurate and scalable method for detecting protein complex. This paper shows about the working of three existing protein complex detection algorithms such as COACH, HUNTER, SPECTRAL CLUSTERING and one proposed algorithm Improved Spectral Clustering Algorithm. The proposed algorithm was applied on the data of DIP dataset. The experimental results show that the proposed algorithm can improve the accuracy and better performance measures on the validation of comparing with known complexes D1 and D2. The work is the result of implementation of these algorithms and tested with 4 yeast datasets namely Collins, DIP, Krogan and Krogan 2006.