Protein Function Prediction by Spectral Clustering of Protein Interaction Network

The increasing availability of large-scale protein-protein interaction (PPI) data has made it possible to understand the basic components and organization of cell machinery from the network level. Many studies have shown that clustering protein interaction network (PIN) is an effective approach for identifying protein complexes or functional modules. A significant number of proteins in such PIN remain uncharacterized and predicting their function remains a major challenge in system biology. We propose a protein annotation method based on spectral clustering, which first transforms the PIN using the normalized Laplacian of the PIN graph, and then employs a classic clustering algorithm like k-means. Protein functions are assigned based on cluster information. Experiments were performed on PPI data from the bakers’ yeast and since the network is noisy and still incomplete, we use pre-processing and purifying. We also performed network weighting based on the annotation correlation between nodes. Results reveal improvement over previous techniques.