An Improved Method Based on Maximal Clique for Predicting Interactions in Protein Interaction Networks

The datasets identified by large-scale, high- throughput methods typically suffer from a relatively high level of noise. Combining the distribution characteristics of noise data and topological properties in the protein interaction network, we described a novel method to improve the reliability of those datasets by predicting missed interactions. The main idea of the method is to predict the interactions among proteins based on the degree of correlation between protein and protein clique, and improve prediction reliability by percolating most amplified noise data. We have applied this approach to some high-throughput datasets. The experimental results show that this method can not only predict more but also higher reliable interactions than the prediction method proposed by Haiyuan Yu in 2006.