Pathway prediction in protein-protein interaction networks based on hierarchical clustering algorithm

Pathway prediction is vital for understanding biological processes and the mechanism of controlling products synthesis, and for identifying drug targets. Pathway prediction has been become a key challenge in system biology. In this paper, a novel computation algorithm is proposed for predicting the genes in the pathways. Firstly, a formula for computing similarity measure is defined by considering that whether the two proteins are direct interaction or not, and that whether the number of common interaction partners is more significant than random. Secondly, a PPI network is constructed through protein-protein interaction dataset, and the similarity measure of each protein pair is computed. At last, hierarchical clustering algorithm is employed for mining its modular structures, namely, protein clusters. Through mapping proteins to the corresponding genes, we can obtain gene clusters and their pathways in the target species. The proposed method is tested on Escherichia coli k-12. Experimental results have shown the effectiveness and attractiveness of the proposed method. Copyright © 2013 American Scientific Publishers. Website © 2013 Publishing Technology.