A network comparison algorithm for predicting the conservative interaction regions in protein-protein interaction network

We presented a network comparison algorithm for predicting the conservative interaction regions in the cross-species protein-protein interaction networks (PINs). In the first place, We made use of the correlated matrix to represent the PINs. Then we standardized the matrix and changed it into a unique representation to facilitate to judge whether the subgraphs is isomorphic. Then we proposed a network comparison algorithm based on the correlated matrix, edge-betweenness and the maximal frequent subgraphs mining. We used the tag grath library composed of the multiple PINs as input data and mined the maximal frequent subgraphs in the cross-species PINs by the algorithm. In the second place, we clustered and merged the similar but different and duplicate locally regions according to the similarity between them and the principle of sigle linkage clustering. In the end we analysed the resulting subgraphs and predicted the conservative interaction regions. The results showed the network comparison algorithm based on mining the frequent subgraplhs can be successfully applied to discover the conservative interaction regions, that is, we can find the functional complexes and predict the protein function. Furthermore, we can predict the interaction will exist in the other species when the conservative regions meet or exceed the threshold of minimum support.