Protein Protein Interaction using Fast Random Walk

Protein-protein interactions (PPI) refer to the associations between proteins and the study of these associations. Recent studies show that a network representation of proteins provides a more accurate model of biological systems and process compared to conventional pair-wise analysis. Many graph analysis has been proposed to the network for interactions prediction, pathway discovery and complex membership prediction. Random walk with restart (RWR) has been demonstrated to be a competitive approach on PPI network in terms of accuracy and effiency. RWR provides a good relevance score between two nodes in a weighted graph. However, with the high-thoroughput of the detection of PPI interactions, the straightforward application of RWR into the problem does not scale up well. We propose fast solutions to this problem. The heart of the approach is to exploit the block-wise structural property of PPI and furthermore, an iterative aggregation and desegregation method is adapted into this problem. The results shows a speedup factor of 10 in terms of time and convergence. We proposed a new approach to integrate proteinprotein pair information into network. Finally, we evaluate the proposed technique on prediction of interaction, pathway and complex membership using three different benchmark data sets. Our methods shows a similar and better results compared with previous work.