Analysis of protein-protein interaction networks using random walks

Genome wide protein networks have become reality in recent years due to high throughput methods for detecting protein interactions. Recent studies show that a networked representation of proteins provides a more accurate model of biological systems and processes compared to conventional pair-wise analyses. Complementary to the availability of protein networks, various graph analysis techniques have been proposed to mine these networks for pathway discovery, function assignment, and prediction of complex membership. In this paper, we propose using random walks on graphs for the complex/pathway membership problem. We evaluate the proposed technique on three different probabilistic yeast networks using a benchmark dataset of 27 complexes from the MIPS complex catalog database and 10 pathways from the KEGG pathway database. Furthermore, we compare the proposed technique to two other existing techniques both in terms of accuracy and running time performance, thus addressing the scalability issue of such analysis techniques for the first time. Our experiments show that the random walk technique achieves similar or better accuracy with more than 1,000 times speed-up compared to the best competing technique.