Authors
Duygu Ucar, Sitaram Asur, Umit Catalyurek, Srinivasan Parthasarathy
Publication date
2006/9/18
Book
European Conference on Principles of Data Mining and Knowledge Discovery
Pages
371-382
Publisher
Springer Berlin Heidelberg
Description
Dense subgraphs of Protein-Protein Interaction (PPI) graphs are believed to be potential functional modules and play an important role in inferring the functional behavior of proteins. PPI graphs are known to exhibit the scale-free property in which a few nodes (hubs) are highly connected. This scale-free topology of PPI graphs makes it hard to isolate dense subgraphs effectively. In this paper, we propose a novel refinement method based on neighborhoods and the biological importance of hub proteins. We show that this refinement improves the functional modularity of the PPI graph and leads to effective clustering into dense components. A detailed comparison of these dense components with the ones obtained from the original PPI graph reveal three major benefits of the refinement: i) Enhancement of existing functional groupings; ii) Isolation of new functional groupings; and iii) Soft clustering of …
Total citations
Scholar articles
D Ucar, S Asur, U Catalyurek, S Parthasarathy - European Conference on Principles of Data Mining …, 2006