Towards Inferring Protein Interactions: Challenges and Solutions
- PDF / 3,056,069 Bytes
- 14 Pages / 600.03 x 792 pts Page_size
- 63 Downloads / 221 Views
Towards Inferring Protein Interactions: Challenges and Solutions Ya Zhang,1, 2 Hongyuan Zha,3 Chao-Hsien Chu,4 and Xiang Ji5 1 Information
and Telecommunication Technology Center, The University of Kansas, Lawrence, KS 66045, USA of Electrical Engineering and Computer Science, The University of Kansas, Lawrence, KS 66045, USA 3 Department of Computer Science and Engineering, School of Engineering, Pennsylvania State University, University Park, PA 16802, USA 4 College of Information Sciences and Technology, Pennsylvania State University, University Park, PA 16802-6823, USA 5 NEC Laboratories America, Inc., Cupertino, CA 95014, USA 2 Department
Received 1 May 2005; Revised 13 October 2005; Accepted 15 December 2005 Discovering interacting proteins has been an essential part of functional genomics. However, existing experimental techniques only uncover a small portion of any interactome. Furthermore, these data often have a very high false rate. By conceptualizing the interactions at domain level, we provide a more abstract representation of interactome, which also facilitates the discovery of unobserved protein-protein interactions. Although several domain-based approaches have been proposed to predict protein-protein interactions, they usually assume that domain interactions are independent on each other for the convenience of computational modeling. A new framework to predict protein interactions is proposed in this paper, where no assumption is made about domain interactions. Protein interactions may be the result of multiple domain interactions which are dependent on each other. A conjunctive norm form representation is used to capture the relationships between protein interactions and domain interactions. The problem of interaction inference is then modeled as a constraint satisfiability problem and solved via linear programing. Experimental results on a combined yeast data set have demonstrated the robustness and the accuracy of the proposed algorithm. Moreover, we also map some predicted interacting domains to three-dimensional structures of protein complexes to show the validity of our predictions. Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.
1.
INTRODUCTION
Proteins usually perform their functions in a collaborative fashion by interacting with each other. Uncovering the complex structures of protein interaction network is essential for understanding how proteins in a cell function together. Many computational efforts have been made to predict interacting proteins. The gene fusion/Rosetta method [1, 2] predicts a pair of proteins to interact if they are encoded separately as two distinct genes in one organism and are encoded by one single gene (fused) in another organism. Several other algorithms explore the use of protein sequences [3], protein structure [4], phylogenetic profiles [5], protein homology [6], gene neighborhood [7], and gene expression correlation [8] for inferring protein-protein interactions. Those methods are mostly based on protein sequence homology or structure
Data Loading...
