Domain-Based Predictive Models for Protein-Protein Interaction Prediction
- PDF / 432,073 Bytes
- 8 Pages / 600.03 x 792 pts Page_size
- 3 Downloads / 238 Views
Domain-Based Predictive Models for Protein-Protein Interaction Prediction Xue-Wen Chen and Mei Liu Bioinformatics and Computational Life-Sciences Laboratory, Information and Telecommunication Technology Center, Department of Electrical Engineering and Computer Science, The University of Kansas, 1520 West 15th Street, Lawrence, KS 66045, USA Received 4 May 2005; Revised 8 September 2005; Accepted 15 December 2005 Protein interactions are of biological interest because they orchestrate a number of cellular processes such as metabolic pathways and immunological recognition. Recently, methods for predicting protein interactions using domain information are proposed and preliminary results have demonstrated their feasibility. In this paper, we develop two domain-based statistical models (neural networks and decision trees) for protein interaction predictions. Unlike most of the existing methods which consider only domain pairs (one domain from one protein) and assume that domain-domain interactions are independent of each other, the proposed methods are capable of exploring all possible interactions between domains and make predictions based on all the domains. Compared to maximum-likelihood estimation methods, our experimental results show that the proposed schemes can predict protein-protein interactions with higher specificity and sensitivity, while requiring less computation time. Furthermore, the decision tree-based model can be used to infer the interactions not only between two domains, but among multiple domains as well. Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.
1.
INTRODUCTION
Proteins play an essential role in nearly all cell functions such as composing cellular structure, promoting chemical reactions, carrying messages from one cell to another, and acting as antibodies. The multiplicity of functions that proteins execute in most cellular processes and biochemical events is attributed to their interactions with other proteins. It is thus critical to understand protein-protein interactions (PPIs) involved in a pathway or a cellular process in order to better understand protein functions and the underlined biological processes. PPI information can also help predict the function of uncharacterized proteins based on the classification of known proteins within the PPI network. Furthermore, a complete PPI map may directly contribute to drug development as almost all drugs are directed against proteins. The recent development of high throughput technologies has provided experimental tools to identify PPIs systematically. These methods include two-hybrid system [1], mass spectrometry [2], protein chips [3], immunoprecipitation [4], and gel-filtration chromatography [5]. Protein interactions can also be measured by biophysical methods such as analytical ultracentrifugation [6], calorimetry [7], and optical spectroscopy [8]. Among those experimental methods, the two-hybrid system is mature and accurate enough to be
used for obtaining the full protein interaction networks of Saccharomyces ce
Data Loading...
