Predicting cell adhesion receptors using protein sequence index

edicting cell adhesion receptors using protein sequence index Original Research (2359) Total Article Views Authors: Wang ZJ, Xu CY, Yu F, Tang L, He JH Published Date June 2011 Volume 2011:3 Pages 97 - 105 DOI: http://dx.doi.org/10.2147/OAB.S17492 Zhijun Wang, Chunyan Xu, Feng Yu, Lin Tang, Jianhua He Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai, People’s Republic of China Abstract: Cell adhesion receptors (CARs) play important roles in signaling, regulation, membrane trafficking, immune response, and transport. For a long time, based on their functional and sequence diversity, CARs have been classified into four classes: cadherin-mediated cell adhesion receptors (CMCARs); immunoglobulin superfamily-mediated cell adhesion receptors (ISMCARs); selectin-mediated cell adhesion receptors (SMCARs); and integrin-mediated cell adhesion receptors (IMCARs). Experimental methods suitable to identify and to determine the kind of CARs are time-consuming. It is, therefore, desirable to explore new methods for predicting CARs directly from protein sequence information. This report shows the application of Protein Sequence Index (PSI) as such a method. Two fuzzy k-nearest neighbor (NN) prediction systems were developed to identify adhesion proteins (APs) and classify APs into different CARs with PSI. In the first fuzzy k-NN predicting system, 619 APs and 1211 nonadhesion proteins (NAPs) were used as a training dataset to identify the APs, and they were evaluated by an independent dataset of 477 APs and 576 NAPs. The computed prediction accuracy was 94.5% and 94.4% for the APs and NAPs respectively, using the independent dataset. In the second fuzzy k-NN predicting system, 1211 noncell adhesion receptors (NCARs), 286 CMCARs, 59 ISMCARs, 38 SMCARs, and 236 IMCARs was used as a training dataset to classify CARs into different types, and they were evaluated by an independent testing dataset of 576 NCARs, 228 CMCARs, 47 ISMCARs, 20 SMCARs, and 182 IMCARs. The predicting accuracy was 94.4%, 92.1%, 95.7%, 95.0%, and 98.9%, for NCARs, CMCARs, ISMCARs, SMCARs, and IMCARs, respectively. These findings suggest the usefulness of PSI for facilitating the identification and classification of CARs. A program, ADHEN, was constructed, which can be used to predict the CARs.