Title: Granular support vector machine to identify unknown structural classes of protein

Authors: Rohayanti Hassan; Razib M. Othman; Zuraini A. Shah

Addresses: Department of Software Engineering, Faculty of Computing, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia ' Department of Software Engineering, Faculty of Computing, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia ' Department of Software Engineering, Faculty of Computing, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia

Abstract: To date, classification of structural class using local protein structure rather than the whole structure has been gaining widespread attention. It is noted that the structural class lies in local composition or arrangement of secondary structure, while the threshold-based classification method has restricted rules in determining these structural classes. As a consequence, some of the structures are unknown. In order to determine these unknown structural classes, we propose a fusion algorithm, abbreviated as GSVM-SigLpsSCPred (Granular Support Vector Machine - with Significant Local protein structure for Structural Class Prediction), which consists of two major components, which are: optimal local protein structure to represent the feature vector and granular support vector machine to predict the unknown structural classes. The results highlight the performance of GSVM-SigLpsSCPred as an alternative computational method for low-identity sequences.

Keywords: local protein structure; structural class prediction; granular SVM; support vector machines; classification; bioinformatics.

DOI: 10.1504/IJDMB.2015.070065

International Journal of Data Mining and Bioinformatics, 2015 Vol.12 No.4, pp.451 - 467

Received: 16 Jul 2013
Accepted: 24 Jun 2014
Published online: 26 Jun 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article