Measuring A Computational Prediction Method For Fast And...

In general, the gap is broadening rapidly between the number of known protein sequences and the number of known protein structural classes. To overcome this crisis, it is essential to develop a computational prediction method for fast and precisely determining the protein structural class. Based on the predicted secondary structure information, the protein structural classes are predicted. To evaluate the performance of the proposed algorithm with the existing algorithms, four datasets, namely 25PDB, 1189, D640 and FC699 are used. In this work, an Improved Support Vector Machine (ISVM) is proposed to predict the protein structural classes. The comparison of results indicates that Improved Support Vector Machine (ISVM) predicts more accurate protein structural class than the existing algorithms. Keywords—Protein structural class, Support Vector Machine (SVM), Naà ¯ve Bayes, Improved Support Vector Machine (ISVM), 25PDB, 1189, D640 and FC699. I. INTRODUCTION (HEADING 1) Usually, the proteins are classified into one of the four structural classes such as, all-ÃŽ ±, all-ÃŽ ², ÃŽ ±+ÃŽ ², ÃŽ ±/ÃŽ ². So far, several algorithms and efforts have been made to deal with this problem. There are two steps involved in predicting protein structural classes. They are, i) Protein feature representation and ii) Design of algorithm for classification. In earlier studies, the protein sequence features can be represented in different ways such as, Functional Domain Composition (Chou And Cai, 2004), Amino Acids