Research of protein structure classification based on rough set and support vector machine

Abstract

A novel method of feature extraction form protein sequences, structures and physicochemical properties has been proposed and obtained a better classification results by the key eigenvector obtained form knowledge reduction combined with the algorithm of support vector machine. Based on Jackknife detecting methods, the comprehensive classification results 78.3% and 90.9% for all-α,all-β,α+β and α/β have been obtained by the method of support vector machine when we tested Z277 and Z498 in database. Moreover, we found that protein physicochemical properties have a strong influence on classification precision of protein structure with Matlab. These results show that the method of feature extraction based on rough set is effective and available, the research of protein structure for support vector machine classification based on rough set is very effective.