关键词: 蛋白质结构类预测, 自相关系数, 伪氨基酸组成（PseAAC）, 支持向量机（SVM）

Abstract: In the traditional prediction methods, only the composition of amino acids was taken into account in constructing feature vector. While both the position and interaction of the amino acids which are at the different locations can be reflected well by the correlation coefficient. Firstly, this paper designs a method which combines amino acid composition and correlation coefficient. Secondly, on the basis of the pseudo-amino acid composition (PseAAC) model proposed by Chou, this paper reconstructs the PseAAC model by extending the information, and combines the PseAAC model and autocorrelation coefficient to construct feature vector. Using the two new methods for coding, several experiments are conducted on the datasets Z277, Z498 and the independent test sets D138 with the prediction tool of support vector machine. The experimental comparison results show that the accuracy of the new method can improve 7.43% and 8.53% on average than the traditional amino acid composition method, which proves that the new method is more effective.

Key words: protein structure class prediction, autocorrelation coefficient, pseudo-amino acid composition (PseAAC), support vector machine (SVM)