Абстрактный

Prediction of four kinds of supersecondary structures in enzymes by using svm based on scoring function

Sujuan Gao, Xiuzhen Hu

Enzymes are a kind of protein that has catalytic function, the study of supersecondary structures in enzymes plays an important role in the structure and function of enzymes. Based on enzyme sequence information, four kinds of supersecondary structures in enzymes were researched for the first time. Amino acids of sites and dipeptide components of sites were selected as parameters, the predictive results were not ideal by using scoring function method; the better performance was obtained by using support vector machine (SVM), 40 scores for five selections of the best fixed-length pattern were selected as input parameters, the overall prediction accuracy in 7-fold cross-validation was 81.2%and the Matthews correlation coefficient was above 0.70. Therefore, SVM based on scoring function is an effective method to predict four kinds of supersecondary structures in enzymes.