基于SVM和KNN的蛋白质耐热性分类

Classification of protein thermostability using Support Vector Machines and K-Nearest Neighbors

以氨基酸含量为特征向量,研究了SVM和KNN预测蛋白质耐热性的准确度。结果表明,基于SVM的分类效果较好,其局部预测率和全局预测率分别为82.4%和83.4%;而基于KNN方法的局部预测率和全局预测率分别为77.6%和79.9%。两种方法的预测率均表明氨基酸含量是影响蛋白质耐热性的主要因素。

Regarding amino acid composition as eigenvector,protein thermostability is classified using Support Vector Machines and K-Nearest Neighbors.h is found that the result of using support vector machines is better than K-Nearest Neighbors.The local accuracy and global accuracy are 82.4% and 83.4% respectively.But the local accuracy and global accuracy are 77.6% and 79.9% respectively using K-Nearest Neighbors.The prediction accuracy of two kinds of methods can both prove that the amino acid composition is the main factor that influences the protein thermostability.