支持向量回归用于氨基酸描述符在肽QSAR建模中的性能评价

Evaluation of Predict Performanceof Five Amino Acid Descriptors in Peptide QSAR by Support Vector Regression

采用支持向量回归方法用3个数据集来评价z-scales、c-scales、ISA-ECI、MS-WHIM、PRIN等5个氨基酸描述符在肽QSAR支持向量回归模型构建中的性能并对核函数进行选择,采用留一法交叉检验的结果显示径向基核函数要好于多项式核函数和线性核函数;在以径向基核函数的支持向量回归模型中表明z-scales的预测准确度要略优于其它描述符,且在同一描述符的情况下SVR的预测效果要好于其它线性方法,说明SVR在肽QSAR模型构建中是一个可行的方法.

Evaluation of predict performance of five amino add descriptors （z-scales, e-scales, ISA-ECI, MS- WHIM, PRIN） in peptide QSAR（Quamitative structure-activity relationships） with three dataset by support vector regression（SVR） is made in the artiele, and RBF is selected as kernel function. Using ＇leave-one-out＇ ert^-validation （LOO-CV）, we suppose that radial basis function （RBF） is better than polynomial function and linear function, as long as our model is considered. Predicting aeeuraey of z-scales is slightly better than the other descriptors in SVR with RBF. Prediction capability of SVR in the same descriptor is better than other linear methods. Therefore. SVR is assumed to be a feasible method in peptide OSAR model.