基于PRIN描述子的肽毛细管区带电泳迁移率支持向量回归预测模型  

Modeling prediction of electrophoretic mobilities of peptides in capillary zone electrophoresis by support vectors regression based on amino acid descriptors(PRIN)

作　　者：印家健[1] 邹平[1] 崔扬健[1] 

机构地区：[1]四川农业大学生命理学院,雅安625014

出　　处：《黑龙江大学自然科学学报》2007年第4期516-521,共6页Journal of Natural Science of Heilongjiang University

基　　金：四川农业大学生命科学与理学院科技创新基金资助

摘　　要：基于氨基酸描述子PRIN,利用加和法构建了102个混杂肽(从2肽到14肽)毛细管区带电泳迁移率的支持向量回归(SVR)预测模型(ε=0.012、σ=5和C=5),并与多元线性回归作了比较研究,结果表明SVR要好于多元线性回归方法,SVR方法对训练集和预测集的预测残差平方和(PRESS)分别是0.0097和0.0135及预测复相关系数分别为0.9774和0.9419,其预测结果与实验值一致,且提出了一个简单的方法并指出迁移率与描述肽的分子结构信息的PRIN参数存在着非线性关系。Based on amino acid descriptors （PRIN） and additive method, modeling prediction of electrophoretic mobilities of 102 promiscuous peptides （from 2 peptides to 14 peptides） in capillary zone electrophoresis by support vectors regression （SVR）. The prediction result of the SVM model （ε=0.012、σ=5 and C =5） is much better than that obtained by multiple linear regression （MLR） method. The PRESS error of the training set and the test set is 0. 0097,0. 0135 and the prediction correlation coefficient is 0. 9774 and 0. 9419, respectively. The prediction results are in agreement with the experimental values. This paper provided a simple and effective method for predicting the electrophoretic behavior of peptide and some insight into what structural features are related to the electrophoretic mobility of peptides. Moreover, it also offered an idea about nonlinear relation between electrophoretic mobility of peptides and their structural descriptors （PRIN）.

关 键 词：肽 定量结构一迁移率关系 支持向量机 PRIN参数 预测 

分 类 号：O629[理学—有机化学] R914[理学—化学]