摘要
基于氨基酸描述子ISA-ECI,利用加和法构建了101个混杂肽高效液相色谱保留时间的支持向量回归(SVR)预测模型(ε=0.001、σ=4和C=100),并与多元线性回归和投影寻踪回归作了比较研究,结果表明SVR要好于其它方法,SVR方法对数据集留一法交叉检验和拟合的预测复相关系数分别为0.8822和0.9530,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保留时间与描述肽的分子结构信息的ISA-ECI参数存在着非线性关系.
Based on amino acid descriptors(ISA-ECI) and addition method, a prediction model of high-performance liquid chromatography retention time of 101 promiscuous peptides by support vectors regression (SVR) has been set up. The prediction result of the SVM model (ε =0. 001, σ = 4 and C = 100) is much better than that obtained by multiple linear regression(MLR) and project pursuit regression(PPR) methods. The prediction correlation coefficient is 0.882 2 by leave-one-out cross validation. The prediction correlation coefficient is 0. 953 in fitting calculation. The prediction results are in agreement with the experimental values. This paper provides a simple and effective method for predicting to the retention time of peptidcs. Moreover, it also indicates the nonlinear relation between the retention time of pcptidcs and their structural descriptors(ISA-ECI).
出处
《四川师范大学学报(自然科学版)》
CAS
CSCD
北大核心
2007年第1期94-98,共5页
Journal of Sichuan Normal University(Natural Science)
基金
四川农业大学生命理学院创新基金的资助.
关键词
肽
定量结构-保留关系
支持向量机
ISA-ECI参数
Pepfide
Quantitative structure-retention relationship
Support vector machine
ISA-ECI parameter