摘要
为辅助开发高活性EP1受体抗拮剂,探讨和研究EP1受体拮抗活性的关键影响因素,选取103个EP1受体抗拮剂分子作为数据集,采用多元线性回归(MLR)法和主成分分析(PCA)法分析每个分子的254个参数进行模拟建模.结果表明,应用MLR和PCA方法都得到了具有良好预测能力的定量构效关系模型.MLR法所建模型结果为:训练集R2=0.77,SEE=0.83,检验集R2=0.74,SEP=0.33.PCA所建模型为:训练集R2=0.72,SEE=0.45,检验集R2=0.71,SEP=0.38.两种方法相比,MLR法所建模型较优,可靠性及预测性强.这些模型及其确定的活性影响参数有助于辅助研发和筛选新型EP1受体抗拮剂.
To develop EPI receptor antagonists with higher activities, the key factors that affect the activities of EP1 antagonists were explored in this study. 103 EP1 antagonists were selected as data sets, and each mol- ecule was calculated based on 254 parameters. Two regression methods of multiple linear regression （MLR） and principal component analysis （PCA） were used. The results show that the quantitative structure - activity relationship models using both the MLR and PCA exhibit good predictive ability. The statistical results by MLR show training set R2 = 0.77, SEE = 0. 83, test set R2 = 0. 74, SEp = 0. 33, and those by PCA show training setR2 = 0.72, SEF = 0.45,test setR2 = 0.71, SEp = 0.38.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2012年第4期121-125,共5页
Journal of Harbin Institute of Technology
基金
国家自然科学基金资助项目(10801025)
关键词
EP1受体抗拮剂
定量构效关系
多元线性回归
主成分分析
EP1 receptor antagonists
quantitative structure-activity relationship
multiple linear regression
principal component analysis
