摘要
将三维原子场作用全息矢量用于表征68个二氢叶酸还原酶抑制剂和48个苦味二肽结构,分别以多元线性回归和偏最小二乘建模,取得优良结果.对前者建模得复相关系数Rm2m=0.893,交互检验相关系数RC2V=0.853,对后者建模得Rm2m=0.936,RC2V=0.849.结果表明3D-HoVAIF能够较好表征两类分子结构,具有物化意义明确及结果易解释特点,值得进一步应用推广.
A novel method based on three-dimensional holographic vector of atomic interaction field （3D-HoVAIF） was employed to express 3D structures of 68 dihydrofolate reductase（DHFR） inhibitors and 48 bitter tasting thresholds （BTT）. Better model was developed based on multiple linear regression （MLR） through a stepwise manner for DHFR and genetic arithmetic-partial least square for BTT, and the following results were obtained： the multiple correlation coefficient （Rmm^2） of 0.893 and 0.936, the cross validated RCV^2 of 0.853 and 0.849 by the leave-one-out procedure, respectively. Satisfactory results showed that information related to biological activities of DHFR and BTT can preferably be expressed by 3D-HoVAIF with definite physical and chemical meaning, easy interpretation, and well understanding, which may become a useful structural expression technique to study quantitative structure-activity relationships （QSAR） for various bioactive substances.
出处
《物理化学学报》
SCIE
CAS
CSCD
北大核心
2006年第3期388-390,共3页
Acta Physico-Chimica Sinica
基金
霍英东基金[98-7-6]
国家春晖计划教育部启动基金[99-1-4/38]
重庆大学研究生创新实践基金[03-5-6
06-01-A]资助项目
