A set of novel structural descriptors (molecular hybridization electronegativity-distance vector, VMEDh) was put forward, and the quantitative structure–activity relationship (QSAR) of a series of 17α-Acetoxyprogest...A set of novel structural descriptors (molecular hybridization electronegativity-distance vector, VMEDh) was put forward, and the quantitative structure–activity relationship (QSAR) of a series of 17α-Acetoxyprogesterones (APs) was investigated. Taking into account the effect of various hybridized orbits on atomic electronegativities, we developed the structure descriptors with amended electronegativities to build a QSAR model. The 10-parameter model based on VMEDh yields a correlation coefficient R=0.972 and standard deviation SD=0.262, which are more desirable than those of the previous molecular electonegativity-distance vector (MEDV-4) (R=0.969, SD=0.275). By stepwise multiple linear regression, several parameters are selected to construct optimal models. The 7-parameter model based on VMEDh has R=0.960 and SD=0.276; its correlation coefficient (RCV) and standard deviation (SDCV) for leave-one-out procedure crossvalidation are respectively RCV=0.890 and SDCV=0.445. The 6-parameter MEDV-4 model has R=0.946, SD=0.304, RCV=0.903 and SDCV=0.406. It is demonstrated that VMEDh has desirable estimation performance and good predictive capability for this series of chemical compounds.展开更多
采用三维全息原子场作用矢量(3D-HoVAIF)对20个TIBO类衍生物抗艾滋病药物进行定量构效关系(QSAR)研究。运用偏最小二乘回归(Partial Least Square Regression,PLS)建模,同时采用内部及外部双重验证的办法对所得模型稳定性能进行深入分...采用三维全息原子场作用矢量(3D-HoVAIF)对20个TIBO类衍生物抗艾滋病药物进行定量构效关系(QSAR)研究。运用偏最小二乘回归(Partial Least Square Regression,PLS)建模,同时采用内部及外部双重验证的办法对所得模型稳定性能进行深入分析和检验。PLS建模的复相关系数(Rcum)、留一法(Leave-one-out,LOO)交互校验(Cross-validation,CV)复相关系数(QCV)和外部样本校验复相关系数(Qext)分别为0.837、0.804、0.812。结果表明,3D-HoVAIF能较好表征TIBO类衍生物抗艾滋病药物分子结构信息,展开更多
基金Funded by Chongqing Medical University Scientific Research Foundation
文摘A set of novel structural descriptors (molecular hybridization electronegativity-distance vector, VMEDh) was put forward, and the quantitative structure–activity relationship (QSAR) of a series of 17α-Acetoxyprogesterones (APs) was investigated. Taking into account the effect of various hybridized orbits on atomic electronegativities, we developed the structure descriptors with amended electronegativities to build a QSAR model. The 10-parameter model based on VMEDh yields a correlation coefficient R=0.972 and standard deviation SD=0.262, which are more desirable than those of the previous molecular electonegativity-distance vector (MEDV-4) (R=0.969, SD=0.275). By stepwise multiple linear regression, several parameters are selected to construct optimal models. The 7-parameter model based on VMEDh has R=0.960 and SD=0.276; its correlation coefficient (RCV) and standard deviation (SDCV) for leave-one-out procedure crossvalidation are respectively RCV=0.890 and SDCV=0.445. The 6-parameter MEDV-4 model has R=0.946, SD=0.304, RCV=0.903 and SDCV=0.406. It is demonstrated that VMEDh has desirable estimation performance and good predictive capability for this series of chemical compounds.
文摘采用三维全息原子场作用矢量(3D-HoVAIF)对20个TIBO类衍生物抗艾滋病药物进行定量构效关系(QSAR)研究。运用偏最小二乘回归(Partial Least Square Regression,PLS)建模,同时采用内部及外部双重验证的办法对所得模型稳定性能进行深入分析和检验。PLS建模的复相关系数(Rcum)、留一法(Leave-one-out,LOO)交互校验(Cross-validation,CV)复相关系数(QCV)和外部样本校验复相关系数(Qext)分别为0.837、0.804、0.812。结果表明,3D-HoVAIF能较好表征TIBO类衍生物抗艾滋病药物分子结构信息,
