Atomic electronegativity interaction vector (AEIV) and atomic hybridization state index (AHSI) were used for establishing the quantitative structure-spectroscopy relationship(QSSR) model of 13C NMR chemical shifts of ...Atomic electronegativity interaction vector (AEIV) and atomic hybridization state index (AHSI) were used for establishing the quantitative structure-spectroscopy relationship(QSSR) model of 13C NMR chemical shifts of isodon diterpenoid compounds.Multiple linear regression (MLR) and computational neural network (CNN) were used to create the models,and the estimation stability and generalization ability of the models were strictly analyzed by both internal and external validations.The established MLR and CNN models were correlated with experimental values and the correlation coefficients of model estimation,leave-one-out (LOO)cross-validation (CV),and predicted values of external samples were Rcum=0.9724,RCV=0.9723,Qext=0.9738 (MLR);Rcum=0.9957,Qext=0.9956 (CNN),respectively.The results indicated that CNN gave significantly better prediction of 13C NMR chemical shifts for isodon diterpenoids than MLR.Satisfactory results showed that AEIV and AHSI were obviously good for modeling 13C NMR chemical shifts of isodon diterpenoid compounds.展开更多
用原子电性距离矢量(atom ic electronegativity d istance vector,AEDV)和原子杂化状态指数(atom ic hybrid i-zation state index,AHSI)对13个雄甾烯酮化合物中247个碳原子进行了结构表征,并与其核磁共振碳谱(13CNMR)建立了多元线性...用原子电性距离矢量(atom ic electronegativity d istance vector,AEDV)和原子杂化状态指数(atom ic hybrid i-zation state index,AHSI)对13个雄甾烯酮化合物中247个碳原子进行了结构表征,并与其核磁共振碳谱(13CNMR)建立了多元线性定量构谱相关(QSSR)模型;运用逐步回归结合统计检测,对模型变量进行了筛选,建模计算值、留一法(leave-one-out,LOO)交互校验(cross-validation,CV)预测值和留分法(leave-molecu le-out,LMO)交互校验预测值的复相关系数(R)分别为0.989 6,0.989 1和0.989 4。结果表明:AEDV,AHSI与13CNMR谱化学位移显著相关。展开更多
文摘Atomic electronegativity interaction vector (AEIV) and atomic hybridization state index (AHSI) were used for establishing the quantitative structure-spectroscopy relationship(QSSR) model of 13C NMR chemical shifts of isodon diterpenoid compounds.Multiple linear regression (MLR) and computational neural network (CNN) were used to create the models,and the estimation stability and generalization ability of the models were strictly analyzed by both internal and external validations.The established MLR and CNN models were correlated with experimental values and the correlation coefficients of model estimation,leave-one-out (LOO)cross-validation (CV),and predicted values of external samples were Rcum=0.9724,RCV=0.9723,Qext=0.9738 (MLR);Rcum=0.9957,Qext=0.9956 (CNN),respectively.The results indicated that CNN gave significantly better prediction of 13C NMR chemical shifts for isodon diterpenoids than MLR.Satisfactory results showed that AEIV and AHSI were obviously good for modeling 13C NMR chemical shifts of isodon diterpenoid compounds.
文摘用原子电性距离矢量(atom ic electronegativity d istance vector,AEDV)和原子杂化状态指数(atom ic hybrid i-zation state index,AHSI)对13个雄甾烯酮化合物中247个碳原子进行了结构表征,并与其核磁共振碳谱(13CNMR)建立了多元线性定量构谱相关(QSSR)模型;运用逐步回归结合统计检测,对模型变量进行了筛选,建模计算值、留一法(leave-one-out,LOO)交互校验(cross-validation,CV)预测值和留分法(leave-molecu le-out,LMO)交互校验预测值的复相关系数(R)分别为0.989 6,0.989 1和0.989 4。结果表明:AEDV,AHSI与13CNMR谱化学位移显著相关。
