四氢咪唑苯二氮卓酮类抗艾滋病药物定量构效关系的研究

采用三维全息原子场作用矢量(3D-HoVAIF)研究89个四氢咪唑苯二氮卓酮(TIBO)类抗艾滋病药物的定量构效关系(QSAR).分别运用偏最小二乘回归、人工神经网络建模,同时采用内部及外部双重验证的办法深入分析和检验模型的稳定性.PLS与ANN建模的复相关系数(R2cum)、留一法(leave-one-out,LOO)交互校验(cross-validation,CV)复相关系数(Q2CV)和外部样本校验复相关系数(Q2ext),分别为0.802、0.710、0.552和0.871、0.864、0.760.表明用3D-HoVAIF表征TIBO类抗艾滋病药物分子结构信息较好,建立QSAR模型的稳定性和预测能力良好,运用ANN建模优于PLS及前人报道的多元线性回归(multiple linear regression,MLR).

Quantitative structure activity relationship study of anti-HIV drug tetrahydroim-idazobenzodiazepinone

A newly developed three-dimensional holographic vector of atomic interaction field(3D-HoVAIF)was used to describe the chemical structures of 89 anti-HIV drugs as tetrahydroimidazobenzodiazepinones(T1BO).Here two quantitative structure activity rela- tionship(QSAR)models were built by partial least square regression(PLS)and artificial neural network(ANN).The estimation sta- bility and generalization ability of these models were strictly analyzed by both internal and external validations.The correlation coeffi- cient(R~2)of established PLS and ANN models,leave-one-out(LOO)cross-validation(CV),predicted values versus experimental ones of external samples were 0.802,0.710,0.552 and 0.871,0.864,0.760,respectively.The results indicated that results of ANN better than those of PLS here and MLR in reference,which had both favorable estimation stability and good prediction capabili- ties.Satisfactory results showed that 3D-HoVAIF could preferably express information related to biological activity of TIBO.