中文摘要：

在B3LYP/6-31G＊水平上采用高斯09全优化计算了50个P450 1A2抑制剂的量子化学参数,应用基于预测的模型变量选择方法（VSMP）选择描述子最佳子集,建立了最高轨道占有能（EHOMO）和分子体积（Vm）与萘、内酯衍生物及其他化合物对细胞色素氧化酶P450 1A2抑制剂的两变量线性QSAR模型,结果表明：所选的2个分子结构描述符与50个抑制剂的活性之间具有很强的线性关系（相关系数r2=0.907 0）和内部预测能力（留多法交叉验证相关系数q2=0.751 7）。同时,将50个化合物分成奇数集和偶数集各自进行筛选建模,并彼此进行外部预测,对全部样本集、奇数集和偶数集样本模型进行了y-Randomization检验,结果表明描述符EHOMO和Vm建立的模型均非常稳定并具有很高的预测能力。

英文摘要：

In order to establish the QSAR model of inhibitors of cytochrome P450 1A2,19 quantum chemical and thermodynamics parameters of 50 compounds,which include naphthalene,lactone and quinoline derivatives,are calculated at B3LYP/6-31G＊ level by density funcional theory method with Gaussian 09 program.Using VSMP（variable selection and modeling based on prediction） technique to select the optimal subset,a two-descriptor QSAR model is constructed.The results of modeling show that there is a strong linear relationship between two descriptors（energies of the highest occupied molecular orbital,EHOMO and the molecular volume,Vm） and the inhibited activity of 50 compounds（correlation coefficient of model,r2=0.907 0）,and high inner predicted ability（correlation coefficient leave-multiple-out cross validation,q2=0.751 7）.Meanwhile,50 compounds are divided into odd set and even set,and two-descriptor QSAR model are constructed with external validated by each other.Then,the y-Randomization validation is performed within the models of all compounds set,odd set and even set.The models of three sets built by EHOMO and Vm have stability and high prediction.