二芳基甲烷衍生物保留因子与分离因子的神经网络理论研究

Theoretical Research on Retention and Separation Factors of Diarylmethane Derivatives by Neural Network Method

为了研究手性二芳基甲烷衍生物的保留因子和分离因子,基于分子结构及邻接矩阵,计算了63个手性二芳基甲烷衍生物的分子连接性指数和电性拓扑状态指数。建立了这些分子保留因子、分离因子与优化得到的结构指数之间的相关性模型,并将筛选的结构参数作为BP神经网络的输入层变量,采用不同的网络结构,获得了令人较为满意的三个预测模型,模型的总相关系数R分别为0.981、0.972和0.992。利用模型计算得到的保留因子和分离因子预测值与实验值的平均误差分别为0.041、0.042和0.010,吻合度较好。结果表明手性二芳基甲烷衍生物的保留因子及分离因子与分子结构参数之间有良好的非线性关系。

In order to study retention factors and separation factors of the chiral diaryhnethane derivatives, the molecular connectivity index and electrotopologieal state index of 63 chiral diarylmethane derivatives were calculated based on the location of molecular structure and conjugation matrix. The relationship model between the retention factors, separation factors and optimized molecular structure parameters of these compounds were developed. Using the structural parameters as the input variables of the neural network, we constructed three satisfying QSRR models with back-propagation algorithm, whose network structure were different. The total correlation coefficient R was 0. 981, 0. 972 and 0. 992 respectively. The mean deviation between the experimental and the predicted values of lgk2,lgkj and lgα was 0. 041, 0. 042 and 0. 010 respectively. The results showed that there was good nonlinear relationship between the lgk2, lgk1 , lgα and the structural parameters.