密度泛函理论方法用于部分农药Kow的QSPR模型研究

QSPR study on Octanol/Water partition coefficients of pesticides

农药的正辛醇/水分配系数(lgKow)是研究农药在水中化学行为的重要参数。本文采用摇瓶法测得25℃时10种农药的正辛醇/水分配系数(lgKow),采用密度泛函理论(DFT)中的B3LYP方法,在6-31G*基组上全优化计算得到了10种农药的结构参数和热力学参数。在10种化合物的lgKow实验数据基础上,采用GQSARF2.0程序,以结构参数和热力学参数作为理论描述符,得到预测lgKow的相关模型,其方程为lgKow=0.108-0.191μ-2.451H/1000,决定系数R2为0.9443。用变异膨胀因子(Variance Inflation Factors,VIF)和t值对模型进行了验证,结果显示模型具有较好的稳定性。而交叉验证法的相关系数q2为0.9269,证明所建模型具有较强的预测能力。用所得模型预测了10种农药的lgKow,与实验测定值相比,方程的预测值与实验值都较接近,可见预测值较可靠,所建模型可用于同类农药lgKow的预测。

This paper is concerned about our study on the Octanol/ Water partition coefficients of pesticides through experimental determination of the values of 10 pesticides by using flask-shaking methods at 25℃ . In doing so, optimized calculations were carried out of the compounds at B3LYP/6 - 31G^＊ level in Gaussian 03 program by means of density functional theory. Detailed work over the structural and thermodynamic parameters has made us possible to obtain the essential output from Gaussian 03 program, which has, in turn, enabled us to develop the structural and thermodynamic parameters as theoretical descriptors, correlation equation between experimental data lgKow and such parameters of 10 pesticides by using the quantitative structure-property relationship with GQSARF 2.0 program. By nature, the optimum equation we have gained may include two parameters （ standard enthalpies （ H^θ ） and dipole moment （ μ ） ） , in which the conventional correlation coefficient （R^2 ） proves to be 0.944 3. Analyzing the model, it is possible for us to reach the following interpretations： （1） The greater standard enthalpy value （which is negative） , the bigger the lgKow will be; （2） Since lgKow is inversely proportional to the dipole moment, large values of the dipole moment would inevitably result in small value of lgKow ; （3） Standard enthalpies tend to exhibit remarkable iniquenee on lgKow and dipole moment has relatively little effect on lgKow. This shows that the molecules would increase in number with the increase of the volume （Vm） and that of the average value of polarizability （a）. And since the molecules tend to dissolve in the organic phase, high lgKo. would be followed by, with the model further validated by the variance inflation factors （VIF） and t-test. In addition, the results of cross-validation also indicate that the model is endowed with high potential power for prediction, It is thus suggested that the present model can be used to nrediet the existence of lgKiw of pesticides with great certaintv,