用支持向量回归方法预测多氯代二苯并呋喃类化合物的正辛醇/水分配系数

The Prediction to n-Octanol/Water Partition Coefficients of Polychlorinated Dibenzofurans by Support Vector Regression

采用Chemoffice8.0中的MOPAC-AM1算法对135个多氯代二苯并呋喃(PCDFs)类化合物的6种量子化学结构参数进行计算,并将筛选后的量化参数作为PCDFs分子的结构描述符,利用支持向量回归方法建立分子结构与正辛醇/水分配系数间的相关模型并进行预测.将所得结果与逐步多元回归方法的结果进行对比分析,支持向量回归方法所得结果均优于多元回归方法.

The structure parameters of the quantum chemistry for polychlorinated dibenzofurans （PCDFs） compounds were calculated by using the MOPAC-AM1 method in Chemoffice8.0. Some parameters are selected as the structure descriptors of PCDFs compound. The molecular structure and the model of n-octanol/water partition co- efficients are constructed and predicted in support vector regression. The results are compared with the results of multiple regression methods. It can be found that the results of support vector regression are better than the multiple regression methods.