基于DFT和分子连接性指数方法研究醇类化合物的水溶解度和分配系数  

Study of water solubility and partition coefficients of alcohols based on DFT descriptors and molecular connectivity indices

作　　者：崔鹏[1] 崔秀君[2] 郭英娜[2] 袁星[4] 张卓勇[2,3] 

机构地区：[1]吉林大学化学学院,吉林长春130012 [2]东北师范大学化学学院,吉林长春130024 [3]首都师范大学化学系,三维信息获取与应用教育部重点实验室,北京100037 [4]东北师范大学城市与环境学院,吉林长春130024

出　　处：《分子科学学报》2008年第3期180-183,共4页Journal of Molecular Science

基　　金：北京市教育委员会科技发展项目资助(KM200310028105)

摘　　要：将DFT方法计算得到的量化参数和分子连接性指数联合应用到60个醇类化合物的溶解度和辛醇/水分配系数的QSPR研究中,分别通过逐步回归得到具有显著统计意义的4个参数和5个参数的QSPR方程.以此4个参数和5个参数分别作为输入参数,采用BPNN,RBFNN方法建立了QSPR预测模型,使用Latin-partition交叉验证方法评价模型的预测能力.BPNN,RBFNN模型对溶解度预测的相关系数分别为0.993和0.994,而对辛醇/水分配系数预测的相关系数分别0.990和0.997,结果令人满意.Physicochemical properties such as aqueous solubility and 1-oetanol/water partition coefficients of 60 alcohol compounds were predicted based on BPNNs, RBFNNs using four descriptors and five descriptors derived from quantum chemical descriptors calculated by DFT and molecular connectivity indices. Equation including four parameters and five parameters with remarkable statistical meaning were obtained by stepwise multiple regressions, respectively. For aqueous solubility of whole data set, the predictive R of BPNNs model, RBFNNs model based Latin-partition cross-validation is 0.993 and 0. 994, respectively. For 1-octanol/water partition coefficients of whole data set, the predictive R of BPNNs model, RBFNNs model based Latin-partition cross-validation is 0. 990 and 01997, respectively. The predicted results are in very good agreement with the experimental values.

关 键 词：定量结构/性质相关 BP网络 径向基网络 密度泛函 分子连接性指数 醇 

分 类 号：O623[理学—有机化学]