介绍了虚拟化学实验室网站提供的10种常用分配系数(正辛醇/水)预测软件(ALOGPs、AC logP、AB/LogP、COSMOFraq、miLogP、ALOGP、MLOGP、KOWWIN、XLOGP2、XLOGP3)。以美国环境保护局《国家推荐水质标准》中的105种有机污染物作为测试集,使用成对t检验和线性回归的方法比较了10种软件的预测值与实验值的差异性。成对t检验结果表明,ALOGPs、AC logP、AB/LogP和KOWWIN软件的预测值和实验值没有显著的差别,对应p-水平分别为0.26517、0.08149、0.43717和0.89145。线性回归的结果表明,ALOGPs和KOWWIN软件的预测值和实验值相关性最好,相关系数分别为0.9919和0.9884,拟合曲线的斜率分别为0.9765和0.9982,截距分别为0.0564和0.0102。比较而言,ALOGPs软件界面友好,操作方便,输入方式多样,通过Internet免费运行并且同时提供多种软件预测值作为参考等优点,因此其具有更大的优势。

Ten kinds of general partition coefficient (octanol/water) prediction software (ALOGPs, AC logP, AB/LogP, COSMOFraq, miLogP, ALOGP, MLOGP, KOWWIN, XLOGP2, and XLOGP3) provided by Virtual Computational Chemistry Laboratory is introduced. Using 105 kinds of organic pollutants listing in the National Recommended Water Quality Criteria of USEPA as a test set, paired t-test and liner regression method are performed on the predictive values and experimental values. The paired t-test results show that there is no significant difference between predictive values and experimental values for ALOGPs, AC logP, AB/LogP, and KOWWIN. The corresponding p-values are 0.26517, 0.08149, 0.43717, and 0.89145, respectively. The results of liner regression method indicate that the ALOGPs and KOWWIN are more accurate predictors than other software. For ALOGPs, the correlation coefficient between predictive values and experimental values is 0.9919. The slope and the intercept of the fitting curve are 0.9765 and 0.0564, respectively. For KOWWIN, the correlation coefficient is 0.9884. The slope and the intercept of the fitting curve are 0.9982 and 0.0102, respectively. Since the ALOGPs has a more user friendly interface, more convenient operation, more methods for entering input data, is able to free online perform, and provides the average lgP value of 10 kinds of software, therefore the ALOGPS has advantage over KOWWIN.