咪唑鎓盐类双子表面活性剂的构效关系研究  

The quantitative model construction of structure-cmc relationship for gemini imidazolium surfactants

作　　者：刘友权[1] 郭延芝[2] 唐永帆[1] 孙川[1] 原励[1] 张燕[1] LIU You-quan;GUO Yan-zhi;TANG Yong-fan;SUN Chuan;YUAN Li;ZHANG Yan(Research Institute of Natural Gas Technology,PetroChina Southwest Oil&Gas Field Company,Chengdu 610213,China;College of Chemistry,Sichuan University,Chengdu 610064,China)

机构地区：[1]中国石油西南油气田分公司天然气研究院,四川成都610213 [2]四川大学化学学院,四川成都610064

出　　处：《化学研究与应用》2019年第7期1284-1292,共9页Chemical Research and Application

基　　金：中国石油天然气股份有限公司重大科技专项基金(2016E-0609)

摘　　要：阐明表面活性剂结构与其临界胶束浓度(critical micelle concentration,cmc)之间的内在关联,并构建机器学习方法模型实现cmc的自动预测,对了解其表面活性的分子机制和设计合成具有潜力的新型表面活性分子具有重要意义。本研究从咪唑鎓盐类双子表面活性剂的3-D结构出发,计算得到大量的能量、电荷、立体构象与拓扑参数等方面的结构描述符。然后利用方差阈值法(variance threshold)对特征变量进行评估与筛选,以剔除冗余变量,实现数据降维。最后采用支持向量回归(Support Vector Regression,SVR)构建咪唑鎓盐类双子表面活性剂的结构-cmc的关系模型。根据测定cmc实验方法的不同,分别构建了两个特异性模型:表面张力法的SVR模型与电导法的SVR模型,其模型的留一法交叉验证结果令人满意,相关系数R分别为0.8791与0.9316,均方根误差(RMSE)分别为0.36与0.34。结果表明,方法可靠可行,可实现对咪唑鎓盐类双子表面活性剂的cmc的定量预测。Elucidating the internal relationship between structure and critical micelle concentration(cmc)and the automatic prediction of cmc values using machine-learning methods are both essential to further understand the molecular mechanism of surfactants and design the new potential surfactant molecules.Based on the 3-D structures of gemini imidazolium surfactants,amount of structural descriptors,including energy,charge,spatial and topological parameters were computed for each molecule.Then variance threshold was used to evaluate and screen these descriptor variables in order to remove redundant ones.Finally,support vector regression(SVR)was applied to construct the quantitative structure-cmc relationship model.Here,according to different experimental methods,two specific models were built.One is the surface tension method-specific model and the other is the conductivity-specific model.Two modelswere tested by leave-one-out cross validation and yield the satisfactory results with the correlation coefficient(R)of 0.8791 and 0.9316 and the root mean square error(RMSE)of 0.36 and 0.34,respectively.The results indicate that the method proposed by us is reliable and feasible and it can be used for quantitativeprediction of cmc values for the gemini imidazolium surfactants.

关 键 词：咪唑鎓盐类双子表面活性剂 临界胶束浓度 定量结构-cmc关系模型 方法阈值法 支持向量回归 

分 类 号：O651[理学—分析化学]