期刊
JOURNAL OF HAZARDOUS MATERIALS
卷 278, 期 -, 页码 320-329出版社
ELSEVIER
DOI: 10.1016/j.jhazmat.2014.06.018
关键词
QSAR; Multiple linear regression (MLR); Support vector machine (SVM); Toxicity; Ionic liquids
资金
- Key Program of National Natural Science Foundation of China [21036007]
- National Basic Research Program of China (973 Program) [2013CB733506]
- Key Technologies R&D Program of China [2013BAC11B02]
- Chih-Jen Lin team (National Taiwan University, Taiwan)
A comprehensive database on toxicity of ionic liquids (ILs) is established. The database includes over 4000 pieces of data. Based on the database, the relationship between IL's structure and its toxicity has been analyzed qualitatively. Furthermore, Quantitative Structure Activity relationships (QSAR) model is conducted to predict the toxicities (EC50 values) of various ILs toward the Leukemia rat cell line IPC-81. Four parameters selected by the heuristic method (HM) are used to perform the studies of multiple linear regression (MLR) and support vector machine (SVM). The squared correlation coefficient (R-2) and the root mean square error (RMSE) of training sets by two QSAR models are 0.918 and 0.959, 0.258 and 0.179, respectively. The prediction R-2 and RMSE of QSAR test sets by MLR model are 0.892 and 0.329, by SVM model are 0.958 and 0.234, respectively. The nonlinear model developed by SVM algorithm is much outperformed MLR, which indicates that SVM model is more reliable in the prediction of toxicity of ILs. This study shows that increasing the relative number of O atoms of molecules leads to decrease in the toxicity of ILs. (C) 2014 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据