摘要
目的运用定量-结构活性关系方法研究离子液体理化结构及其与青海弧菌Q67毒性的关系。方法选取30个离子液体,随机分为24个训练集,6个测试集。以离子液体对青海弧菌Q67微板毒性分析(MTA)方法测定的百分发光抑制率(EC50)值为毒性指标,取EC50负对数值(pEC50)作为本次研究的实验值。应用Hyperchem软件先进行MM+分子力学优化,再采用半经验的方法对其进行几何优化,最终应用CODESSA软件计算离子液体的分子描述符,包括构成、拓扑、几何、静电和量子化学描述符。用启发式算法(HM)筛选描述符参数,在此基础上建立多元线性回归模型并运用基因表达式编程算法(GEP)建立非线性预测模型。结果 HM算法训练集和测试集相关系数分别为0.95和0.91,GEP算法训练集和测试集相关系数分别为0.97和0.88。结论两种算法对离子液体毒性预测均有良好的可靠性,GEP算法在离子液体毒性预测中优于HM算法。
Objective To investigate the physical and chemical structures of ionic liquids using the quantitative structureactivity relationship method, as well as its association with the toxicity of Vibrio Qinghaiensis sp.-Q67. Methods A total of 30 ionic liquids were randomly divided into training set with 24 liquids and test set with 6 liquids. The luminescence inhibition rate （EC50） of Vibrio Qinghaiensis sp.-Q67 measured by microplate toxicity analysis using ionic liquids was used as the toxicity index, and the negative logarithmic value of EC50 （pECs0） was used as the experimental value of this study. The Hyperchem software was used for MM-t- molecular mechanics optimization, a semi-empirical approach was used for geometry optimization, and then the CODESSA software was used to calculate the molecular descriptors of ionic liquids, including constitutional, topological, geometrical, electrostatic, and quantum chemical descriptors. The heuristic method （HM） was used for the screening of descriptor para meters, and then a multiple linear regression model was established and gene expression programming （GEP） was used to establish a nonlinear prediction model. Results HM achieved a correlation coefficient of 0.95 in the training set and 0.91 in the test set, and GEP achieved a correlation coefficient of 0.97 in the training set and 0.88 in the test set. Conclusion Both algorithms are reliable in predicting the toxicity of ionic liquids, and GEP has a better predictive effect than HM.
出处
《青岛大学医学院学报》
CAS
2017年第3期312-316,319,共6页
Acta Academiae Medicinae Qingdao Universitatis
关键词
离子液体
定量结构-活性关系
算法
毒性
ionic liquids
quantitative structure-activity relationship
algorithm
toxicity
