几种QSAR建模方法的研究进展与应用  

Research Progress and Application in the Several QSAR Modeling Method

作　　者：冯丽萍[1] 郭正元[1] 梁菁[1] 周井刚[1] 

机构地区：[1]湖南农业大学农业环保研究所,湖南长沙410128

出　　处：《农业环境科学学报》2007年第B10期651-655,共5页Journal of Agro-Environment Science

摘　　要：定量构效关系(QSAR)对药物设计和新药研制、环境毒物的毒性评价与预测有着显著的作用,本文简单介绍了几种QSAR建模方法:多元线性回归(MLR)、主成分分析(PCA)、偏最小二乘法(PLS)、人工神经网络(ANN)和支持向量机(SVM),并对这几种不同的建模方法在实际中的应用进行举例。可以看出PLS和ANNS是优秀的建模方法,预测能力强,SVM通过结构风险最小化原则建模,有效将期望风险降至最低,模型预测力得到显著提高,在环境毒物评价中具有广阔的应用前景。Quantitative structure - activity relationship of the design and ：levelopment of new drugs, toxicity evaluation and prediction of toxic environment plays a role. This paper briefly described several QSAR modeling methods ： multiple linear regression, principal component analysis, partial least squares, artificial neural network and support vector machine, and that several different modeling methods for the application in practice. It showed that the PLS and ANN were excellent modeling methods, predictive capability was very good, SVM modeling through structural risk minimization principle, to effectively minimized the risks expectations that the model prediction had improved significantly, and the environmental toxicology evaluation and prediction had broad application prospects.

关 键 词：定量构效关系 多元线性回归 偏最小二乘法 支持向量机 人工神经网络 

分 类 号：X592[环境科学与工程—环境工程]