几种QSAR建模方法的研究进展与应用.pdfVIP

414 第二届全国农业环境科学学术研讨会论文集 2007年7月 几种QSAR建模方法的研究进展与应用 冯丽萍，郭正元，梁菁，周井刚 (湖南农业大学农业环保研究所，湖南长沙410128) 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 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 环境毒物砰价中具有广阔的应用前景。 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 Research in and theseveral method progressapplication QSARmodeling FENG Li—ping,GUO Zheng—yuml，LIANGJing，ZHOUJing—gang (InatitoteofAm-EnvironmentalPn=,to．ction，HnnanA—culturalUniversity，Changsha410128，ChixⅢ) ofthe and ofDeW evaluationand Abstract：Quantitativeatructure-aclivityrelationsMpdesigndevelopment drugs，toxicity oftoxic a describesseveral Linear environmentrole．This prediction plays paperbriefly QSARmodelingmethods：Muhiple Least NeuralNetworkand Vector Regression，PrincipalComponentAnalysis，PartialSquares，Artificial Support Machine,and that severaldifferent methodsforthe in Call∞ethattheP坞andANNareexcellent m·xteling 8ppucationpractice．We modeling is methods，Pmdiefive structuralriskminimization minimize capabilityverygood，SVMmodelingthrough principle，toeffectively therisks thatthemodel has the Envlronmemalevaluationand expectatiom predictionimprovedslgnJficandy，and Toxiexdogy hasbroad prediction applicationprospects． Structure Linear Least Vector Keywords：QuantitativeActivityRelationship；MuhipleRegression；Partial Squares；Supporl