支持向量回归方法用于取代芳烃急性毒性的预测  

Support vector regression for the prediction of acute toxicities of chloro-aromatic compounds for scenedesmus obliquus

作　　者：戴雨茹[1] 陆文聪[1] 严悦然[1] 李国正[2] 包伯荣[1] 

机构地区：[1]上海大学理学院化学系,上海200444 [2]上海大学计算机科学与工程学院,上海200072

出　　处：《计算机与应用化学》2007年第1期74-76,共3页Computers and Applied Chemistry

基　　金：国家自然科学基金(20503015)

摘　　要：采用支持向量回归方法建立了取代芳烃对斜生栅列藻的急性毒性值(-lgEC_50)与分子体积(v)、近似表面积(ASA)、网格表面积(GSA)、水化能(HE)、疏水系数(LogP)间的定量关系,用留一法预报了取代芳烃对斜生栅列藻的急性毒性值(- lgEC_50),相对误差绝对值平均为5．803％,其结果优于偏最小二乘法建模所得结果(相对误差绝对值平均为9．49％)。In this paper, support vector regression （SVM） was applied to QSAR study of 18 chloro-aromatic compounds. The molecular descriptors include approximate surface area, grid surface area, volume, hydrated energy and LogP. Based on the leave-one-out cross validation results, the mean relative error （MRE） SVR model is 5.803%, while that of PLS model is 9.49% for the prediction of acute toxieities of Scenedesmus Obliquus. The result indicates SVR is suitable to find the regularities from the small data set, i.e. data set with fewer samples, giving resuits of modeling with good generalization ability.

关 键 词：支持向量回归 取代芳烃 斜生栅列藻 急性毒性值预测 

分 类 号：TQ24[化学工程—有机化工]