摘要
将支持向量机分类方法用于醚菊酯类似物构效关系的研究,所用分子描述符为物理化学参数,包括该类化合物的两个取代基的Hammett 常数σ_A、σ_B,摩尔折射M_A、M_B,疏水值之和П。建立了醚菊酯类似物生物活性预报的支持向量机预报模型,其留一法交叉检验的预报正确率为90.3%。将支持向量机的预报结果与主成分分析(PCA)、人工神经网络(ANN)、最近邻(KNN)的预报结果进行比较,所得SVM 的预报正确率高于PCA、ANN、KNN 的结果。因此,SVM 方法有望成为研究药物构效关系的有力工具。
The support vector classification(SVC),as a novel approach,was employed to discriminate between high and low activitiesof ethofenprox analogous based on the molecular descriptors,which were physicochemical parameters including two Hammett constants(σ_A and σ_B),two molar refractivities(M_A and M_B)and the sum of hydrophobic parameters(Ⅱ)of substituents A and B respectively.SVC model was obtained for the qualitative prediction of activities of the ethofenprox analogous,with the accuracy of prediction 90. 3%of leave-one-out cross-validation(LOOCV).The results indicated that the performance of SVC model was better than those of principalcomponent analysis(PCA),artificial neural network(ANN)and K-Nearest Neighbor(KNN)models for this real world data.There-fore,SVC method could be a promising tool in the field of the structure-activity relationship(SAR)research.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2004年第6期795-799,共5页
Computers and Applied Chemistry
基金
国家自然科学基金(20373040)