PCA-SVM在芳香族化合物生物降解性QSBR研究中的应用

定量结构-生物降解性能关系(QSBR)通过分析有机物结构与其生物降解性之间的定量关系,实现生物降解性的定量预测。针对影响生物降解性的基团结构多、传统方法难以消除基团数据之间的冗余,导致预测精度较低的问题,提出了一种基于主成分分析(PCA)-支持向量机(SVM)相结合的预测方法。首先利用主成分分析消除对该类化合物生物降解性影响较大的基团结构之间的冗余,降低数据维数,获取样本集主要信息;然后利用网格-交叉验证法优化后的支持向量机,建立预测模型。并与全要素的SVM模型及BP网络模型进行了比较,结果表明,该模型预测精度较高,具有通用性。

Application of PCA-SVM in the QSBR study on biodegradability of aromatic compounds

Quantitative structure-biodegradability relationship(QSBR)was chiefly studied to build quantitative relationship between the compounds structures and its biodegradability,and with this to predict its biodegradability.There were many molecule structures which affected the compounds biodegradability,and traditional method couldn’t eliminate data redundancy between these structures,so prediction accuracy was very low.In order to improve the biodegradability prediction accuracy of aromatic compounds,a new method was proposed based on principal component analysis(PCA)and support vector machine(SVM).Firstly,principal component analysis,which could eliminate the nonlinarity of these interrelated structures,was used to reduce the data dimension and acquire the characteristic information of the data set;then used support vector machine to establish biodegradability prediction model and the optimal parameters of the model were obtained via gridsearch-cross validation method.Compared this model with another SVM model which used the full structures that affected the biodegradability and BP neural network model,the results indicated that this model had better prediction accuracy and it was deserved to be popularized.