基于QSAR模型预测有机化合物的生物富集因子  

Prediction of the bioconcentration factor of organic compounds based on QSAR model

作　　者：万金玉 李雪娇 WAN Jinyu;LI Xuejiao(College of Pharmacy,Quanzhou Medical College,Quanzhou 362000,Fujian,China;Shanghai Institute of Applied Physics,Chinese Academy of Sciences,Shanghai 201800,China)

机构地区：[1]泉州医学高等专科学校,福建泉州362000 [2]中国科学院上海应用物理研究所,上海201800

出　　处：《化学研究》2023年第2期140-146,共7页Chemical Research

基　　金：福建省自然科学基金计划(2019J01474);泉州市科技计划(2018N121S)。

摘　　要：生物富集因子(BCF)是REACH法规要求的生物积累危害评价指标的关键参数,用实验的方法测定BCF值人力物力花费较大,而用计算的方法预测BCF值可以有效地替代昂贵的实验过程。该研究使用E-Dragon计算了数据集中每个分子的1666种描述符,并用筛选后的描述符与lgBCF建立了QSAR模型:采用随机森林与支持向量机建立的分类预报模型,随机森林分类模型的准确率为0.89、敏感度为0.89;用基于准确率递减和基尼系数方法的随机森林分类模型筛选出对lgBCF值有重要影响的30个描述符,这些描述符包括ALOGPS_lg P、MATS6v、TPSA.NO.、GATS7v等;ALOGPS_lg P和ALOGPS_lg S是用支持向量机分类模型筛选出的对lgBCF值有重要影响的描述符。The bioconcentration factor(BCF)is a key parameter about bioaccumulation hazard assessment metric required by the REACH regulation.It is expensive and time-consuming to measure,while silico methods to predict BCF values is an effective alternative to expensive experimental procedures.In this paper,1666 descriptors for each molecule in the dataset were calculated by E-Dragon,and a QSAR model was established between the filtered descriptors and lgBCF:the classification prediction models were established by random forest and support vector machine,and the accuracy and sensitivity of random forest model are 0.89 and 0.89 respectively.Then 30 descriptors that have important influence on lgBCF value were filtered by random forest classification model based on decreasing accuracy method and Gini coefficient method.Those descriptors include ALOGPS_lg P,MATS6v,TPSA.NO.,GATS7v and so on.ALOGPS_lg P and ALOGPS_lg S have a major impact on lgBCF filtered by support vector machine classification model value.

关 键 词：富集因子 定量构效关系 随机森林 支持向量机 

分 类 号：X171.5[环境科学与工程—环境科学]