中文关键词:有机磷化合物 定量结构活性关系 分子描述符 支持向量机 英文关键词:Organophosphorouscompounds, Quantitativestructure-activityrelationship, Moleculardescriptors 基金项目:福建省自然科学基金计划项目（2019J01474）和泉州市科技计划项目（2018N121S）

作者	单位	E-mail
万金玉	泉州医学高等专科学校药学系 泉州	1406852345@QQ.COM
刘怡飞	福建医科大学附属第二医院 泉州

摘要点击次数:1048 全文下载次数:0 中文摘要: 随着有机磷化合物(OPs)的广泛应用,有机磷化合物在越来越多的环境介质中检测出来。大多数有机磷化合物具有毒性,人们往往缺乏快速且有效的预测手段来对毒性进行评估。本文将结合dragon软件计算的分子描述符,采用不同的QSAR模型对36个有机磷化合物的毒性进行预报。文中采用后退法作为描述符筛选结果,以均方根误差(RMSE)作为评价标准,我们共找到14个对线性核函数支持向量机(SVM)模型贡献较大的描述符,在最终得到的SVM模型交叉验证结果中,计算值与实际值的相关系数为0.913,均方根误差为0.388。外部测试验证结果中,平均相对误差为8.46%。此外,我们也采用多元线性回归(MLR)及人工神经网络(ANN)模型对有机磷化合物的毒性进行预报,交叉验证结果显示,两个模型的计算值与实际值的相关系数分别为：0.878与0.686,结果没有支持向量机模型预测能力好。因此采用线性核函数的支持向量机模型对有机磷化合物的毒性预报是一个行之有效的方法。 英文摘要: With the wide application of organophosphorus compounds (OPs), OPs have been detected in an increasing number of environmental media, most organophosphorus compounds are poisonous, people often lack of fast and effective prediction methods to evaluate the toxicity. In this paper, the toxicity of 36 organophosphorus compounds was predicted by using different QSAR models combined with molecular descriptors by computer with dragon software. In this paper we use the backward method as the screening result, root mean square error (RMSE) as evaluation standard, we have found 14 molecular descriptors have great influence on the linear kernel function of support vector machine (SVM) model, in the cross-validation results of support vector machine models, the correlation coefficient between the calculated value and the actual value was 0.913, the root mean square error is 0.388. In the results of external test verification, the average relative error was 8.46%. In addition, we also used multiple linear regression (MLR) and artificial neural network (ANN) models to predict the toxicity of organophosphorus compounds. The cross- 查看全文查看/发表评论下载PDF阅读器 相关附件：版权转让声明书 -->