使用相关向量机及一致性建模预测大鼠经口急性毒性
详细信息 查看官网全文
- 英文论文题名：Accurate prediction of rat oral acute toxicity using relevance vector machine and consensus modeling
- 论文作者：雷太龙 ; 李有勇 ; 宋云龙 ; 李丹 ; 孙慧涌 ; 侯廷军
- 英文论文作者：Tailong Lei ; Youyong Li ; Yunlong Song ; Dan Li ; Huiyong Sun ; Tingjun Hou ; College of Pharmaceutical Sciences ; Zhejiang University ; FUNSOM ; Soochow University ; School of Pharmacy ; Second Military Medical University ; State Key Lab of CAD&CG ; Zhejiang University
- 年：2016
- 作者机构：浙江大学药学院;苏州大学功能纳米与软物质研究院;第二军医大学药学院药物化学系;浙江大学计算机辅助设计和图形学国家重点实验室;
- 论文关键词：毒性预测 ; 大鼠经口急性毒性 ; 相关向量机 ; 一致性模型
- 会议召开时间：2016-07-01
- 会议录名称：中国化学会第30届学术年会摘要集-第二十五分会：化学信息学与化学计量学
- 语种：中文
- 分类号：TQ460.72
- 学会代码：ZGHY
- 会议名称：中国化学会第30届学术年会-第二十五分会 ; 化学信息学与化学计量学
- 会议地点：中国辽宁大连
- 主办单位：中国化学会
- 学会名称：中国化学会
- 页数：1
- 文件大小：119k
- 原文格式：D
- 会议级别：全国
摘要
急性毒性的测定是药物开发中最重要的步骤,可以用半数致死量(LD_(50))来表示。由于急性毒性的体内试验非常昂贵和耗时,因此亟需开发其计算机预测模型。本研究中,我们基于7314个多样性化合物及其大鼠经口LD_(50)值,建立了一系列相关向量机(RVM)回归模型以预测化合物的经口急性毒性。与此同时,我们还将其与最邻近回归、随机森林(RF)、支持向量机、局部近似高斯过程、多层感知器集群和extreme梯度推进算法等六种方法构建的模型进行了比较。在各模型中,卡方统计量被用于分子描述符和结构指纹(PubChemFP或SubFP)的选择。单个模型的测试集预测精度q_(ext)~2为0.572~0.659。从测试集的总体预测精度来看,拉普拉斯核RVM和RF算法预测能力更好。我们还将各单独模型平均以建立四个一致性模型,取得了较好的预测精度(q_(ext)~2=0.669-0.689)。最后,我们还识别和分析了一系列急性毒性的描述符和子结构警报。我们认为最优一致性模型的预测精度较高,可以作为一个可靠的大鼠经口急性毒性预测工具。
In this study,based on 7314 diverse chemicals with rat oral LD_(50) values,relevance vector machine(RVM)technique was employed to build the regression models for the prediction of oral acute toxicity in rat,which were compared with those built using other six approaches,including k-nearest-neighbor,random forest(RF),support vector machine,local approximate Gaussian process,multilayer perceptron ensemble,and eXtreme gradient boosting.Dimension reduction was done by the Chi-squared statistics.The prediction capabilities(q_(ext)~2) of individual models ranged 0.572-0.659 for the test set.Considering the overall prediction accuracy for the test set,RVM with Laplacian kernel and RF were recommended to build models of rat oral acute toxicity.Then four consensus models were developed and predicted better for the test set(q_(ext)~2 = 0.669-0.689).Finally,some property or substructure alerts to oral acute toxicity were identified and analyzed.We believe that the best consensus model can be used as a reliable tool to filter out compounds with high rat oral acute toxicity.
In this study,based on 7314 diverse chemicals with rat oral LD_(50) values,relevance vector machine(RVM)technique was employed to build the regression models for the prediction of oral acute toxicity in rat,which were compared with those built using other six approaches,including k-nearest-neighbor,random forest(RF),support vector machine,local approximate Gaussian process,multilayer perceptron ensemble,and eXtreme gradient boosting.Dimension reduction was done by the Chi-squared statistics.The prediction capabilities(q_(ext)~2) of individual models ranged 0.572-0.659 for the test set.Considering the overall prediction accuracy for the test set,RVM with Laplacian kernel and RF were recommended to build models of rat oral acute toxicity.Then four consensus models were developed and predicted better for the test set(q_(ext)~2 = 0.669-0.689).Finally,some property or substructure alerts to oral acute toxicity were identified and analyzed.We believe that the best consensus model can be used as a reliable tool to filter out compounds with high rat oral acute toxicity.
引文
[1]Lei,T.;Li,Y.;Song,Y;Li,D.;Sun,H.;Hou,T.,ADMET evaluation in drug discovery:15.Accurate prediction of rat oral acute toxicity using relevance vector machine and consensus modeling.J.Cheminformatics 2016,8:1-19.