文献类型：期刊文献

中文题名：环氧化合物分子结构与致变活性的关系

英文题名：Quantitative Structure and Activity Relationship for Epoxides

作者：白乃彬[1];顾玛力[2];谢飞[2];王兴平[2]

第一作者：白乃彬

机构：[1]中国科学院生态环境研究中心;[2]北京联合大学化学工程学院

第一机构：中国科学院生态环境研究中心,北京100085

年份：2001

卷号：22

期号：6

起止页码：112-115

中文期刊名：环境科学

外文期刊名：Chinese Journal of Enviromental Science

收录：CSTPCD;;Scopus;北大核心:【北大核心2000】；CSCD:【CSCD2011_2012】；PubMed;

基金：国家自然科学基金项目 (2 98770 32 )

语种：中文

中文关键词：分子结构;神经网络;环氧化物;致变活性;构效关系

外文关键词：Back\|Propagation Method(BP); neural networks;epoxides;mutagenic activity;QSAR

摘要：由化学物质毒性效应登录 (RTECS) 1 998年光盘系统检索获得 86种环氧化合物对鼠沙门氏菌致变活性数据 .应用主成分分析法从 1 1种分子描述符中选择出 5种对致变活性有明显影响的描述符 :碳原子数 ,苯环数 ,氢原子数 ,烷基数和氧原子数 .通过样本学习集训练并优化神经网络结构 ,建模分类预报 :对于 86个样本 ,低和高 2类活性 ,正确分类率达到 92 % .结果表明 :3～
Mutagenic activity data of 86 epoxides were collected from a CD system of Registry of Toxic Effects of Chemical Substances (RTECS 1998 Version). By using principal component analysis,the 5 kinds from 11 kinds of molecular structure descriptors, such as the number of carbon atom, the number of benzene cycle, the number of hydrogen atom, the number of substituted alkyl group and the number of oxygen atom which can have influence on their mutagenic activity very much were chosen. After learning of above samples, The Back\|Propagation Method(BP) network structure was optimized, the learning sets of sample were classified , then unknown samples were predicted one by one ; for 86 samples, the correctly classified rate can reach to 92 %between the low class and the high class. Results shown that the special molecular structure, the epoxides of polycyclic aromatic hydrocarbons contained 3～4 cycles, can be responsible for their very high mutagenic activities.