摘要
通过检索美国国立职业与卫生研究所化学物质毒性效应登录 ( RTECS) 1 998年版光盘系统 ,搜集了 88种醛类化合物对大鼠急性毒性口服 L D50 数据 ,并利用遗传神经网络建立了醛类化合物分子结构与对大鼠急性毒性关系的 BP神经网络模式 ,模式的交互检验相关系数达 0 .83～ 0 .87,具有较强的预报能力 .遗传神经网络是由遗传算法和神经网络耦合而成 ,文章详细讨论了遗传神经网络的构造以及利用遗传神经网络建立 QSAR模型的方法 .
Acute toxicity data of 88 aldehydes to rat LD 50 were collected from optical disk system of Registry of Toxic Effects of Chemical Substances (RTECS) developed by National Institute for Occupational Safety and Health (NIOSH).Some models of artificial neural networks based on BP algorithm for relationships between acute toxicity of aldehydes to rat LD 50 and their molecular structure descriptors got from structural formula were then built by genetic neural network approach.The cross validation correlation coefficient of these models were 0 83～0 87.Genetic neural network was obtained by integrating genetic algorithm and neural networks.How a genetic neural network was constructed and how a QSAR model was built by using it was described in detail in this paper.
出处
《环境科学》
EI
CAS
CSCD
北大核心
2000年第5期89-93,共5页
Environmental Science
基金
国家自然科学基金!资助项目 ( 2 98770 32 )
