论文部分内容阅读
由化学物质毒性效应登录(RTECS)1998年版光盘系统检索获得52种环醚和34种氯代环醚急性毒性数据.应用主成份分析法从11种分子描述符中选择出对急性毒性有明显影响的5种:氯原子数,取代苯环上氢的氯原子数,取代烷基氢的氯原子数,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,低中和高剧两类毒性,正确分类率达到91%;而对于低、中、高、剧4类毒性,正确分类率达到80%.结果表明,氯代环醚的特殊结构具有很强的致毒作用.
Acute toxicity data of 52 kinds of cyclic ethers and 34 kinds of chlorinated cyclic ethers were obtained from the 1998 edition of the RTECS (RTECS) Registration System. The principal component analysis method was used to select from 11 kinds of molecular descriptors, which had obvious effects on acute toxicity Of the five kinds: the number of chlorine atoms, the number of chlorine atoms to replace the hydrogen on the benzene ring, the number of chlorine atoms, the number of alkyl groups and the number of oxygen atoms of the substituted alkylhydrogen.Through the sample learning set training and optimization of neural network structure, modeling classification, forecast , For 86 samples, low and medium high drama two types of toxicity, the correct classification rate of 91%; for low, medium, high, class 4 toxicity, the correct classification rate of 80% .The results showed that the chlorinated cyclic ether Special structure has a strong toxic effect.