摘要
应用自适应分子结构描述符生成方法和神经网络数值模式，研究６２种烹调食品过程中产生的杂环芳胺结构与致变活性间的关系。该模式预报结果与实验测定结果符合良好；致变剂、非致变剂和勉强有致变活性剂三类间正确分类率超过９０％。通过６次自适应分子结构描述符选择迭代，得到５种特征分子描述符为：芳环碳原子取代甲基数、芳环氮原子取代甲基数、母体共轭环数、端环增活结构数和端环抑活结构数。其中，增活结构为具有共轭烯烃碎片结构；抑活结构具有短共轭烯烃结构或不具有共轭烯烃结构。最后，对该构效关系的起因进行了定性解释。
In this paper,a structure-mutagenicity study of 62 heterocyclic amines formed during the cooking of food has been carried out using both feed-forward neural networks and a adaptive selection of procedure from molecular descriptors. Prediction of mutagenicities is in a good agreement with their experiment data. Correctly classified rates in the three classes (mutagen, nonmutagen and marginal) are all more than 90%. After sixth iteration of adaptive selection ofmolecular structure descriptors,5 from them have been chosen as very important molecular descriptors. They are number of methy1 group connected with a carbon atom in aromatic cycle,number of methy1 group connected with a nitrogen atom in aromatic cycle, number of conjugate cycle in the parent part of molecular, number of enhanced mutagenicity sub-structure and number of inhibited mutagenicity sub-structure. The enhanced sub-structures are some longer conjugate alkene fragments existed in termainal cycle, while the inhibited sub-structure has some short conjugate alkene fragments in the termainal cycle. Finally, causes of the enhancement or the inhibition have been also discussed.
出处
《中国环境科学》
EI
CAS
CSSCI
CSCD
北大核心
1995年第6期447-451,共5页
China Environmental Science
基金
国家自然科学基金
