论文部分内容阅读
利用电矩矢量(Vmed)的分子结构数字技术,通过不同原子类型分类方法探索芳香胺类的基因毒性与组成原子类型之间的构效关系。随机多组的样本划分方案研究揭示,恰当的回归相关系数(R~2)能够在一定程度上避免模型的“过拟合”状态,原子划分类型越详细,就越容易陷入“过拟合”状态中。全分子集的4种原子类型分类方案综合研究揭示,化合物的基因毒性与分子结构的芳香性碳原子多寡、芳香氮原子的替换和稠环化程度等因素相关联。分析结果表明:通过不同的原子类型划分方案的综合研究,Vmed方法能够比较直观反映出分子结构的原子类型变化与性质变化关系,有效的指导化合物设计。
The structure-activity relationship between the genotoxicity of aromatic amines and the compositional atomic types was explored by using different molecular classification methods based on molecular structure of Vmed. The study of the randomized multi-group sample partitioning scheme reveals that the appropriate regression correlation coefficient (R ~ 2) can avoid the “overfitting” state of the model to a certain degree. The more detailed the atomic classification type, the easier it is to get stuck Fitting "state. A comprehensive study of the four atomic species classification schemes of the full-molecular set revealed that the genotoxicity of the compounds is related to the number of aromatic carbon atoms in the molecular structure, the substitution of aromatic nitrogen atoms and the degree of thick cyclization. The results show that the Vmed method can intuitively reflect the relationship between the change of atomic type and the change of the molecular structure and effectively guide the design of the compound through the comprehensive study of different atomic type partition schemes.