摘要
本文应用网络科学的理论与方法,将组成分子的原子看作是网络节点,而一个分子的平面结构图就是由这些节点通过边连接而成,从而构造出分子结构网络,进一步计算出分子结构网络的平均度、平均路程长度等18个网络静态特征变量,作为构效分析的候选自变量。为了更好的进行构效分析,利用BP(Back Propagate,反向传播)人工神经网络从18个候选自变量中筛选出对活性贡献大的接近度中心性最大值等4个自变量,分别用支持向量机回归(SVR)和BP神经网络对分子结构网络的4个自变量与大鼠经口毒性LD_(50)进行定量构效关系(QSAR)研究,实验结果表明:基于分子结构网络静态参数的支持向量机回归模型具有良好的预测能力。说明应用分子结构网络静态参数建模,对具有同一(或多种)属性/活性的一类物质进行构效关系分析研究的这种新方法具有一定的应用前景。
The paper utilize the network of scientific theory and method, the atom that composed the molecule as a network node, a planar structure of molecular is passed by the edges connecting these nodes together to construct the molecular structure of the network; the 18 network static characteristic variables that calculate the average degree, the average path length, etc the molecular structure of the network as QSAR candidate arguments. In order to better carry out QSAR analysis, using BP artificial neural network variable from 18 candidates were selected from the active contribution of the proximity of the center of a large maximum, etc. 4 independent variables, then with support vector regression （SVR） and BP neural network on the molecular structure of the network of 4 independent variables and the rat by oral LDs0 for QSAR studies respectively. Experimental results show that： support vector machine regression that based on the molecular structure of the network static parameters model has good predictive ability. This new method that applying for network static parameter of molecular structure establish model has certain application prospect for the same （or more） property/activity of a class of substances QSAR analysis and research.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2014年第4期471-476,共6页
Computers and Applied Chemistry
基金
浙江省科技厅重点项目(2009C14024)
浙江海洋学院
东海科技学院研究性与创新性实验项目(2012.05)
关键词
复杂网络
网络静态特征
食品添加剂
定量构效关系
支持向量机回归
complex networks
networks static characteristics
food preservative
QSAR
support vector machine regression
