An analysis of the autocorrelation descriptor for molecules

An analysis of the autocorrelation descriptor for molecules
复制标题

DOI:
10.1023/a:1023247831238
复制
发表时间:
2003-02-01
影响因子:
1.7
通讯作者:
Hollas, B
Hollas, B
中科院分区:
化学3区
文献类型:
--
作者:
Hollas, B

文献摘要

被引文献

相似文献

我们讨论的自相关描述符,一个拓扑描述符编码的分子结构和物理化学性质的分子的性质。我们介绍了两个随机图模型的分子,并表明,这种描述符可能会表现出不必要的相关性,使生成的数据无法用于结构-活性关系的研究。这个缺点可以很容易地消除居中属性,方便随后的统计分析。
We discuss properties of the autocorrelation descriptor, a topological descriptor encoding both molecular structure and physico-chemical properties of a molecule. We introduce two random graph models for molecules and show that this descriptor may exhibit unwanted correlation properties, making the generated data unusable for structure-activity relationship studies. This shortcoming can easily be eliminated by centering properties, facilitating subsequent statistical analysis.