Abstract

Several chemical and medical experimentation reveals a dependence of physicochemical and biological properties of a compound on its molecular structure. Molecular/topological descriptors/indices retrieve this dependence by employing mathematical/statistical tools to generate quantitative structure property/activity relationship (QSAR/QSPR) models. QSAR/QSPR models are regression models which theoretically relate a physicochemical/biological property to a molecular descriptor. By converting a compound to a chemical graph, graph-theoretic topological indices use algorithmic graph theory to generate QSAR/QSPR models. Degree-related molecular indices have been well-known for correlating better with certain physicochemical characteristics of compounds. In this paper, we conduct a comparative analysis to put forward the best degree-based index for correlating with physicochemical characteristics of polycyclic aromatic hydrocarbons. Two different computational techniques have been put forward and then used to conduct our comparative analysis. Our comparative testing generates several favorable and unexpected conclusions as the well-known indices such as the augmented Zagreb index, the atom-bond connectivity index, the geometric arithmetic index deliver relatively poor performance. Moreover, exceptional performance of the generalized variants of Randić and sum-connectivity indices is somehow unexpected as they significantly outperform the well-studied and well-researched degree-based indices. This warrants to study further the general Randić and sum-connectivity indices for QSAR/QSPR models. Applications to certain families of carbon polyhex aromatic nanotubes are reported.

摘要

一些化学和医学实验揭示了化合物的物理化学和生物学特性对其分子结构的依赖性。分子/拓扑描述符/索引通过使用数学/统计工具生成定量结构特性/活性关系 (QSAR/QSPR) 模型来检索这种依赖性。QSAR/QSPR 模型是理论上将物理化学/生物学特性与分子描述符相关联的回归模型。通过将化合物转换为化学图，图论拓扑指数使用算法图论生成 QSAR/QSPR 模型。众所周知，与度数相关的分子指标与化合物的某些物理化学特征具有更好的相关性。在本文中，我们进行了对比分析，提出了与多环芳烃理化特性相关的最佳度数指标。已经提出了两种不同的计算技术，然后用于进行我们的比较分析。我们的比较测试产生了几个有利和出乎意料的结论，因为众所周知的指数，如增强萨格勒布指数、原子键连接指数、几何算术指数表现相对较差。此外，Randić 和 sum-connectivity 指数的广义变体的出色表现在某种程度上是出乎意料的，因为它们明显优于经过充分研究和研究的基于学位的指数。这有必要进一步研究 QSAR/QSPR 模型的一般 Randić 和总连接指数。