QSPR modeling of polychlorinated biphenyls using degree-based molecular descriptors: a comparative study with linear, polynomial, and ridge regression

مقال فى مجلة

2026

تم النشر فى: https://link.springer.com/article/10.1140/epje/s10189-025-00551-x

of 174

QSPR modeling of polychlorinated biphenyls using degree-based molecular descriptors: a comparative study with linear, polynomial, and ridge regression

By

Wei, H (Wei, Huang) [1] ; Noureen, S (Noureen, Sadia) [2] ; Maryam, A (Maryam, Amna) [2] ; Tchier, F (Tchier, Fairouz) [3] ; Aslam, A (Aslam, Adnan) [4]

 

 

 (provided by Clarivate) 

Source

EUROPEAN PHYSICAL JOURNAL E

Volume

49

Issue

1-2

DOI

10.1140/epje/s10189-025-00551-x

 

Article Number

8

Published

JAN 29 2026

Indexed

2026-02-03

Document Type

Article

Jump to

Abstract

Chemical graph theory provides a mathematical framework for representing molecular structures as graphs, where atoms correspond to vertices and chemical bonds to edges. This approach enables the use of molecular descriptors to extract reliable structural information and model physicochemical properties. In this study, we investigate the use of recently introduced degree-based molecular descriptors including Euler Sombor, elliptic Sombor, reverse Sombor, reverse elliptic Sombor, reverse Euler Sombor, Lanzhou, and ad-hoc Lanzhou indices to model key properties of polychlorinated biphenyls (PCBs). Experimentally reported properties such as melting point, relative retention time, octanol-water partition coefficient, enthalpy of formation, and Henry's law constant were analyzed. Quantitative structure-property relationship models were developed using linear, polynomial, and ridge regression techniques. The predictive performance of these models was evaluated through comparison of actual and predicted values, cross-validation, and bootstrapping. Results indicate that the selected descriptors, particularly the elliptic Sombor and reverse Euler Sombor indices, exhibit strong correlations with PCB properties, demonstrating their utility in predicting physicochemical behavior. These models hold potential for applications in chemical ecology, environmental risk assessment, and computational molecular design.