ISSN print edition: 0366-6352 ISSN electronic edition: 1336-9075 Registr. No.: MK SR 9/7 Published monthly

Topological and statistical regression study of chemical structures using graph-theoretic descriptors: applications to cancer therapeutics

Wakeel Ahmed, Ghulam Fatima, Shahid Zaman, Asad Ullah, Emad E. Mahmoud, and Tamseela Ashraf

Department of Mathematics, COMSATS University Islamabad, Lahore, Pakistan

E-mail: wakeelahmed784@gmail.com

Received: 13 August 2025  Accepted: 22 September 2025

Abstract:

This study integrates computational chemistry and machine learning to explore the relationship between topological indices and physicochemical properties of compounds. Artificial neural networks (ANNs) and random forest (RF) models were developed, with ANN consistently achieving superior predictive performance. ANN attained significantly lower errors than RF for enthalpy of vaporization, with similar trends across properties such as density, boiling point, and surface tension. The analysis revealed that topological indices played a key role in predicting physicochemical properties such as molar refractivity and polar surface area. These findings underscore the robustness of ANN in capturing complex nonlinear structural property relationships and highlight its potential as a scalable computational framework for molecular design and drug discovery.

Keywords: Artificial neural networks; Cheminformatics; Python algorithm; Topological descriptors; Anti-skin cancer drugs

Full paper is available at www.springerlink.com.

DOI: 10.1007/s11696-025-04404-x

Chemical Papers 80 (1) 461–483 (2026)