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

A graph-theoretical and machine learning framework for structure–property analysis of phytochemicals via matching energy

Sadia Noureen, Nazma Ashraf, and Majid Hussain

Department of Mathematics, University of Gujrat, Gujrat, Pakistan

E-mail: sadia.tauseef@uog.edu.pk

Received: 2 April 2026  Accepted: 12 May 2026

Abstract:

This paper shows a graph-based exploration of the structure-property correlation among some specific phytochemicals in terms of matching energy as a topological measure. The individual molecules are modeled as graphs, and the corresponding polynomials are calculated to obtain corresponding energy as a topological characteristic. A sample of varied plant-based bioactive substances is examined to examine the connection between matching energy and various physicochemical characteristics, such as molecular weight (MW), molar refractivity (MR), Bertz complexity (BertzCT), NHA, Heavy Atoms, Chi 0v, Chi 1v, and Labute ASA. The high correlations between matching energy and all the properties under study showed that linear regression analysis is a good predictive parameter. To achieve greater predictive accuracy, the use of the XGBoost regression model is employed, which produced greater results in terms of high values of the \(r^{2}\) and low error in prediction. Train-test validation, k-fold cross-validation, and bootstrap analysis are used to test the reliability of the models. The findings, in general, underscore the usefulness of the integration of both graph theory and machine learning to precisely predict the physicochemical properties of phytochemical compounds.

Keywords: Phytochemicals; Matching polynomial; Matching energy; Topological indices; Structure–property relationship; Linear regression; XGBoost model; 05C92; 92E10; 62J05; 68T07

Full paper is available at www.springerlink.com.

DOI: 10.1007/s11696-026-05057-0

Chemical Papers 80 (9) 10967–10985 (2026)