Data-driven QSPR analysis of anti-cancer drugs using python-based topological techniques

• Published in: Journal of the Indian Chemical Society Vol. 102; no. 10; p. 101993
• Main Authors: Kara, Yeliz, Sağlam Özkan, Yeşim, Bektaş, Ali Berkan
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.10.2025
• Subjects: Anti-cancer drugs; Chemical and physical properties; Python algorithm; QSPR models; Topological indices; QSPR models; Chemical and physical properties; Python algorithm; Topological indices; Anti-cancer drugs
• ISSN: 0019-4522; 2667-2847
• DOI: 10.1016/j.jics.2025.101993

This study proposes a machine learning-based Quantitative Structure–Property Relationship (QSPR) model for predicting the physicochemical properties of anti-cancer drugs by utilizing topological descriptors. The development of anti-cancer drugs poses a significant challenge due to the intricate relationship between drug efficacy and chemical structure. The present study utilizes machine learning regression models in combination with leave-one-out cross-validation (LOOCV) to predict a range of physicochemical properties, including boiling point, enthalpy, molar refractivity, complexity, molecular weight, heavy atom count, flash point, and polarizability. The models are developed using data from thirty anti-cancer drugs and assessed using performance metrics such as the correlation coefficient (R), the coefficient of determination (R2) and root mean square error (RMSE). The findings are encouraging, with a thorough comparison made between the observed values and the values predicted by the QSPR models. [Display omitted] •The study focuses on the use of QSPR models to analyze 30 anticancer drugs used in different cancer therapies.•To compute degree-based topological indices (TIs) of anticancer drugs, a Python-based code is developed.•The study develops QSPR models that use degree-based TIs to correlate molecular descriptors with drug properties such as boiling point, enthalpy, molar refractivity, complexity, molecular weight, heavy atom count, flash point, and polarizability.•The study provides a comparative analysis between the linear, quadratic and cubic regression models.