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

Degree-based topological indices and QSPR analysis of novel anti-glaucoma drugs using machine learning

Uma Ramachandran, N. Kavitha, and H. Naresh Kumar

Deparment of Mathematics, Government College of Engineering, Thanjavur, India

E-mail: ruma@gcetj.edu.in

Received: 3 December 2025  Accepted: 22 February 2026

Abstract:

Quantitative prediction of physicochemical properties is crucial for rational drug design. In this study, chemical graph theory was used to model the molecular structures of glaucoma drugs, representing atoms as vertices and bonds as edges. Topological indices derived from these graphs were employed as molecular descriptors to develop quantitative structure–property relationship (QSPR) models using linear regression (LR), support vector regression (SVR), and CatBoost regression. Model performance was evaluated using coefficient of determination (\(R^2\)), root mean square error (RMSE), mean absolute error (MAE), mean percentage error (MPE), and mean absolute percentage error (MAPE). To ensure a reliable and unbiased assessment, leave-one-out cross-validation (LOOCV) was applied to all models. The results indicate that SVR and CatBoost outperform linear regression in predictive accuracy. CatBoost achieved superior in-sample performance, while SVR demonstrated strong generalisation capability. These findings confirm the effectiveness of topological descriptors combined with machine learning for predicting drug properties and supporting the development of novel anti-glaucoma agents.

Keywords: Topological indices; Anti-glaucoma drugs; CatBoost; QSPR analysis; Linear regression; Support vector regression

Full paper is available at www.springerlink.com.

DOI: 10.1007/s11696-026-04763-z

Chemical Papers 80 (6) 6237–6269 (2026)