A hybrid chemoinformatics approach integrating caputo fractional-derivative graph invariants and artificial neural network for modeling of anti-sickle cell compounds

Abstract

Abstract Cheminformatics focuses on the accurate prediction of physicochemical properties, which plays a crucial role in molecular modeling and materials science, as experimental determination is often time-consuming and costly. In this study, we propose a machine learning based framework for predicting physicochemical properties using fractional derivative based topological indices as feature variable. Two widely used machine learning algorithms, namely Artificial Neural Networks (ANN) and Random Forest (RF), were employed to model the nonlinear relationships between topological descriptors and the corresponding physicochemical properties. A comprehensive set of fractional derivative based topological indices was used as independent variables, while multiple physicochemical properties were considered as targeted variables. The predictive performance of the developed models was rigorously evaluated using standard statistical metrics, including the coefficient of determination ( $$R^2$$ ), mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE). The results demonstrate that ANN model consistently exhibited superior performance, achieving higher values of $$R^2$$ (generally exceeding 0.82) and lower error measures compared to the RF based models for the majority of the properties under investigation. This highlights the enhanced capability of ANN in capturing complex nonlinear patterns inherent in fractional derivative based topological indices. The proposed approach confirms that fractional derivative based topological indices, when integrated with advanced machine learning algorithms, serve as powerful predictors of physicochemical properties. The findings suggest that the ANN-based framework, in particular, can be effectively utilized as a reliable and efficient tool for property prediction in chemical and molecular informatics, thereby reducing experimental effort and accelerating material and compound design.