Advanced QSPR modeling of profens using machine learning and molecular descriptors for NSAID analysis

In this paper, we present a predictive model based on artificial neural network (ANN) to evaluate principal physicochemical properties of a set of anti-inflammatory drugs based on chosen topological indices. The molecular descriptors were calculated from molecular structures and employed as the inputs to the ANN model. Normalization of the feature set was carried out before training to maintain convergence and stability of the model. The ANN exhibited excellent predictive ability based on a [Formula: see text] value of 0.94 and a mean squared error (MSE) of 0.0087 on the test set. The chemical structure data used were mainly retrieved from ChemSpider. The method showcases the promise of machine learning models to facilitate better virtual screening and assist in rational drug design by making accurate predictions of properties.