Data driven QSPR modeling of psychiatric drugs using degree based graph invariants

The combination of graph-theoretical descriptors and artificial intelligence techniques developed in recent years has greatly improved the predictive modeling of molecular properties. Herein, we demonstrate the applicability of various degree-based topological indices in QSPR analysis for different drug compounds including psychiatric drugs. In this paper, we compute a comprehensive set of topological descriptors that characterize molecular structures and study their relationships with various physicochemical properties using both classical regression models as well as artificial neural networks (ANNs). Although linear and polynomial regression models are able to give some indication about the dependence of structural descriptors on molecular properties, they can miss complex nonlinear relationships. To overcome this limitation, ANN models have been developed that are able to increase predictive accuracy and robustness. The results show that models built with neural networks outperform traditional approaches, especially for properties that behave non-linearly. Moreover, it addresses the importance of certain topological indices associated with molecular properties, thus providing potential implications for drug discovery and optimization. Thus, this work reinforces the perspective that a combination of topological indices through AI based techniques can be an efficient framework for accurate prediction of drug properties.