부산대학교 도서관

This study aimed to decipher the encoded information within the molecular structure of a dataset of GSK-3β inhibitors through a comprehensive quantity structural-activity relationship (QSAR) investigation employing traditional physicochemical descriptors. Various statistical techniques were applied, encompassing linear methods such as Multiple Linear Regression (MLR) and Partial Least Squares (PLS), along with non-linear approaches such as Artificial Neural Networks (ANN). Rigorous validation using diverse statistical tools confirmed the models’ precision and predictability. The model exhibited exceptional statistical relevance, as evidenced by standard parameters: S-value (0.37), F-value (37.17), r (0.93), r2 (0.855), and r2CV (0.78). They assessed predictive power and robustness by involving specific statistical parameters. Key descriptors analyzed, including Verloop L (subs 2), Lipole Z component (whole molecule), and VAMP Dipole Z component, offered crucial insights into their respective contributions. This analysis implies that targeted modifications in the substitution pattern potentially enhance GSK-3β inhibitory activity significantly. The established model clarified how bioactivity depends on structure and offered valuable recommendations for creating new compounds with better inhibitory activity profiles against the GSK-3β enzyme.