QSAR Modeling of Pyridone Derivatives as α-amylase Inhibitors Using Chemical Descriptors and Machine Learning

Abstract

Diabetes is a prevalent chronic metabolic disorder that affects the lives and health of millions of individuals annually. α-amylase, a key digestive enzyme, plays a critical role in carbohydrate digestion. Inhibition of α-amylase activity can effectively slow the digestion of carbohydrates, thereby aiding in the maintenance of stable blood glucose levels. Consequently, the identification and development of potent α-amylase inhibitors have become a significant research focus in diabetes management. This study employed a modeling approach based on chemical descriptors and machine learning techniques to systematically explore the relationship between the chemical structures of 32 pyridone derivatives and their α-amylase inhibitory activity. A robust and predictive quantitative structure-activity relationship (QSAR) model was developed through optimization with the Sparrow algorithm, Monte carlo domain applicability evaluation, and Y-randomization testing. Utilizing this model in conjunction with data from the ZINC15 database, 23 potential compounds exhibiting favorable activity were designed. Further evaluation through SwissADME performance predictions identified three compounds with high inhibitory potential. Molecular docking studies provided insights into the potential binding modes and mechanisms of action of these compounds. The results of this study offer valuable theoretical support for the development of pyridone derivatives as potential therapeutic agents for diabetes and provide novel insights for the discovery of α-amylase inhibitors.