Discovering cholinesterase inhibitors from Chinese herbal medicine with deep learning models

Abstract

Traditional Chinese medicine (TCM) holds distinctive advantages in the management of Alzheimer’s disease. Nonetheless, a considerable gap remains in our understanding of its pharmacologically active constituents. In this study, we harnessed the potential of deep learning models to swiftly and precisely predict drug-target interactions. We conducted a systematic screening of cholinesterase (ChE) inhibitors from an extensive array of TCM ingredients, followed by rigorous validation through in vitro experiments. We constructed both a drug-target interactions (DTI) model and a blood-brain barrier permeability (BBBP) model, with both models achieving an AUPRC score exceeding 0.9. Subsequently, we conducted a screening process that identified six compounds for in vitro ChE inhibitory assay. Notably, all six compounds exhibited a robust inhibitory effect on acetylcholinesterase (AChE), while four of the six compounds demonstrated potent inhibitory activity against butyrylcholinesterase (BChE). Our findings underscore the promise of leveraging deep learning to discover inhibitors from TCM.