AI-Driven discovery of brain-penetrant mTOR-independent autophagy enhancers for Alzheimer's disease.

Current Alzheimer's disease therapies offer limited efficacy and are often accompanied by significant side effects, underscoring the urgent need for new treatment strategies. Enhancing autophagy represents a promising therapeutic approach, yet most known autophagy inducers act through the mTOR-dependent pathway, which broadly affects cellular metabolism and proliferation, and their clinical potential is further limited by poor blood-brain barrier (BBB) penetration. To address these twin challenges, an artificial intelligence (AI)-driven platform named DeepDrugDiscovery was developed, shifting the focus from traditional structure-based screening toward a mechanism-centric strategy for identifying mTOR-independent autophagy enhancers with brain penetrability. The platform screened over one million molecules and identified two lead compounds, Ombuin and 2-Hydroxycinnamic acid, which were experimentally shown to clear pathogenic tau and amyloid-β aggregates and restore memory function in both worm and mouse models of Alzheimer's disease. Notably, Ombuin exhibited robust brain exposure, confirming accurate BBB prediction. Released as an open-source resource, DeepDrugDiscovery demonstrates a scalable, AI-powered pipeline for discovering mechanism-based therapeutics.