AI Drug Discovery Platform Identifies Novel Blood-Brain Barrier-Permeable Alzheimer's Compounds — Nature Biomedical Engineering

An open-source AI drug screening platform called DeepDrugDiscovery, developed by Mindrank AI, was published in Nature Biomedical Engineering on April 25, 2026. The platform uses mechanism-aware deep learning with integrated ADMET (absorption, distribution, metabolism, excretion, toxicity) predictions and blood-brain barrier permeability modeling to identify novel Alzheimer's disease drug candidates. Unlike conventional virtual screening approaches, DeepDrugDiscovery selects for compounds based on their mechanism of action — in this case, targeting autophagy enhancement to clear amyloid-beta and tau protein aggregates — while simultaneously filtering for favorable pharmacokinetic properties. Lead candidate compounds from the screen successfully crossed the blood-brain barrier in computational and cell-line validation, cleared amyloid and tau aggregates in C. elegans (nematode worm) and mouse Alzheimer's disease models, and restored memory function — without relying on mTOR pathway inhibition, which causes systemic immunosuppression side effects in existing autophagy drugs. The platform is publicly available at deepdrugdiscovery.mindrank.ai, making the same AI tool used in this study accessible to research institutions globally, including those in low- and middle-income countries where pharmaceutical AI infrastructure is limited. The Alzheimer's Association estimates 50 million people live with Alzheimer's globally, with 10 million new cases per year. The high failure rate of Alzheimer's drugs in clinical trials (99% over the past two decades) reflects the difficulty of BBB penetration and off-target effects — precisely the parameters DeepDrugDiscovery is designed to address computationally before synthesis and testing.