DeepDrugDiscovery identifies blood–brain barrier permeable autophagy enhancers for Alzheimer’s disease | Nature Biomedical Engineering

Your privacy, your choice

We use essential cookies to make sure the site can function. We also use optional cookies for advertising, personalisation of content, usage analysis, and social media, as well as to allow video information to be shared for both marketing, analytics and editorial purposes.

By accepting optional cookies, you consent to the processing of your personal data - including transfers to third parties. Some third parties are outside of the European Economic Area, with varying standards of data protection.

See our privacy policy for more information on the use of your personal data.

Manage preferences for further information and to change your choices.

Accept all cookies Reject optional cookies

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Abstract

Dysfunctional autophagy, a key cellular cleaning process, is a key driver of brain ageing and neurodegenerative diseases such as Alzheimer’s disease (AD). However, developing effective treatments by enhancing autophagy has been challenging, as most known compounds act through the broad mTOR pathway, risking side effects, and few can effectively penetrate the brain. To address this, we developed DeepDrugDiscovery—a mechanism-aware, AI-powered screening platform incorporating ADMET and blood–brain barrier penetrability predictions. Here we show that this platform successfully identified novel, mTOR-independent autophagy enhancers, with two lead compounds demonstrating an ability to cross the blood–brain barrier, clear AD-related protein aggregates and restore memory function in worm and mouse AD models. By releasing DeepDrugDiscovery as an open-source, modular tool, we offer a user-friendly AI platform that enables customized therapeutic screening. Our work establishes a scalable, AI-driven pipeline that integrates cross-species validation to rapidly discover mechanism-based therapeutics for diseases with high unmet medical need.

Access through your institution

Buy or subscribe

This is a preview of subscription content, access via your institution

Access options

Access through your institution

Access Nature and 54 other Nature Portfolio journals

Get Nature+, our best-value online-access subscription

$32.99 /30 days

cancel any time

Learn more

Subscribe to this journal

Receive 12 digital issues and online access to articles

$119.00 per year

only $9.92 per issue

Learn more

Buy this article

• Purchase on SpringerLink
• Instant access to the full article PDF.

USD 39.95

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Workflow of the DeepDrugDiscovery for ligand-based virtual screening and predictive modelling.

The alternative text for this image may have been generated using AI.

Fig. 2: DeepDrugDiscovery screening results of BBB penetrating mTOR-independent autophagy inducers.

The alternative text for this image may have been generated using AI.

Fig. 3: Validation of candidate compounds for autophagy modulation.

The alternative text for this image may have been generated using AI.

Fig. 4: Characterization of candidate mTOR-independent autophagy enhancers.

The alternative text for this image may have been generated using AI.

Fig. 5: Clearance of AD-related pathological proteins by candidate compounds.

The alternative text for this image may have been generated using AI.

Fig. 6: Omb and 2-HCA regulate autophagy and exhibit neuroprotective effects in the C. elegans model of AD.

The alternative text for this image may have been generated using AI.

Fig. 7: Omb and 2-HCA improve cognitive dysfunction in the 3×Tg-AD mouse model.

The alternative text for this image may have been generated using AI.

Fig. 8: Omb and 2-HCA demonstrate the ability to clear abnormal protein aggregates in the 3×Tg-AD mouse model.

The alternative text for this image may have been generated using AI.

Similar content being viewed by others

DeepDrug as an expert guided and AI driven drug repurposing methodology for selecting the lead combination of drugs for Alzheimer’s disease

Article Open access 15 January 2025

Discovery of a small molecule secreted clusterin enhancer that improves memory in Alzheimer’s disease mice

Article Open access 02 May 2025

Identification of therapeutic targets for Alzheimer’s Disease Treatment using bioinformatics and machine learning

Article Open access 31 January 2025

Data availability

Public candidate compound dataset is available via figshare at https://figshare.com/articles/dataset/Compound_Library_for_DeepDrugDiscovery_platform/31123354. DeepDrugDiscovery web interface is available at https://deepdrugdiscovery.mindrank.ai/. ADMET prediction web interface is available at https://admet.mindrank.ai/. Source data are provided with this paper.

Code availability

Primary GitHub repository is available at https://github.com/XiangLuXiao/DeepDrugDiscovery. This repository contains the complete code for model training and inference, example datasets, tutorials, documentation on computational requirements and configurations for Docker containers or Conda environments.

References

1. GBDDF Collaborators. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health7, e105–e125 (2022).

ArticleGoogle Scholar

1. Zhang, J. et al. Recent advances in Alzheimer’s disease: mechanisms, clinical trials and new drug development strategies. Signal Transduct. Target. Ther.9, 211 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Aman, Y. et al. Autophagy in healthy aging and disease. Nat. Aging1, 634–650 (2021).

ArticlePubMedPubMed CentralGoogle Scholar

1. Nixon, R. A. & Rubinsztein, D. C. Mechanisms of autophagy-lysosome dysfunction in neurodegenerative diseases. Nat. Rev. Mol. Cell Biol.25, 926–946 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Zhang, X. W. et al. Targeting autophagy in Alzheimer’s disease: animal models and mechanisms. Zool. Res.44, 1132–1145 (2023).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Kerr, J. S. et al. Mitophagy and Alzheimer’s disease: cellular and molecular mechanisms. Trends Neurosci.40, 151–166 (2017).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Kobro-Flatmoen, A. et al. Re-emphasizing early Alzheimer’s disease pathology starting in select entorhinal neurons, with a special focus on mitophagy. Ageing Res. Rev.67, 101307 (2021).

ArticleCASPubMedGoogle Scholar

1. Deng, Z. et al. Pharmacological modulation of autophagy for Alzheimer’s disease therapy: opportunities and obstacles. Acta Pharm. Sin. B12, 1688–1706 (2022).

ArticleCASPubMedGoogle Scholar

1. Date, Y. et al. Novel autophagy inducers by accelerating lysosomal clustering against Parkinson’s disease. eLife13, e98649 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Sarkar, S. Regulation of autophagy by mTOR-dependent and mTOR-independent pathways: autophagy dysfunction in neurodegenerative diseases and therapeutic application of autophagy enhancers. Biochem. Soc. Trans.41, 1103–1130 (2013).

ArticleCASPubMedGoogle Scholar

1. Dong, Y. et al. Autophagy modulator scoring system: a user-friendly tool for quantitative analysis of methodological integrity of chemical autophagy modulator studies. Autophagy16, 195–202 (2020).

ArticleCASPubMedGoogle Scholar

1. Zhang, H. et al. Heterophyllin B enhances transcription factor EB-mediated autophagy and alleviates pyroptosis and oxidative stress after spinal cord injury. Int. J. Biol. Sci.20, 5415–5435 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Williams, A. et al. Novel targets for Huntington’s disease in an mTOR-independent autophagy pathway. Nat. Chem. Biol.4, 295–305 (2008).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Zhang, X. et al. MTOR-independent, autophagic enhancer trehalose prolongs motor neuron survival and ameliorates the autophagic flux defect in a mouse model of amyotrophic lateral sclerosis. Autophagy10, 588–602 (2014).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Mak, K. K. et al. Artificial intelligence in Drug Discovery and Evaluation: Safety and Pharmacokinetic Assays (eds Franz, J. H. & Michael, K.) 1461–1498 (Springer, 2024).

2. Rehman, A. U. et al. Role of artificial intelligence in revolutionizing drug discovery. Fundam. Res.5, 2667–3258 (2024).

Google Scholar

1. Niu, Y. Z. & Lin, P. Advances of computer-aided drug design (CADD) in the development of anti-Azheimer’s-disease drugs. Drug Discov. Today28, 103665 (2023).

ArticleCASPubMedGoogle Scholar

1. Zhou, G. et al. An artificial intelligence accelerated virtual screening platform for drug discovery. Nat. Commun.15, 7761 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Zhang, K. et al. Artificial intelligence in drug development. Nat. Med.31, 45–59 (2025).

ArticleCASPubMedGoogle Scholar

1. Xiang, H. et al. Drug discovery by targeting the protein-protein interactions involved in autophagy. Acta Pharm. Sin. B13, 4373–4390 (2023).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Yang, X. et al. Concepts of artificial intelligence for computer-assisted drug discovery. Chem. Rev.119, 10520–10594 (2019).

ArticleCASPubMedGoogle Scholar

1. Pham, T. H. et al. A deep learning framework for high-throughput mechanism-driven phenotype compound screening and its application to COVID-19 drug repurposing. Nat. Mach. Intell.3, 247–257 (2021).

ArticlePubMedPubMed CentralGoogle Scholar

1. Xie, C. et al. Amelioration of Alzheimer’s disease pathology by mitophagy inducers identified via machine learning and a cross-species workflow. Nat. Biomed. Eng.6, 76–93 (2022).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Mizushima, N., Yoshimori, T. & Levine, B. Methods in mammalian autophagy research. Cell140, 313–326 (2010).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Tanida, I. et al. Lysosomal turnover, but not a cellular level, of endogenous LC3 is a marker for autophagy. Autophagy1, 84–91 (2005).

ArticleCASPubMedGoogle Scholar

1. Marwaha, R. & Sharma, M. DQ-Red BSA trafficking assay in cultured cells to assess cargo delivery to lysosomes. Bio Protoc.7, e2571 (2017).

ArticlePubMedGoogle Scholar

1. Dong, Y. et al. Toosendanin, a novel potent vacuolar-type H+-translocating ATPase inhibitor, sensitizes cancer cells to chemotherapy by blocking protective autophagy. Int. J. Biol. Sci.18, 2684–2702 (2022).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Wang, Y. & Mandelkow, E. Tau in physiology and pathology. Nat. Rev. Neurosci.17, 5–21 (2016).

ArticlePubMedGoogle Scholar

1. Jonsson, T. et al. A mutation in APP protects against Alzheimer’s disease and age-related cognitive decline. Nature488, 96–99 (2012).

ArticleCASPubMedGoogle Scholar

1. DeLoach, K. E. Characterization of Mice Expressing APPSwe/Ind in DG, CA3 or CA1 of the Hippocampus. MSc thesis, UC San Diego (2012).

2. Cai, Z. & Yan, L. J. Rapamycin, autophagy, and Alzheimer’s disease. J. Biochem. Pharmacol. Res.1, 84–90 (2013).

PubMedPubMed CentralGoogle Scholar

1. Fang, E. F. et al. Mitophagy inhibits amyloid-beta and tau pathology and reverses cognitive deficits in models of Alzheimer’s disease. Nat. Neurosci.22, 401–412 (2019).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Pasquier, B. SAR405, a PIK3C3/Vps34 inhibitor that prevents autophagy and synergizes with MTOR inhibition in tumor cells. Autophagy11, 725–726 (2015).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Yamamoto, A. et al. Bafilomycin A1 prevents maturation of autophagic vacuoles by inhibiting fusion between autophagosomes and lysosomes in rat hepatoma cell line, H-4-II-E cells. Cell Struct. Funct.23, 33–42 (1998).

ArticleCASPubMedGoogle Scholar

1. Griffin, E. F., Caldwell, K. A. & Caldwell, G. A. Genetic and pharmacological discovery for Alzheimer’s disease using Caenorhabditis elegans. ACS Chem. Neurosci.8, 2596–2606 (2017).

ArticleCASPubMedGoogle Scholar

1. Cao, S. Q. et al. Chemotaxis assay for evaluation of memory-like behavior in wild-type and Alzheimer’s-disease-like C. elegans models. STAR Protoc.4, 102250 (2023).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Strand, B. H. et al. Survival and years of life lost in various aetiologies of dementia, mild cognitive impairment (MCI) and subjective cognitive decline (SCD) in Norway. PLoS ONE13, e0204436 (2018).

ArticlePubMedPubMed CentralGoogle Scholar

1. Larson, E. B. et al. Survival after initial diagnosis of Alzheimer disease. Ann. Intern. Med.140, 501–509 (2004).

ArticlePubMedGoogle Scholar

1. Sterniczuk, R. et al. Characterization of the 3xTg-AD mouse model of Alzheimer’s disease: part 1. Circadian changes. Brain Res.1348, 139–148 (2010).

ArticleCASPubMedGoogle Scholar

1. Sterniczuk, R. et al. Characterization of the 3xTg-AD mouse model of Alzheimer’s disease: part 2. Behavioral and cognitive changes. Brain Res.1348, 149–155 (2010).

ArticleCASPubMedGoogle Scholar

1. Michaud, F. et al. Altered firing output of VIP interneurons and early dysfunctions in CA1 hippocampal circuits in the 3xTg mouse model of Alzheimer’s disease. eLife13, RP95412 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Park, S. et al. Distinct disruptions in CA1 and CA3 place cell function in Alzheimer’s disease mice. iScience28, 111631 (2025).

ArticlePubMedPubMed CentralGoogle Scholar

1. Hussain, B., Fang, C. & Chang, J. Blood-brain barrier breakdown: an emerging biomarker of cognitive impairment in normal aging and dementia. Front. Neurosci.15, 688090 (2021).

ArticlePubMedPubMed CentralGoogle Scholar

1. Tsartsalis, S. et al. A single nuclear transcriptomic characterisation of mechanisms responsible for impaired angiogenesis and blood-brain barrier function in Alzheimer’s disease. Nat. Commun.15, 2243 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Tsuji, A. Small molecular drug transfer across the blood-brain barrier via carrier-mediated transport systems. NeuroRx2, 54–62 (2005).

ArticlePubMedPubMed CentralGoogle Scholar

1. Ohtsuki, S. & Terasaki, T. Contribution of carrier-mediated transport systems to the blood-brain barrier as a supporting and protecting interface for the brain; importance for CNS drug discovery and development. Pharm. Res.24, 1745–1758 (2007).

ArticleCASPubMedGoogle Scholar

1. Franken, H. et al. Thermal proteome profiling for unbiased identification of direct and indirect drug targets using multiplexed quantitative mass spectrometry. Nat. Protoc.10, 1567–1593 (2015).

ArticleCASPubMedGoogle Scholar

1. Xie, Z. et al. Gene set knowledge discovery with Enrichr. Curr. Protoc.1, e90 (2021).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Tang, D. et al. SRplot: a free online platform for data visualization and graphing. PLoS ONE18, e0294236 (2023).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Szklarczyk, D. et al. The STRING database in 2023: protein–protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res.51, D638–D646 (2023).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Passaro, S. et al. Boltz-2: towards accurate and efficient binding affinity prediction. Preprint at https://www.biorxiv.org/content/10.1101/2025.06.14.659707v1 (2025).

2. Bhatt, P. et al. Artificial intelligence in pharmaceutical industry: revolutionizing drug development and delivery. Curr. Artif. Intell.2, E051223224198 (2024).

ArticleGoogle Scholar

1. Aundhia, C. et al. Impact of artificial intelligence on drug development and delivery. Curr. Top. Med. Chem.25, 1165–1184 (2025).

ArticleCASPubMedGoogle Scholar

1. Jung, S., Vatheuer, H. & Czodrowski, P. VSFlow: an open-source ligand-based virtual screening tool. J. Cheminform.15, 40 (2023).

ArticlePubMedPubMed CentralGoogle Scholar

1. Ni, B. et al. AutoDock-SS: AutoDock for multiconformational ligand-based virtual screening. J. Chem. Inf. Model.64, 3779–3789 (2024).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Cleves, A. E., Johnson, S. R. & Jain, A. N. Electrostatic-field and surface-shape similarity for virtual screening and pose prediction. J. Comput. Aided Mol. Des.33, 865–886 (2019).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Hawkins, P. C., Skillman, A. G. & Nicholls, A. Comparison of shape-matching and docking as virtual screening tools. J. Med. Chem.50, 74–82 (2007).

ArticleCASPubMedGoogle Scholar

1. Yan, X. et al. Enhancing molecular shape comparison by weighted Gaussian functions. J. Chem. Inf. Model.53, 1967–1978 (2013).

ArticleCASPubMedGoogle Scholar

1. Puertas-Martin, S. et al. OptiPharm: an evolutionary algorithm to compare shape similarity. Sci. Rep.9, 1398 (2019).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Ballester, P. J. & Richards, W. G. Ultrafast shape recognition to search compound databases for similar molecular shapes. J. Comput. Chem.28, 1711–1723 (2007).

ArticleCASPubMedGoogle Scholar

1. Fu, L. et al. ADMETlab 3.0: an updated comprehensive online ADMET prediction platform enhanced with broader coverage, improved performance, API functionality and decision support. Nucleic Acids Res.52, W422–W431 (2024).

ArticlePubMedPubMed CentralGoogle Scholar

1. Daina, A., Michielin, O. & Zoete, V. SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci. Rep.7, 42717 (2017).

ArticlePubMedPubMed CentralGoogle Scholar

1. Gu, Y. et al. admetSAR3.0: a comprehensive platform for exploration, prediction and optimization of chemical ADMET properties. Nucleic Acids Res.52, W432–W438 (2024).

ArticlePubMedPubMed CentralGoogle Scholar

1. Pires, D. E., Blundell, T. L. & Ascher, D. B. pkCSM: predicting small-molecule pharmacokinetic and toxicity properties using graph-based signatures. J. Med. Chem.58, 4066–4072 (2015).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Moulis, M. & Vindis, C. Methods for measuring autophagy in mice. Cells6, 14 (2017).

ArticlePubMedPubMed CentralGoogle Scholar

1. Liu, J. & Li, L. Targeting autophagy for the treatment of Alzheimer’s disease: challenges and opportunities. Front. Mol. Neurosci.12, 203 (2019).

ArticleCASPubMedPubMed CentralGoogle Scholar

1. Choo, M. Z. Y. & Chai, C. L. L. The polypharmacology of natural products in drug discovery and development. Annual Reports in Medicinal Chemistry (ed. Karl-Heinz, A.) Vol. 61, 55–100 (Elsevier, 2023).

2. Naskar, R. et al. A critical appraisal of geroprotective activities of flavonoids in terms of their bio-accessibility and polypharmacology. Neurochem. Int.180, 105859 (2024).

ArticleCASPubMedGoogle Scholar

Download references

Acknowledgements

This study was supported by the Dr. Stanley Ho Medical Development Foundation (file number SHMDF-OIRFS/2024/002), the National Natural Science Foundation of China (number 82271455), the Science and Technology Development Fund, Macau SAR (file numbers 0040/2024/RIB1 and 0002/2025/NRP), the Guangdong Basic and Applied Basic Research Foundation (file number 2024A1515012740) and the University of Macau grants (file numbers MYRG-GRG2024-00238-ICMS-UMDF and MYRG-GRG2023-00089-ICMS-UMDF) awarded to J.-H.L. The Wellcome Leap Dynamic Resilience programme (co-funded by Temasek Trust) was awarded to X.X. and Z.N. The EPSRC Doctoral Training Programme was awarded to X.X. The ERC IMI (101005122), the H2020 (952172), the MRC (MC/PC/21013), the Royal Society (IEC\NSFC\211235), the NVIDIA Academic Hardware Grant Program, the SABER project supported by Boehringer Ingelheim Ltd., NIHR Imperial Biomedical Research Centre (RDA01), the Wellcome Leap Dynamic Resilience programme (co-funded by Temasek Trust), UKRI guarantee funding for Horizon Europe MSCA Postdoctoral Fellowships (EP/Z002206/1), UKRI MRC Research Grant, TFS Research Grants (MR/U506710/1), and the UKRI Future Leaders Fellowship (MR/V023799/1) were awarded to G.Y. We thank the members of the Faculty of Health Sciences Animal Facility at the University of Macau for their experimental and technical support. Parts of figures were created using templates from Servier Medical Art (https://smart.servier.com/), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

Author information

Author notes

1. These authors contributed equally: Yu Dong, Xianglu Xiao, Xu-Xu Zhuang, Wenfan Wu.

Authors and Affiliations

1. State Key Laboratory of Mechanism and Quality of Chinese Medicine, Institute of Chinese Medical Sciences, University of Macau, Macau SAR, China

Yu Dong (董雨),Xu-Xu Zhuang (庄旭旭),Zi-Ying Wang (王子颖),Shuang Zhang (张爽),Jin-Tao Li (李进涛),Ke Zhang (张可),Wen-Yu Fu (付文钰),Jun-Ming Chen (陈俊铭),Shi Hang Xiong (熊世航),Huanxing Su,Jian-Bo Wan,Hua Yu (余华),Defang Ouyang&Jia-Hong Lu (路嘉宏)

1. Bioengineering Department and Imperial-X, Imperial College London, London, UK

Xianglu Xiao (肖祥路),Shenglong Deng (邓胜龙),Krinos Li (李泽瑞)&Guang Yang (楊光)

1. Mindrank AI Ltd, Hangzhou, China

Xianglu Xiao (肖祥路),Wenfan Wu (吴文凡),Chao Ma (马超),Wangzhen Jin (金王震),Xurui Jin (晋旭锐),Qiwei Cai (蔡绮薇)&Zhangming Niu (牛张明)

1. AI Research Center, MindRank Technologies Limited, London, UK

Xianglu Xiao (肖祥路),Shenglong Deng (邓胜龙)&Zhangming Niu (牛张明)

1. Department of Bioinformatics and Systems Biology, Huazhong University of Science and Technology, College of Life Sciences and Technology, Wuhan, China

Wenfan Wu (吴文凡)

1. Guangzhou National Laboratory, Guangzhou, China

Wenfan Wu (吴文凡)

1. Interdisciplinary Institute for Personalized Medicine in Brain Disorders, School of Traditional Chinese Medicine, Jinan University, Guangzhou, China

Zi-Ying Wang (王子颖)

1. Department of Chemistry, School of Science, Southern University of Science and Technology, Shenzhen, China

Shuang Zhang (张爽)&Chris Soon Heng Tan

1. Department of Biomedical Sciences, Faculty of Health Sciences, Ministry of Education Frontiers Science Center for Precision Oncology, University of Macau, Macau SAR, China

Han-Ming Shen (沈汉明)

1. Mr. & Mrs. Ko Chi-Ming Centre for Parkinson’s Disease Research, School of Chinese Medicine, Hong Kong Baptist University, Hong Kong SAR, China

Min Li (李敏)

1. Department of Public Health and Medicinal Administration, Faculty of Health Sciences, University of Macau, Macau SAR, China

Defang Ouyang

1. Faculty of Life and Health Sciences, Shenzhen University of Advanced Technology (SUAT), Shenzhen, China

Keqiang Ye (叶克强)

1. Department of Clinical Molecular Biology, University of Oslo and Akershus University Hospital, Lørenskog, Norway

Evandro F. Fang (方飛)

1. The Norwegian Centre on Healthy Ageing (NO-Age) and the Norwegian National anti-Alzheimer’s Disease (NO-AD) Networks, Oslo, Norway

Evandro F. Fang (方飛)

1. National Heart and Lung Institute, Imperial College London, London, UK

Guang Yang (楊光)&Zhangming Niu (牛张明)

1. Cardiovascular Research Centre, Royal Brompton Hospital, London, UK

Guang Yang (楊光)

1. School of Biomedical Engineering & Imaging Sciences, King’s College London, London, UK

Guang Yang (楊光)

Authors

1. Yu Dong (董雨)View author publications Search author on:PubMedGoogle Scholar
2. Xianglu Xiao (肖祥路)View author publications Search author on:PubMedGoogle Scholar
3. Xu-Xu Zhuang (庄旭旭)View author publications Search author on:PubMedGoogle Scholar
4. Wenfan Wu (吴文凡)View author publications Search author on:PubMedGoogle Scholar
5. Zi-Ying Wang (王子颖)View author publications Search author on:PubMedGoogle Scholar
6. Shuang Zhang (张爽)View author publications Search author on:PubMedGoogle Scholar
7. Jin-Tao Li (李进涛)View author publications Search author on:PubMedGoogle Scholar
8. Ke Zhang (张可)View author publications Search author on:PubMedGoogle Scholar
9. Wen-Yu Fu (付文钰)View author publications Search author on:PubMedGoogle Scholar
10. Jun-Ming Chen (陈俊铭)View author publications Search author on:PubMedGoogle Scholar
11. Shi Hang Xiong (熊世航)View author publications Search author on:PubMedGoogle Scholar
12. Shenglong Deng (邓胜龙)View author publications Search author on:PubMedGoogle Scholar
13. Krinos Li (李泽瑞)View author publications Search author on:PubMedGoogle Scholar
14. Chao Ma (马超)View author publications Search author on:PubMedGoogle Scholar
15. Wangzhen Jin (金王震)View author publications Search author on:PubMedGoogle Scholar
16. Xurui Jin (晋旭锐)View author publications Search author on:PubMedGoogle Scholar
17. Qiwei Cai (蔡绮薇)View author publications Search author on:PubMedGoogle Scholar
18. Han-Ming Shen (沈汉明)View author publications Search author on:PubMedGoogle Scholar
19. Min Li (李敏)View author publications Search author on:PubMedGoogle Scholar
20. Huanxing SuView author publications Search author on:PubMedGoogle Scholar
21. Jian-Bo WanView author publications Search author on:PubMedGoogle Scholar
22. Hua Yu (余华)View author publications Search author on:PubMedGoogle Scholar
23. Defang OuyangView author publications Search author on:PubMedGoogle Scholar
24. Keqiang Ye (叶克强)View author publications Search author on:PubMedGoogle Scholar
25. Evandro F. Fang (方飛)View author publications Search author on:PubMedGoogle Scholar
26. Chris Soon Heng TanView author publications Search author on:PubMedGoogle Scholar
27. Guang Yang (楊光)View author publications Search author on:PubMedGoogle Scholar
28. Zhangming Niu (牛张明)View author publications Search author on:PubMedGoogle Scholar
29. Jia-Hong Lu (路嘉宏)View author publications Search author on:PubMedGoogle Scholar

Contributions

J.-H.L. and Z.N. conceived and supervised the project. Y.D. designed and conducted the experiments, acquired and analysed the data, and drafted the paper. X.X. and W.W. contributed to the development of AI-driven drug screening and network platform. X.-X.Z. and Z.-Y.W. contributed to the mouse and nematode experiments. S.Z. and C.S.H.T. contributed to the thermal proteome profiling. K.Z. contributed to the plasmid construction. J.-T.L., W.-Y.F., J.-M.C., S.H.X., J.-B.W. and H.Y. contributed to the LC-MS/MS analysis. S.D., K.L., C.M., W.J., X.J. and Q.C. contributed to the AI-driven drug screening and network platform. H.-M.S., H.S., M.L., D.O., K.Y., E.F.F. and G.Y. contributed to the revision of the paper. All authors reviewed, read and approved the final paper.

Corresponding authors

Correspondence to Zhangming Niu (牛张明) or Jia-Hong Lu (路嘉宏).

Ethics declarations

Competing interests

E.F.F. is a co-owner of Fang-S Consultation AS (organization number 931 410 717) and NO-Age AS (organization number 933 219 127); he has an MTA with LMITO Therapeutics Inc., a CRADA arrangement with ChromaDex, a commercialization agreement with Molecule AG/VITADAO, and MTAs with GeneHarbor (Hong Kong) Biotechnologies Limited and Hong Kong Longevity Science Laboratory; he is a consultant to MindRank AI, NYO3, AgeLab (Vitality Nordic AS) and Hong Kong Longevity Science Laboratory. Z.N., X.X., W.W., W.J., C.M., X.J., Q.C. and S.D. are employers of MindRank AI.

Peer review

Peer review information

Nature Biomedical Engineering thanks Agustín Ibáñez and Ho Ko for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Physicochemical property distribution of compounds and variational autoencoder training performance.

a,b, Key physicochemical and drug-likeness parameters of the virtual screening compound library and 50 reference mTOR-independent autophagy enhancers. Properties analyzed include molecular weight (MW), lipophilicity (LogP), hydrogen bond donors (HBDs), hydrogen bond acceptors (HBAs), topological polar surface area (TPSA), and rotatable bond count. c, Training and validation loss curves for the variational autoencoder on the DeepDrugDiscovery platform. The left panel shows the total loss (combined reconstruction and Kullback-Leibler (KL) divergence loss), and the right panel shows the Kullback-Leibler divergence loss alone.

Extended Data Fig. 2 Benchmarking of our AI platform against public platforms.

a, Comparison of computational time for DeepDrugDiscovery versus VSFlow, AutoDock-SS, eSim-pfastf, ROCS, WEGA, OptiPharm, and USR (s: seconds; h: hours; d: days). b-d, Comparison of Caco-2 permeability, MDCK efflux ratio, and blood-brain barrier penetration classification performance across ADMET Ranker, ADMETlab 3.0, SwissADME, admetSAR 3.0, pkCSM (Deep-PK), along with high-confidence prediction subsets from ADMET Ranker and pkCSM predictions. Performance metrics include accuracy, AUC, F1-score, precision, and recall. Missing bars indicate that a given platform does not generate predictions for the specific endpoint or metric. e, Scatter plot showing the correlation between experimentally derived LogBB values and the predicted blood-brain barrier penetration scores from ADMET Ranker. The dashed red line indicates the ideal correlation (predicted = experimental), and cyan circles represent individual data points (compounds).

Extended Data Fig. 3 Validation of mTOR-independent autophagy inducers.

a,b, WB analysis of LC3B-II and SQSTM1/p62 protein levels in N2a cells treated with the indicated compounds (20 μM) or Torin1 (1 μM) for 24 h (n = 3 independent biological replicates). One-way ANOVA followed by Dunnett’s multiple comparisons test (a,b). c-f, WB analysis of LC3B-II and SQSTM1/p62 protein levels in N2a cells treated with the indicated compounds (20 μM) or Torin1 (1 μM) in the absence or presence of Bafilomycin A 1 (100 nM) for 24 h (n = 3 independent biological replicates). One-way ANOVA followed by Tukey’s multiple comparisons test (c-f). Quantitative data are shown as mean ± SD. ns, not significant, *P< 0.05, **P< 0.01, ***P< 0.001.

Extended Data Fig. 4 Evaluation of abnormal protein degradation effect and autophagic flux by candidate autophagy enhancers.

a, Chemical structures, molecular formulas and weights of four novel autophagy inducers. b-f, WB analysis of total tau, LC3B-II and SQSTM1/p62 protein levels in PC12 cells stably expressing Tau P301L or APP KM595/596NL, V642F treated with the indicated compounds (20 μM) (n = 3 independent biological replicates). g, PC12 cells stably expressing mRFP-EGFP-LC3B were treated with the indicated compounds (20 μM) in the absence or presence of BAF (100 nM) or SAR405 (1 μM) for 24 h, and Opera Phenix plus high-content screening system was applied to capture the fluorescent images (scale bar, 10 μm) (n = 2 independent biological replicates). h-i, PC12 cells stably expressing mRFP-EGFP-LC3B were treated with the indicated compound for 24 h or 48 h, and Opera Phenix plus high-content screening system was applied to capture the fluorescent images (scale bar, 10 μm). The average number of red and yellow dots per cell was quantified (cells≥30, n = 3 independent biological replicates). One-way ANOVA followed by Dunnett’s multiple comparisons test (b-f, i). Quantitative data are shown as mean ± SD. ns, not significant, *P< 0.05, **P< 0.01, ***P< 0.001.

Extended Data Fig. 5 Ombuin and 2-Hydroxycinnamic acid clear abnormal proteins in an autophagy-dependent manner.

a-n, WB analysis of total tau, phosphorylated tau (at multiple sites), CTF-β, and CTF-α protein levels in PC12 cells stably expressing Tau P301L or APP KM595/596NL, V642F treated with the indicated compounds (20 μM) in the absence or presence of BAF (100 nM) or SAR405 (1 μM) (n = 3 independent biological replicates). One-way ANOVA followed by Tukey’s multiple comparisons test (a-f, i-n). Quantitative data are shown as mean ± SD. ns, not significant, *P< 0.05, **P< 0.01, ***P< 0.001. Original unprocessed WB gel data are in Source Data Fig. 6.

Source data

Extended Data Fig. 6 Evaluation of the neuroprotective effects of Ombuin and 2-Hydroxycinnamic acid in the 3×Tg-AD mouse models.

a,b, Immunofluorescence analysis of phosphorylated tau and Aβ in the hippocampal CA1 region. Relative fluorescence intensity for each treatment group was quantified (n = 7 mice per group). c-d, WB analysis of CTF-α and LC3B-II protein levels in mouse hippocampal tissues (n = 3 mice per group). e, Body weight changes in 3×Tg-AD mice throughout the treatment period (n = 7 mice per group). One-way ANOVA followed by Dunnett’s multiple comparisons test (a-d). Quantitative data are shown as mean ± SD. ns, not significant, *P< 0.05, **P< 0.01, ***P< 0.001.

Extended Data Fig. 7 Integrated profiling suggests candidate mechanisms for autophagy induction by Ombuin and 2-Hydroxycinnamic acid.

a, Schematic workflow of the multi-faceted approach used to elucidate the mechanisms of action of the compounds. b,c, Thermal proteome profiling of compound-treated lysates identifies drug-binding candidates. Differential thermal stability was assessed by comparing compound-treated and vehicle-control samples using moderated t-tests implemented in the limma package (v3.62) in R. (P< 0.05, red dot: stabilized, blue dot: destabilized; P< 0.05, log2FC > 0.5, protein name: labeled). The dashed line indicates the P = 0.05 threshold. d,e, Protein-protein interaction (PPI) networks were constructed for candidate targets of Ombuin (Pip4k2a, Mtmr1) or 2-Hydroxycinnamic acid (Mapk8ip3, Ikbkb) using the STRING database to explore their coordinated roles in autophagy regulation. Edges represent predicted functional associations, colored by evidence type: purple (experimental), light blue (database), green (neighborhood), red (fusion), blue (co-occurrence), black (co-expression), light green (text mining), light purple (homology). f,g, Predicted binding poses for each compound-target pair, generated using the Boltz-2 deep learning method and visualized with PyMOL. Hydrogen bonds and hydrophobic interactions are indicated by blue lines and gray dashed lines, respectively.

Extended Data Fig. 8 Integrated profiling reveals candidate mechanisms for autophagy induction by Ombuin and 2-Hydroxycinnamic acid.

a-d, Integrated Sankey and dot plot illustrating the biological process and signaling/metabolic pathways (analyzed using Enrichr) associated with target proteins of Ombuin (Omb) and 2-Hydroxycinnamic acid (2-HCA), identified by thermal proteome profiling (P< 0.05, log2FC > 0.5). Statistical significance was determined by Fisher’s exact test, and multiple testing was corrected using the Benjamini-Hochberg method. Detailed data are provided in Supplementary Tables 7–10. e-h, GO and KEGG enrichment analysis of the protein–protein interaction (PPI) network constructed using the STRING database for Omb and 2-HCA candidate targets, highlighting functional pathways related to autophagy regulation. i-j, Predicted binding poses between each compound and its target protein, generated using the Boltz-2 deep learning method and visualized with PyMOL. Hydrogen bonds and hydrophobic interactions are indicated by blue lines and gray dashed lines, respectively.

Extended Data Fig. 9 Molecular dynamics analysis of Ombuin and 2-Hydroxycinnamic acid interactions with target proteins.

a-c, Analysis plots from molecular dynamics simulations, including: (1) root mean square deviation (RMSD) of protein and ligand; (2) radius of gyration (Rg) of the complex; (3) root mean square fluctuation (RMSF) of protein residues; (4) solvent-accessible surface area (SASA) buried upon binding; (5) binding free energy; and (6) number of hydrogen bonds formed during the simulation.

Supplementary information

Supplementary Information (download PDF )

Supplementary Tables 1–14 and Supplementary Methods.

Reporting Summary (download PDF )

Source data

Source Data Fig. 3 (download PDF )

Unprocessed western blots for Fig. 3c–f.

Source Data Fig. 4 (download PDF )

Unprocessed western blots for Fig. 4a,c–e.

Source Data Fig. 5 (download PDF )

Unprocessed western blots for Fig. 5a,e,h,i.

Source Data Fig. 6 (download PDF )

Unprocessed western blots for Fig. 6c.

Source Data Fig. 8 (download PDF )

Unprocessed western blots for Fig. 8c–e.

Source Data Extended Data Fig. 5 (download PDF )

Unprocessed western blots for Extended Data Fig. 5g,h.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Cite this article

Dong, Y., Xiao, X., Zhuang, XX. et al. DeepDrugDiscovery identifies blood–brain barrier permeable autophagy enhancers for Alzheimer’s disease. Nat. Biomed. Eng (2026). https://doi.org/10.1038/s41551-026-01667-x

Download citation

• Received: 04 April 2025

• Accepted: 23 March 2026

• Published: 24 April 2026

• Version of record: 24 April 2026

• DOI: https://doi.org/10.1038/s41551-026-01667-x

Anyone you share the following link with will be able to read this content:

Get shareable link

Sorry, a shareable link is not currently available for this article.

Copy shareable link to clipboard

Provided by the Springer Nature SharedIt content-sharing initiative

Subjects

• Alzheimer's disease
• High-throughput screening
• Machine learning

Close