Ischemic stroke remains a leading cause of death and long-term disability worldwide. Current treatments, including thrombolysis and mechanical thrombectomy, can restore cerebral blood flow, but their narrow therapeutic windows, strict eligibility requirements, and bleeding risks limit their use. Developing drugs for stroke is particularly difficult because ischemic injury involves interconnected processes such as oxidative stress, inflammation, mitochondrial dysfunction, neuronal death, and vascular damage. Conventional screening is often too slow and narrowly targeted to address this complexity. Therefore, an integrated approach is needed to rapidly identify clinically accessible compounds capable of acting across multiple stroke-related pathways while also revealing their mechanisms and optimizable chemical features.
A study (DOI: 10.48130/targetome-0026-0034) published in Targetome on 12 August 2026 by Guangji Wang's & Xinuo Li's team, China Pharmaceutical University, reports a multilayered screening and validation strategy that identified CBD and linked its neurovascular protection to NRF2-dependent antioxidant activity and BMAL1-mediated angiogenesis.
The researchers first constructed pathway-specific machine-learning classifiers using molecular fingerprints and bioactivity data related to antioxidant activity, reactive oxygen species, NRF2, inflammatory signaling, and stroke. After genetic analyses suggested an association between anxiety-related traits and stroke risk, 255 annotated anxiolytic compounds were screened as a repurposing library. The models achieved an average area under the receiver operating characteristic curve above 0.89 on held-out test data. A Latent Gene Expression Graph Neural Network, or LGE-GNN, was then trained on LINCS L1000 perturbation profiles to predict drug-induced expression changes across 978 landmark genes. It reached a Pearson correlation of 0.9517 and was used with stroke single-cell transcriptomic data to rank compounds according to their ability to reverse disease-associated endothelial signatures. A Drug Decompose Net further analyzed molecular fragments and prioritized 43 candidates, ultimately converging with the other models on CBD and doramectin. CBD showed lower cytotoxicity and greater protection of oxygen-glucose-deprived neuronal cells. Fragment analysis identified CBD’s (+)-dipentene-containing region as a potentially favorable structural feature, although this computational prediction requires direct testing with CBD analogues. In mice subjected to middle cerebral artery occlusion, CBD at 20 and 40 mg/kg reduced infarct volume, improved neurological and motor performance, suppressed inflammatory mediators, lowered reactive oxygen species and malondialdehyde, and restored antioxidant activity. LGE-GNN predictions and molecular experiments identified NRF2 as a central mediator. Blocking NRF2 pharmacologically or deleting Nrf2 weakened CBD’s antioxidant and neuroprotective effects. DeepD2V modeling, CUT&Tag sequencing, RNA sequencing, spatial transcriptomics, and ChIP-qPCR identified Bmal1 as an NRF2-associated downstream effector. Silencing Bmal1 in endothelial cells impaired CBD-enhanced migration, wound closure, and capillary-like tube formation, supporting its role in vascular repair. Pharmacokinetic testing also showed that intranasal CBD produced faster absorption and higher early brain uptake than intraperitoneal delivery while achieving comparable overall brain exposure.
Overall, the study presents CBD as a proof-of-concept repurposed candidate rather than a newly discovered compound. More importantly, it demonstrates how AI can connect candidate selection with mechanistic validation, drug-delivery evaluation, and pharmacophore-guided optimization. Independent datasets, experiments in female animals, fuller dose-response studies, and direct testing of CBD derivatives will be needed before clinical translation. Nevertheless, the framework provides a promising roadmap for accelerating multi-target drug discovery in stroke and other multifactorial disorders.
###
References
DOI
10.48130/targetome-0026-0034
Original Source URL
https://doi.org/10.48130/targetome-0026-0034
Funding information
This study was supported by the Fundamental Research Funds for the Central Universities (No. 2632025ZD10, China); the Targeted Commissioned Project – In Vivo Fate and Intelligent Delivery of Multi-target Natural Drugs (No. SKLNMZZ202403, China); the Young Elite Scientists Sponsorship Program by CAST (No. 2024-2026QNRC001, China); the Sanming Project of Medicine in Shenzhen (No. SZSM202301035, China); the Haihe Laboratory of Cell Ecosystem Innovation Fund (No. 22HHXBSS00005, China); and the General Program of the National Natural Science Foundation of China (No. 82371365, China).
About Targetome
Targetome refers to the complete collection of molecular targets (e.g., proteins, RNA or DNA) that interact with and mediate the effect of a specific biomolecule, such as a drug, toxin, metabolites, transcription factor or microRNA, within a biological system. Targetome is an open access journal publishing rigorously peer-reviewed original research articles, reviews, break-through methods, and perspectives that advance our understanding, identification and validation of molecular targets for new drug development.