Abstract: SA-PO0092
Artificial Intelligence (AI) Co-Scientist-Guided Drug Repurposing Identifies Diosmin as a GPER1-Dependent Antifibrotic Candidate in Renal Fibrosis
Session Information
- Artificial Intelligence: Kidney Imaging, Pathology, and Precision Phenotyping
October 24, 2026 | Location: Exhibit Hall A, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Artificial Intelligence, Digital Health, and Data Science
- 300 Artificial Intelligence, Digital Health, and Data Science
Authors
- Gu, Xiangchen, Shanghai University of Traditional Chinese Medicine Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Shanghai, China
- Wang, Chenxi, Shanghai University of Traditional Chinese Medicine Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Shanghai, China
- Gu, Yaodong, Shanghai University of Traditional Chinese Medicine Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Shanghai, China
- Cheng, Ye, Shanghai University of Traditional Chinese Medicine Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Shanghai, China
- Chen, Min, Shanghai University of Traditional Chinese Medicine Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Shanghai, China
Background
Mechanistically validated renal targets often fail to advance into therapy because actionable modulators are difficult to identify and validate. GPER1, a non-classical estrogen receptor, was previously shown by our group to regulate macrophage activation and renal fibrosis, but its therapeutic translatability remains unexplored. We asked whether a kidney-focused AI co-scientist workflow could convert this mechanism into a testable drug-repurposing hypothesis.
Methods
We developed the Virtual Kidney Drug Repurposing Lab, a multi-agent AI co-scientist framework integrating LLM-orchestrated scientific agents, structure-guided receptor-ligand modeling informed by the cryo-EM GPR30-Gq complex, repurposing-library prioritization, cross-agent critique, and kidney-specific validation design. GPER1 was predefined as the biological target, and 2,308 repurposing candidates were ranked using a composite evidence score incorporating predicted pocket occupancy, receptor-ligand interactions, structural confidence, and translational feasibility. Top candidates were advanced to calcium influx assays, primary renal tubular epithelial cell injury models, and an aristolochic acid I-induced chronic kidney injury model in wild-type and Gper1-deficient mice.
Results
The framework prioritized diosmin as a top-ranked candidate predicted to occupy the GPER1 ligand-binding pocket and engage GLU115, GLU218, and GLU275. Diosmin induced calcium influx in GPER1-expressing cells, consistent with agonist-like receptor activation, and suppressed stress-induced inflammatory activation in primary renal tubular epithelial cells. In AAI-induced chronic kidney injury, therapeutic-phase diosmin reduced tubular injury, collagen deposition, serum creatinine, and renal expression of fibrotic and inflammatory markers, including Col1a1, Tgfb1, Fn1, Acta2, Nlrp3, Il1b, Ccl2, Cd68, and Cd86. These effects were abolished in Gper1-deficient mice, establishing genetic target dependency.
Conclusion
A kidney-focused AI co-scientist framework identified diosmin as a functionally active, genetically validated GPER1-dependent anti-fibrotic candidate. This work provides a target-resolved model of AI-assisted drug repurposing in nephrology and supports human-AI co-scientist workflows to accelerate translation from renal disease mechanisms to testable therapeutics.
Acknowledgment
This study is supported by the National Natural Science Foundation of China (Grant No. 82474175, to X. G.)
Funding
- Government Support – Non-U.S.