AI-guided strategy uncovers new gp130 inhibitor for colorectal cancer
image: Design strategy and structure optimization of novel gp130-targeting compounds.
Credit: Targetome
A research team has developed an artificial intelligence-assisted strategy to discover new inhibitors of glycoprotein 130 (gp130), a key signaling receptor involved in inflammation-driven tumor progression, and identified a promising small molecule for colorectal cancer treatment. By combining transfer learning, natural-product screening, chemical optimization, and biological validation, the team discovered compound 8a, an indolopyridine-based molecule that directly binds gp130, suppresses the JAK2/STAT3 signaling pathway, and induces cancer cell apoptosis.
gp130, also known as IL-6β or CD130, is the shared signal-transducing receptor for the interleukin-6 cytokine family and plays a central role in chronic inflammation, immune regulation, and cancer progression. Abnormal activation of the IL-6/gp130/JAK/STAT3 signaling axis has been linked to tumor growth, immune escape, and poor therapeutic response, making gp130 an attractive anticancer target. However, only a small number of gp130 inhibitors have been reported, and many suffer from weak binding affinity, limited antitumor validation, or poor suitability for clinical translation. At the same time, the scarcity of known gp130-targeting compounds makes it difficult to build reliable conventional prediction models, creating a need for new drug-discovery strategies that can work under data-limited conditions.
A study (DOI: 10.48130/targetome-0026-0010) published in Targetome on 20 March 2026 by Yixian Liao's & Wenying Yu's team, China Pharmaceutical University, reports that an AI-guided transfer learning model enabled the discovery and optimization of compound 8a as a direct gp130 inhibitor with significant activity against colorectal cancer.
The researchers first built a transfer learning model using a larger dataset of STAT3 inhibitors as the source task, because STAT3 is a key downstream effector in the gp130 signaling pathway. The model was then fine-tuned with a small literature-curated dataset of gp130-related compounds. Compared with a model trained only on limited gp130 data, the transfer learning model showed better predictive performance and more stable ranking ability, supporting its use for prioritizing candidate molecules. The team then screened 2,560 natural products from a commercial natural-product library. After ADMET-based filtering, 122 candidates remained, and the fine-tuned model ranked them according to predicted gp130-inhibitory potential. Evodiamine was selected as a lead scaffold because of its predicted activity and structural complexity. Based on evodiamine and the related natural product rutaecarpine, the researchers synthesized and optimized a series of derivatives, ultimately identifying compound 8a as the best candidate. Biological assays showed that 8a displayed stronger antiproliferative activity against HT-29 colorectal cancer cells than evodiamine, rutaecarpine, and bazedoxifene. Isothermal titration calorimetry confirmed that 8a directly bound gp130 with a dissociation constant of 2.17 μmol·L−1, indicating markedly improved affinity compared with the parent scaffolds. Molecular docking suggested that 8a interacts with the gp130 D1 domain through hydrogen-bonding, van der Waals, and hydrophobic interactions. Drug affinity responsive target stability and cellular thermal shift assays further supported gp130 as the direct cellular target of 8a. Mechanistic experiments showed that 8a reduced phosphorylation of JAK2 and STAT3, blocked STAT3 DNA-binding activity, and downregulated STAT3-related proteins including Bcl-2 and Cyclin D1. In gp130-knockdown cells, the inhibitory effect of 8a was weakened, further confirming gp130 dependency. The compound also inhibited colony formation and induced mitochondria-related apoptosis through activation of caspase-9, caspase-3, and PARP cleavage. In an HT-29 xenograft mouse model, daily treatment with 8a achieved 56.20% tumor growth inhibition at 20 mg·kg−1.
Overall, this study demonstrates how transfer learning can help overcome limited target-specific data in drug discovery and guide the identification of new anticancer compounds. By integrating artificial intelligence, medicinal chemistry, and experimental validation, the researchers identified compound 8a as a promising gp130-targeted small molecule for colorectal cancer research. The strategy may also be extended to other signaling pathways where therapeutic targets are important but available inhibitor data remain scarce.
###
References
DOI
Original Source URL
https://doi.org/10.48130/targetome-0026-0010
Funding information
This work was funded by the National Natural Science Foundation of China (Grant No. 82473791) and the Basic Research Program of Jiangsu Province (Grant No. BK20250102).
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.