Discovery of EP4 antagonists with image-guided explainable deep learning workflow
Problem
Manual compound library design and slow wet-lab screening limit efficient discovery of novel EP4-targeted cancer immunotherapy drugs.
Why it matters
EP4 antagonists could enhance anti-tumor immune responses, improving cancer immunotherapy outcomes.
What the AI did
LeadDisFlow, an image-guided explainable deep learning workflow, rapidly identified four potent EP4 antagonist compounds.
Results
Four potent antagonists with single-nanomolar activity identified; best compound ZY001 showed IC50 of 0.51 ± 0.02 nM with high selectivity and impaired PGE2-induced immunosuppressive gene expression in macrophages.
Impact
Yielded a potent, selective EP4 antagonist candidate for anti-tumor immunotherapy.
Record
| Venue | National Science Open |
|---|---|
| Published | 2025 |
| Work type | article |
| Authors | Pengsen Ma, Zhiyuan Cheng, Zhixiang Cheng, Yijie Wang, Xiaolei Chai, Bo Feng, Hongxin Xiang, Li Zeng et al. |
| Citations | 1 (OpenAlex, at import time) |
| References | 81 |
| Model or method | LeadDisFlow |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1360/nso/20240015 |
| Open access | diamond |
| Keywords | EP4 antagonist, deep learning, drug discovery, explainable AI, cancer immunotherapy, IC50 |
Importance rating
significant — thousands to millions affected
Cancer immunotherapy target with single-nanomolar antagonist found; abstract states potency (IC50 0.51 nM) but not disease burden or patient scale.