Disarming Cancer's Immune Evasion, Owkin Doses First Patient With AI-Optimized Therapy
Owkin has entered the clinic with a first-in-class EP2/EP4/DP1 triple inhibitor designed for solid tumors.
After May 2024 expansion of its oncology pipeline through a licensing agreement with Idorsia and the development of a dual EP2/EP4 inhibitor, Owkin has now entered the clinic with OKN4395—a first-in-class EP2/EP4/DP1 triple inhibitor designed for solid tumors.
The first patient in the INVOKE trial was dosed on January 22, 2025, marking the start of a study that will evaluate the drug’s early anti-tumor activity, safety, and tolerability. It is the first drug from Owkin to be fully developed using its AI-powered K1.0 operating system, from asset selection all the way to clinical trial design.
What’s special about it?
Prostaglandin E2 (PGE2) and Prostaglandin D2 (PGD2) are hormone-like molecules that the body naturally produces to manage inflammation. Historically, nonsteroidal anti-inflammatory drugs (NSAIDs) have been used to block the Cyclooxygenase enzyme (COX), which is responsible for producing these prostaglandins. Notably, epidemiological studies have linked long-term NSAID use with a reduced risk of certain cancers. Under normal circumstances, when PGE2 binds to its receptors—EP2 and EP4—or when PGD2 binds to its DP1 receptor on immune cells, the resulting signal helps to moderate the immune response, which is basic for preventing excessive inflammation and protecting healthy tissue.
Within the tumor environment, however, this normally protective mechanism becomes counterproductive. By dampening immune cell activity via these receptors, the process reduces the immune system's ability to recognize and attack cancer cells. Essentially, the tumor “exploits” this natural anti-inflammatory response to create an immune-suppressive niche that supports its growth and spread.
OKN4395 is engineered to block all three receptors—EP2, EP4, and DP1—thereby disrupting the immunosuppressive signals triggered by PGE2 and PGD2. This targeted inhibition is intended to maintain an active immune response in the tumor's vicinity, enhancing the body’s capacity to identify and eliminate tumor’s cells. While earlier strategies focused on inhibiting EP4 alone, combining EP2, EP4, and DP1 blockade is a new approach.
The trial
The Phase I study is split into two parts:
Phase Ia: Dose escalation to determine the safest and most effective dosage, both as a standalone treatment and in combination with pembrolizumab (Keytruda).
Phase Ib: Expansion into additional patient groups, targeting over 100,000 patients annually across the US, EU4, and UK.
The trial is being run by Epkin, an Owkin France-affiliated company, and will take place across multiple global sites.
Artificial Intelligence
OKN4395 is a test case for Owkin’s AI-driven approach to clinical trials. The company’s K1.0 operating system, formally announced at the JP Morgan Healthcare Conference, is designed to speed up drug development by integrating multimodal patient data with data-derived intelligence, and experimental validation.
K1.0 was used to:
Analyze disease biology and confirm the role of EP2, EP4, and DP1 in cancer immunosuppression, reinforcing the existing scientific understanding of these targets by analyzing large-scale datasets and integrating prior research.
Select the best patient populations, using AI-based biomarker analysis to identify subgroups most likely to benefit.
Use external control arms (digital twins) to generate early anti-tumor activity insights in Phase I, potentially reducing the need for large placebo-controlled trials.
Optimize trial design, refining endpoints, dosage strategies, and escalation protocols to enhance the probability of success.
See also: The Pivotal Role of AI in Clinical Trials: From Digital Twins to Synthetic Control Arms
Owkin describes K1.0 as an end-to-end system for drug discovery and development that merges computational analysis with real-world patient data from a federated network of research institutions. Built over eight years, K1.0 leverages multimodal patient data curated from thousands of cohorts—covering over a million patients—and incorporates data from projects like MOSAIC (spatial atlas of cancer cells).
Looking ahead, the next iteration—K2.0—will build on this infrastructure by introducing agentic AI that will take a more active role in scientific research and experimental planning. In this future model, AI will function as a collaborative partner, assisting researchers with data analysis, lab automation, and hypothesis generation, with the long-term goal of developing a “biological AGI” capable of untangling complex disease mechanisms.