GB-5267, an AI-designed cell therapy from Generate Biomedicines, has entered clinical testing with the first patient now dosed in a Phase I trial. The study targets ovarian cancer patients whose disease has either come back or stopped responding to platinum-based chemotherapy—one of the toughest groups to treat. The company is betting that its artificial intelligence platform can build a better CAR T cell, one that lasts longer and works harder inside the body, not just one that binds tightly to its target.
On October 1, 2026, Roswell Park Comprehensive Cancer Center and Generate Biomedicines announced the first patient had received GB-5267 in a clinical trial registered as NCT07489287. According to an official press release from Roswell Park, this is the first time GB-5267 is being tested in humans. The main goal is to check safety and tolerability, with Emese Zsiros, MD, PhD, head of gynecologic oncology at Roswell Park, leading the trial. At the time of the announcement, no results on safety or effectiveness had been shared.
For each patient, T cells are collected, genetically modified, and expanded at Roswell Park's GMP manufacturing facility before being returned to the patient, confirming the autologous nature of the therapy and the center's in-house production capabilities.
This trial breaks from standard CAR T protocols by skipping lymphodepleting chemotherapy before infusion. By leaving out this step, the team hopes to avoid the toxic side effects that often come with lymphodepletion, a common hurdle in cell therapy. GB-5267 is given both through the vein and directly into the abdomen, using a dual-delivery approach in a planned group of about 18 patients at Roswell Park in Buffalo, New York.
Targeting MUC16 with armored CAR T cells
GB-5267 is built to go after MUC16, also called CA125, a protein found on the surface of more than 80% of ovarian tumors. MUC16 is tied to tumor growth, spread, and the cancer's ability to dodge the immune system. The therapy is designed to spot the membrane-bound form of MUC16 on tumor cells, not the free-floating CA125 in the blood. Generate Biomedicines explains this in their technical background. What makes GB-5267 different is its "armored" setup: the CAR T cells are engineered to release IL-18 at the tumor site. This local cytokine release is meant to fight off the immune-suppressing environment around the tumor and help the T cells multiply and stick around. In preclinical studies, IL-18-secreting, MUC16-targeted CAR T cells showed stronger expansion and persistence in ovarian tumor models.
Generate Biomedicines says its AI-driven process shaped the CAR design by focusing on how T cells behave—how strong they are, how well they multiply, and how long they last—rather than just how tightly they bind to MUC16. This reflects a shift in cell therapy, where more companies are using computational tools to fine-tune therapies before they reach patients.
The company has set near-term goals for the program that include establishing a safety profile, determining a reproducible dose, and conducting an initial assessment of antitumor activity; as of now, no patient results have been published.
Competitive landscape and clinical context
For women whose ovarian cancer has stopped responding to platinum-based chemotherapy, current treatments include single-agent chemotherapy, bevacizumab-based regimens, and mirvetuximab soravtansine-gynx (Elahere), an antibody-drug conjugate targeting folate receptor alpha that gained full US FDA approval in 2024. Other therapies aimed at MUC16 are also moving forward. Regeneron Pharmaceuticals is developing ubamatamab (REGN4018), a MUC16×CD3 bispecific antibody now in Phase I/II and Phase II trials for recurrent ovarian cancer. This bispecific antibody works differently from GB-5267's autologous CAR T approach.
The field is seeing a range of CAR T strategies for ovarian cancer. As reported earlier, other CAR T therapies targeting different ovarian cancer antigens have shown early signs of disease stabilization, but lasting benefits have yet to be proven.
Editorial perspective
By starting a first-in-human trial of an AI-optimized, IL-18-armored CAR T therapy without lymphodepletion, Generate Biomedicines is testing whether computational design and local cytokine release can break through the tough barriers of the ovarian tumor environment. Skipping lymphodepleting chemotherapy is a bold step that, if it works, could change how cell therapies are given for solid tumors. Still, with only preclinical data on T-cell persistence and no clinical results yet, it's too soon to know if this approach will pay off. The field is watching to see if this AI-driven strategy can finally move the needle for patients where so many other treatments have fallen short.
