Cell Therapy · global
Bringing CAR T Cells into Ovarian Cancer: AI-Assisted Therapy Reaches First Patient Dosing
GB-5267 combines a tumor-recognizing receptor with IL-18 “armor” in an attempt to help immune cells keep fighting in solid tumors. Its first-in-human trial has begun, and whether it can work safely is only now being put to the test.
New treatment options remain urgently needed for patients with recurrent ovarian cancer or ovarian cancer resistant to platinum-based chemotherapy. Getting immune cells to recognize cancer cells is only the first hurdle for cell therapy; they must also reach the tumor, survive in an environment that suppresses immune responses, and sustain their attack. GB-5267, jointly developed by Generate Biomedicines and Roswell Park Comprehensive Cancer Center in the United States, has now begun testing this design in humans.
On October 1, 2026, the partners announced that the first patient had been dosed in this first-in-human Phase 1 trial. The study, numbered NCT07489287 and led by Emese Zsiros, Chair of Gynecologic Oncology at Roswell Park, is enrolling patients with recurrent or platinum-resistant ovarian cancer. Its primary focus is to assess safety and tolerability, while identifying a tolerable dose and preliminary antitumor activity. The first dosing marks the start of clinical investigation; it does not yet allow conclusions about efficacy.
GB-5267 is an autologous CAR T-cell therapy: researchers collect a patient’s T cells, genetically modify and expand them in culture, and then infuse them back into the body. It targets MUC16, an antigen that, according to the center, is overexpressed in more than 80% of ovarian tumors. Generate further states that the design recognizes MUC16 remaining on the tumor cell membrane, with the aim of concentrating the attack on cancer cells and reducing interference from antigen circulating freely in the blood.
The “armor” refers to the incorporation of the cytokine IL-18, intended to help T cells resist immune suppression in the tumor microenvironment. AI is used to explore and evaluate different chimeric antigen receptor architectures. The company says its screening considers not only how tightly a receptor binds to its target, but also killing capacity, cell proliferation, and sustained activity. The real test is whether these functions can all hold up together in patients.
The trial also adjusts the administration approach. Because ovarian cancer often spreads within the peritoneal cavity, the study will evaluate both intravenous infusion and intraperitoneal administration to explore how to get the therapeutic cells to tumor sites. The study does not require lymphodepleting chemotherapy before infusion, in an attempt to reduce the burden of this conditioning step. Whether the cells can expand sufficiently and maintain their activity without that conditioning will also require clinical data to answer.
Omitting lymphodepleting chemotherapy does not mean that the cell therapy itself is without risks. Generate says the medical team will closely monitor cytokine release syndrome, immune effector cell-associated neurotoxicity syndrome, and other unexpected reactions. A safety monitoring committee must review the available data before each dose escalation.
The two announcements currently provide no patient efficacy results and do not detail the AI training data, comparative receptor design data, or quantified preclinical results. They are therefore insufficient to determine how much improvement AI has contributed. The trial must next establish whether the engineered cells can safely reach tumors, persist, and produce measurable antitumor effects. This human evidence will determine how far the strategy can go.