Researchers at the Perelman School of Medicine at the University of Pennsylvania have developed a human-in-the-loop artificial intelligence framework to accelerate the identification of target antigens for CAR T cell therapy. Published in Cell, the study demonstrates that combining large language models (LLMs) with expert human oversight effectively identifies viable targets, such as GPNMB in skin cancer, which showed tumor-killing activity in mouse models. This approach addresses the primary bottleneck in expanding CAR T cell therapy beyond blood cancers.
How AI Solves the Target Discovery Bottleneck
Finding a new target for CAR T cell therapy is currently a labor-intensive process, often compared to finding a “needle in a haystack,” according to lead author Daniel Baker, PhD. The volume of genomic and proteomic sequencing data has outpaced the capacity for manual analysis by human researchers. By integrating LLMs, the Penn Medicine team created a systematic method to scan massive datasets that would be impossible for a human to process individually.
The framework relies on the specific strengths of both entities: AI handles the breadth of data, while human experts provide the necessary deep-dive validation. This partnership ensures that the targets selected are not only statistically likely to be present on tumor cells but also biologically relevant for clinical treatment.
CAR T cell therapy was pioneered at Penn Medicine. While it has successfully treated various blood cancers for over a decade, expanding its reach to solid tumors—which make up the majority of cancer diagnoses—has been limited by the difficulty of identifying unique surface antigens that avoid harming healthy tissue.
Why GPNMB is a Target of Interest
As a proof-of-concept for their AI-driven approach, the team identified glycoprotein non-metastatic melanoma protein B (GPNMB) as a candidate for skin cancer treatment. According to the research published in Cell, this target demonstrated robust tumor-killing activity in preclinical mouse models. The selection of GPNMB highlights the utility of the framework in identifying targets in solid tumors, where immune-based therapies have historically faced significant hurdles.
This success is particularly notable because skin cancer, specifically melanoma, has shown a historical sensitivity to other immunotherapies like immune checkpoint inhibitors and tumor-infiltrating lymphocyte (TIL) therapy. By adding CAR T to this repertoire, researchers hope to create more comprehensive treatment options for patients who do not respond to existing standard-of-care therapies.
What Happens Next for CAR T Research
The success of the Penn Medicine AI framework suggests a shift toward more automated, data-driven drug discovery. Future efforts will likely focus on applying this model to other solid cancers, such as pancreatic or lung cancer, where identifying specific surface antigens remains the primary obstacle to developing effective CAR T therapies.
Beyond cancer, the methodology could potentially be adapted to identify targets for non-cancerous conditions. As the amount of available sequencing data continues to grow, the ability of AI models to “read” this data will become a standard component of clinical research, potentially cutting years off the traditional discovery timeline.
Frequently Asked Questions
- What is a “human-in-the-loop” AI framework? It is a system where AI performs data processing and initial screening, but human experts must review and validate the findings before they are considered for clinical development.
- Why is it hard to treat solid tumors with CAR T? Solid tumors often lack unique, surface-level antigens that are not also found on vital healthy tissues, making it difficult to engineer cells that attack the tumor without causing systemic toxicity.
- Is this therapy currently available for patients? No. The GPNMB-targeted CAR T cells described in this study were tested in mouse models. Clinical trials in humans are the necessary next step to determine safety and efficacy.
If you are interested in the evolution of immunotherapy, monitor updates from the June Lab at Penn Medicine, which continues to lead innovation in T cell engineering and clinical applications.
Are you interested in the intersection of AI and modern medicine? Subscribe to our newsletter to receive the latest updates on cancer research breakthroughs and clinical trial advancements delivered directly to your inbox.
