Penn researchers use AI to find new CAR T therapy targets

Researchers at Penn Medicine have developed an artificial intelligence framework that combines large language models with human scientific expertise to systematically discover new targets for CAR T cell therapy, publishing their findings Wednesday in the journal Cell.miragenews+2

The team, led by Daniel Baker, who completed his doctorate at Penn in December 2025 under the mentorship of CAR T cell therapy pioneer Carl June and Zoltan Arany, chair of Physiology at Penn, built what they describe as a "human-in-the-loop" AI system designed to address one of the field's most persistent bottlenecks: identifying which antigens CAR T cells should be engineered to attack.miragenews+1

A Needle in a Growing Haystack

"Discovering a good CAR target is like trying to find a needle in a haystack, except the haystack keeps growing as more sequencing data becomes available," Baker said. "We thought this would be a strong use-case for AI because one of the strengths of large language models is the amount of data they can consider. Human experts excel at going deep, while LLMs are good at looking across a broad range of data."miragenews+1

The framework integrates single-cell RNA sequencing datasets and runs LLM-based simulations to nominate and prioritize potential targets, producing a shortlist for expert biological validation. The approach was designed to be disease-agnostic and compatible with future AI models.sciencedirect+1

From Algorithm to Tumor-Killing Cells

As a proof of concept, the team focused on skin cancer and identified glycoprotein non-metastatic melanoma protein B (GPNMB) as the top candidate. They then engineered CAR T cells targeting GPNMB, which showed robust tumor-killing activity in mouse models of multiple cancer types, including melanoma, leukemia, and colorectal cancer.cell+3

An accompanying Cell commentary noted that GPNMB CAR T treatment led to remission without relapse in xenograft models.cell

Expanding Beyond Blood Cancers

While CAR T cell therapy has transformed treatment for several blood cancers since its development at Penn, current FDA-approved versions target antigens widely expressed only in blood malignancies. Extending the therapy to solid tumors has been hampered by the difficulty of finding appropriate targets — a process the researchers say their framework can compress from months of manual work to weeks.x+2

The system was built to allow researchers to apply it across disease types without redesigning the underlying architecture, potentially accelerating CAR T development for cancers and non-cancerous conditions alike.miragenews+1