AI Framework Helps Identify New CAR T Target

A human-in-the-loop AI strategy helped narrow more than 10,000 candidates to a CAR T target now validated in mouse models

July 6, 2026
3 min

The Medicine Maker

Overview

Researchers at Penn Medicine have developed an AI-assisted framework that identifies new targets for CAR T-cell therapy, successfully nominating GPNMB as a potential target across multiple cancers. This approach integrates public datasets and large language models to streamline the target discovery process.

Background

The expansion of CAR T-cell therapy beyond hematologic malignancies faces significant challenges, particularly in identifying suitable surface antigens that are expressed on tumors while minimizing damage to healthy tissues.

Data Highlights

The study utilized a human-in-the-loop strategy, integrating public single-cell RNA sequencing datasets and large language models to filter over 10,000 potential targets for CAR T development.

Key Findings

The AI framework successfully nominated GPNMB as a promising target for CAR T-cell therapy.
GPNMB was confirmed to have surface expression across multiple tumor types.
Engineered GPNMB-directed CAR T cells demonstrated antitumor activity in mouse models of melanoma, monoblastic leukemia, and colorectal adenocarcinoma.
The framework is modular and disease-agnostic, applicable to other cancers as datasets improve.
This study marks one of the first applications of large language models in cell and gene therapy.

Clinical Implications

Further validation of GPNMB's specificity and safety will be essential for clinical translation.

Conclusion

The AI-assisted framework developed by Penn Medicine represents a new approach in the search for CAR T-cell therapy targets, with GPNMB emerging as a candidate for further investigation.