Bruce Levine, Barbara and Edward Netter Professor in Cancer Gene Therapy at the University of Pennsylvania, shared a post on X:
“NEW – AI-driven discovery of GPNMB CAR T cells as a multi-cancer therapy.
AI did not ‘invent’ a CAR T therapy here. It helped focus the search for a plausible target in a very large haystack of public single-cell and expression data. That distinction matters.
The important advance is human expertise plus AI triage, followed by wet lab validation. GPNMB CAR T activity across melanoma, leukemia, and colorectal models is intriguing. The hard human enabled part is demonstrating specificity, safety, clinical translation.
Target discovery has long been one of the bottlenecks for extending CAR T beyond B cell malignancies into solid tumors.
This paper is a useful example of where AI may actually help: not replacing biology, but organizing vast datasets, nominating candidates, and accelerating the path to experiments.
GPNMB is not a clinical answer yet, but it is a serious hypothesis now supported by preclinical data across multiple tumor models.
The lesson from this GPNMB CAR T study is not ‘AI cures cancer.’ It is that public datasets, thoughtful prompts, repeated model runs, expert review, and biological validation can compress the target discovery cycle.
That is meaningful for cell therapy. The caution is equally important: antigen expression does not equal therapeutic window. The field still has to prove tumor selectivity, persistence, trafficking, toxicity, and benefit in patients. In the meantime, mouse model data makes this target look very promising.
”
Title: AI-driven discovery of GPNMB CAR T cells as a multi-cancer therapy
Authors: Daniel J. Baker, Leon M. Frommer, Ugur Uslu, Kisha K. Patel, Daniel Zhu, Nils W. Engel, James M. George, Wencao Zhao, Samuel I. Kim, Lisa Sun, Christopher Roselle, Philipp C. Rommel, Regina M. Young, Jonathan A. Epstein, Sikander Hayat, Zoltan Arany, Carl H. June
Other articles featuring Bruce Levine on OncoDaily.