Daniel Baker

AI-driven discovery of GPNMB CAR T cells as a pan-cancer therapy

Daniel J. Baker1,2,3*, Leon Frommer4, Ugur Uslu1,2, Nils W. Engel1,2, Daniel Zhu1,2, Samuel I. Kim1,2, Lisa Sun1,2, Kisha K. Patel1,2, James M. George5, Christopher Roselle1,2, Philipp C. Rommel1,2, Regina M. Young1,2, Jonathan A. Epstein3,6,7,8, Sikander Hayat4*, Zoltan Arany3,6*, Carl H. June1,2,9* Affiliations: 1Center for Cellular Immunotherapies, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104, USA 2Department of Pathology and Laboratory Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104, USA 3Cardiovascular Institute, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, PA 19104, USA. 4Department of Medicine 2, RWTH Aachen University, Medical Faculty, Aachen, Germany 5Michigan Center for Translational Pathology, University of Michigan, Ann Arbor, MI, USA. 6Department of Medicine, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, USA. 7Institute for Regenerative Medicine, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, USA. 8Department of Cell and Developmental Biology, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA, USA 9Parker Institute for Cancer Immunotherapy at University of Pennsylvania, Philadelphia, PA 19104, USA

CAR T cells have demonstrated curative potential in hematologic cancers and increasing efficacy in solid tumors and non-malignant diseases. However, target identification remains a major bottleneck. We developed a systematic, AI-driven approach for rational CAR T target discovery by integrating single-cell RNA sequencing datasets from human melanoma and healthy tissues. This list was refined using public datasets to optimize tumor expression, tissue specificity, and clinical feasibility. Large language models were applied to prioritize and nominate targets with broad therapeutic promise. Glycoprotein non-metastatic melanoma protein B (GPNMB) emerged as the top hit, exhibiting expression across hematologic and solid tumors. We engineered a novel human CAR T cell targeting GPNMB, which showed potent antitumor activity in mouse models of histiocytic lymphoma, melanoma, and colorectal adenocarcinoma. These findings establish a scalable, human-in-the-loop pipeline for CAR T target discovery and support the clinical translation of GPNMB-directed CAR T cells as a pan-cancer therapeutic.