Deepcell Announces Collaboration with Gilead Sciences to Co-Develop an AI Foundation Model for CHO Cell Line Development
The companies will build a custom Chinese hamster ovary (CHO) cell morphology foundation model on Deepcell’s REM-I platform to help identify high-quality antibody producer cell lines.
MENLO PARK, Calif., August 20, 2026. Deepcell, a pioneer in AI-powered, label-free single-cell morphology analysis company, today announced a collaboration with Gilead Sciences, Inc. (Nasdaq: GILD) to co-develop a custom foundation model for Chinese hamster ovary (CHO) cell morphology. The model will be built on Deepcell’s REM-I imaging platform and trained to characterize the CHO cell lines used in the development and manufacture of biotherapeutics.
The collaboration is designed to give cell line development teams an earlier, more efficient read on cell line quality. By learning directly from single-cell morphology, the model aims to help scientists identify cell lines with desirable attributes such as titer, stability, aggregation, and metabolic profile weeks earlier than conventional workflows and with significantly greater efficiency. Because Deepcell’s approach is free of detection reagents, it applies across host lines and molecule formats.
“Beyond scoring cell line quality, the REM-I platform gives teams a window into transfection efficiency, timing, and culture and media conditions, and those same insights may extend to monitoring cells in a bioreactor,” said Mahyar Salek, President and Chief Technology Officer, Deepcell. “We built REM-I to read biology that humans can’t, and a CHO foundation model puts that capability directly into the hands of process development scientists.”
Under the collaboration, Deepcell will image Gilead’s CHO cell lines on REM-I and train the model using self-supervised learning, so that it learns morphological structure without manual labeling. The parties intend to publish an open-weight version of the model, making it available to the broader research and bioprocessing community.
The initial model is expected to be developed over the coming months, with evaluation on held-out Gilead cell lines.
About Deepcell
Deepcell is a life sciences company pioneering AI-powered, label-free single-cell morphology analysis and sorting. Its REM-I platform captures, analyzes, and isolates cells based on high-dimensional morphological features, and its foundation-model approach surfaces biological distinctions that support research across oncology, cell therapy, immunology, and bioprocessing. Deepcell was spun out of Stanford University in 2017 and is based in Menlo Park, California. Learn more at www.deepcell.com.
Deepcell: info@deepcellbio.com
This press release contains forward-looking statements, including statements regarding the expected benefits, timing, and outcomes of the collaboration between Deepcell and Gilead Sciences described above. These statements are based on current expectations and assumptions and are subject to risks and uncertainties that could cause actual results to differ materially, including risks related to the development, performance, and availability of the model described herein. Deepcell undertakes no obligation to update these forward-looking statements, except as required by law.