Deepcell and Gilead Sciences announced a collaboration to co-develop a custom AI foundation model for Chinese hamster ovary (CHO) cell line development using single-cell morphology data. The project will use Deepcell’s REM-I high-dimensional imaging and sorting platform to train a self-supervised model without manual labeling. Deepcell said the model is intended to provide earlier readouts on CHO line quality by characterizing morphological structure linked to traits like titer, stability, aggregation risk, and metabolic profiles—factors that often drive downstream bioprocess outcomes. The companies expect to develop the initial model over the coming months and evaluate it on held-out Gilead cell lines. Deepcell also indicated plans for an open-weight version of the model, while declining to disclose IP ownership and financial terms.