← Back to Home

Selecting Donors Before Scaling Up: ImmuneBridge Uses AI to Seek a Path to Mass-Producing Cell Therapies

Whether cells can become more widely accessible medicines also depends on where manufacturing begins. ImmuneBridge is applying machine learning to donor screening alongside stem cell expansion technology, hoping to reduce variability and increase output; costs and clinical benefits remain to be validated.

By SURL BioNews

The challenges of turning living cells into medicines often begin with selecting the starting material. Cells from different donors may differ in growth and function; finding a source that performs suitably and can be expanded sufficiently is a necessary hurdle for consistent mass production. San Francisco biotech company ImmuneBridge is trying to bring artificial intelligence into this step to give donor selection a stronger basis.

CBS News reported on this approach in October. According to ImmuneBridge’s public description of its technology, its machine learning models use screening data covering hundreds of donors and manufacturing processes. The aim is to use information obtained during hematopoietic stem cell expansion to predict how candidate cells will perform later, eliminate weaker combinations early, and reduce the resources devoted to subsequent differentiation and testing.

AI’s role here is to help prioritize experiments. Whether the cells ultimately have the required functions still needs to be tested. In a sponsored interview published by News-Medical in May 2025, Chief Technology Officer Rui Tostoes used natural killer cells (NK cells) as an example, explaining that the team conducts small-scale cultures in 10-milliliter bioreactors, followed by in vitro and animal testing, to select the top 1% to 10% of candidates that perform best for a particular indication.

After screening, another question is how to retain enough material. The design Tostoes described uses cord blood material from selected donors throughout subsequent clinical and commercial manufacturing, with small bioreactors corresponding to larger-scale processes. This approach aims to reduce the need to change cell sources or substantially modify culture methods during development, providing greater continuity between early experiments and mass production.

Supporting expansion is the company’s proprietary molecule IBR403. ImmuneBridge says it promotes hematopoietic stem cell proliferation while preserving their ability to differentiate into multiple types of immune cells. Combined with its bioreactor process, the company claims that a single cord blood unit can yield tens of thousands of doses of immune cells. This is a manufacturing capacity disclosed by the company; it cannot be directly translated into the number of patients who have completed treatment, nor does it mean that all these cell products have been approved for use.

The main limitation of the currently available public information is the lack of validation details. Although the company says its models can accurately predict candidate performance, the sources above do not provide model accuracy figures that can be assessed, results from independent test sets, or quantitative comparisons with existing donor screening methods. The sponsored interview adds an explanation of the manufacturing process, but it remains the company’s own technical account and is not yet sufficient to demonstrate that AI-selected cells can deliver better outcomes for patients.

Moving from mass production to affordable treatment still presents two challenges: cost and quality. These materials do not disclose the full cost per dose, the prices patients actually pay, or comparative results demonstrating price reductions. For this platform to fulfill its goal of making treatment widely accessible, what is needed next is reproducible batch quality, functional data after manufacturing scale-up, and evidence of clinical safety and efficacy. Selecting candidates faster is only the starting point on that path.

References

  1. CBS News
  2. ImmuneBridge
  3. News-Medical