Interview Update: Astellas Pharma's Epistra AI x Robot Case Study -- Behind the 50-100x Yield Improvement

We have published a case study on our website about how Astellas Pharma tackled the challenge of reproducibility in cell culture through the integration of AI and robotics. The article is presented as an interview with Mr. Atsushi Inoue (Astellas Pharma), who led this initiative.

For Those Working on Cell Therapy Research and Development

Do you face these challenges?

• High variability in long-term cell culture, resulting in low experimental reproducibility
• Unable to find culture conditions that meet targets through experience-based or literature-based condition searches
• Want to efficiently advance cell manufacturing process development, but lack concrete strategies

These challenges have long been barriers to the practical application of cell therapies. Especially in long-term cultures requiring differentiation induction, the enormous time and cost involved in condition searches, along with cell culture-specific issues such as “unable to test many conditions” and “results not meeting expectations,” have plagued researchers.

The Key to the Solution Was AI x Robotics

A three-way collaboration between Astellas Pharma, Robotic Biology Institute (RBI), and Epistra applied a new approach to “NK cell differentiation process optimization,” addressing the aforementioned challenges and presenting a novel methodology toward next-generation cell therapy manufacturing.

Result 1: “Culture Without Variability”

Robotic experiment automation achieved “culture without variability” (CV 5.9%)

Result 2: “Dramatic Yield Improvement”

High-precision data from robotic experiments served as the foundation for condition optimization using Epistra’s experimental condition optimization AI, Epistra Accelerate. In just 3 months, numerous conditions achieving 50 to 100 times the yield compared to literature reports were discovered.

Result 3: “Design Space Prediction”

Condition space modeling by Epistra Accelerate successfully predicted the effective condition range (design space). This enabled QbD (Quality by Design)-conscious evaluation from the early stages of process development.

The interview features Mr. Atsushi Inoue of Astellas Pharma Inc., who led the project, discussing the background and results of this initiative. If you are interested in process development utilizing AI and robotics, we encourage you to read the full article.