From BioLector to Ambr250: Using models to predict behavior across scales

Strain and media selection have a direct and significant impact on the cost of goods in large-scale microbial fermentation processes. To accelerate development and reduce cost, high-throughput screenings are typically performed at microliter scale, with only a subset of top-performing candidates advanced to larger-scale bioreactors. However, performance often does not translate reliably across scales, leading to suboptimal performance.

Join Dr. Moritz von Stosch and Dr. Jing Geng as they present a collaborative study between DataHow and Procelys applying hybrid modeling and transfer learning to address this challenge. The project demonstrates how scale dependencies can be learned from historical data and how large-scale performance of new strains or yeast extracts can be predicted from small-scale screening data, enabling more informed and efficient development decisions.

This insightful webinar will include:

• Introduction to the technologies and methods used
• How transfer learning reduced experimental needs by 38% at BioLector scale and 50% at ambr®250 scale. ´
• Project results & operational implications for Microbial development