Pfizer scales intensified perfusion process from 100 L to 500 L
Pfizer reports scaling its HILVOP dynamic perfusion process to 500 L, with TFF membrane fouling as a key scale-up risk.
Upstream
Preprint: kidney organoids generated on microcarriers in stirred bioreactors
Cornell, Harvard and Utrecht researchers grow kidney ureteric bud and collecting duct organoids on ECM microcarriers in stirred bioreactors, per a new preprint.
Preprint: fault batches in soft-sensor training cut simulated penicillin deviation RMSE 19.7%
A bioRxiv preprint on simulated penicillin batches shows adding deviation batches to soft-sensor training cuts deviation-batch RMSE 19.7% without hurting normal-run accuracy.
AGC Biologics to Scale Up Flen Health’s Recombinant Wound-Care Enzymes
Flen Health taps AGC Biologics’ Heidelberg site to make recombinant lactoperoxidase and glucose oxidase to GMP, pharma-grade standards.
Perfusion, explained: three flows, one balance, and where the product goes
We read 38 primary sources on CHO perfusion. Two flows deserve most of the attention on cost: the medium you buy and the product the bleed throws away.
From 2 L to 2,000 L: what actually breaks when you scale a CHO process
We read 38 primary sources on CHO scale-up. The trouble at 2,000 L is on the gas side and at harvest, not in impeller eddies.
AGC Biologics commits half of Yokohama’s mammalian capacity before the site opens
An undisclosed biopharma has booked half of the mammalian capacity at AGC Biologics’ Yokohama site: five products, 35+ batches a year, four PPQ campaigns in two years.
Predicting scale-up with small data: hybrid models plus transfer learning
A Manchester and Xiamen study adapts a lab-built hybrid model with pilot data only and predicts industrial yeast fermentation at 23.2% MAPE, per the abstract.
Why Raman models break across scales, and a preprocessing fix from mini-bioreactor data
Five of eight Raman models calibrated on Ambr 250 data lost accuracy at 5 L. Choosing preprocessing per target cut cross-scale error by 14.0 to 56.1% for those five.
Model-driven monitoring of an E. coli Fab fed-batch across scales
TU Berlin and Wacker fit one kinetic model from 30 L to 15 mL: batch growth aligns, pulse feeding does not, and mL-scale adaptations, chiefly IPTG, doubled specific yield (abstract).