Machine learning for enhanced understanding of cell culture bioprocess development

A central part of the biopharmaceutical manufacturing process are bioreactors, i.e. sterile containers in which genetically modified cell cultures are grown in a controlled environment to produce drugs or intermediate compounds. One of the major – and mostly unsolved –challenges is to understand how the vast array of bioreactor settings influence amount and quality of the drugs produced.

Biopharmaceutical drug manufacturing is a very common but also expensive process, hence small improvements in efficiency have large impacts on manufacturers’ operations and the availability of public healthcare. One major reason for currently high costs of biopharmaceutical drugs is that the manufacturing process based on bioreactors is highly complex and difficult to understand using the more classical engineering tool set.

AstraZeneca is driving efforts to make full use of the available bioreactor sensor dataa to enhance process control in a smart and automated way. This report presents the results from the Data Study Group, a week-long collaboration between AstraZeneca and The Alan Turing Institute, with the main goal to use modern data science and machine learning techniques in order to find out if one can accurately predict the amount of drugs produced and discover key controllable variables.

Citation information

Data Study Group team. (2019, August 13). Data Study Group Final Report: AstraZeneca. Zenodo. http://doi.org/10.5281/zenodo.3367412

Additional information

Tracy Ballinger, University of Edinburgh
Magda Bucholc, Ulster University
Jingjing Cui, Queen Mary University of London
Alex Gao, University of Toronto
Tobias Hoejgaard Dovmark, Oxford University
Sangyu Lee, University of Leeds
Matthew Levine, Caltech
Markus Loning, UCL
Daniela Perry, University of Munich
Emma Vestesson, The Health Foundation

Turing affiliated authors