Computational platform for antibody glycosylation prediction

Therapeutic glycoproteins represent 40% of biologics currently approved by the European Medicines Agency. THese drugs are predominantly produced in mammalian cell hosts, such as Chinese hamster overy (CHO) cells, which are capable of delivering high quality recombinant proteins with human-like post-translational modifications. One such modification is N-linked glycosylation, which involves the covalent addition of oligosaccharides to the protein backbone. It is well documented that these glycan structures can affect protein stability as well as efficacy and are therefore monitored during R&D and production. However, they have not traditionally been among the key performance indicators angainst which cell lines are creened during the early stages of cell line development campaigns. This is because of a variety of reasons ranging from the low culture volume at these stages, which does not allow extensive analytical characterisation of cells or products, to the cost of performing such analyses for a large number of samples.

Recently available instrumentation such as the Ambr system together with tools such as Raman spectroscopy are now enabling parallel experimentation with online or at line monitoring of cell population, nutrients and metabolites. This setup provides a more controlled environment than previous shakeflask-based protocols and allows GSK to build more substantial and reliable databases to better understand these processes. GSK also has at its disposal a wealth of computational resources for data analysis as well as for mechanistically describing cellular behaviour. One such tool is a recient Imperial generated mathematical model that relates extracellular conditions in terms of cell population dynamics and nutrient availability to intracellular metabolism and eventually protein glycosylation in and IgG-producing hybridoma cell line (Jedrzejewski et al, 2014, Int. J. Mol. Sci. 15(3):4492-4522). GSK can therefore envision a platformin which it takes an increasingly large number of bioprocess parameters at low scale and feeds these into computational platforms that generate predictions for difficult-to-measure but critical product properties.

The aim of this project is to develop and validate the computational aspects of this Imperial platform to support the development and selection of GSK antibody-producing CHO cell lines. The specific project objectives are:
- To use the model created by Jedrzejewski et al. to design informative experiments for the adaptation of this approach to current GSK CHO cell line development protocols;
- To validate the tool for a wide range of experimental conditions that are commonly encountered in production reactors and a range of cell lines;
- To create a user-friendly interface for use by cell culture scientists.

It is envisaged that this tool will find application in cell line development for informing the screening of cell lines by generating predictions for the product glycoform distribution base on macroscopic measurements. It will further be useful in process development and optimisation, where it can be used to explore the design space in silico and identify a subset of operating parameter ranges that lead to the desired glycan structures. The integration of mathematical modelling to current practice has high impact potential in this industry and, in the long term, can support the implementation of the Quality by Design initiative.

For the avoidance of doubt, all IP specifically relating to compounds, biological materials, including cell lines, and associated data provided by GSK for the purpose of this project fall within the definition of GSK Background IP.