Chinese hamster ovary (CHO) cells are vital to modern medicine with an estimated 70% of all approved biologics—and nearly all monoclonal antibodies—being produced using CHO-based expression systems. Despite this near ubiquity, until recently industry only had limited tools available to model culture performance.
But things are changing, says Ioscani Jiménez del Val, PhD, assistant professor at University College Dublin’s School of Chemical & Bioprocess Engineering, who believes advances in analytics will increase the use of transferable process models.
“Before the current push toward biopharma 4.0, models describing cell culture bioprocesses were used relatively sparingly in industry and, when used, were developed for specific processes products. Traditionally, upstream bioprocess models were used to interrogate and, in ideal cases, optimize the performance of specific cell culture processes.
“Although such work laid strong foundations, model transferability to other products processes was perceived as being onerous and requiring technical expertise and bespoke software. These perceptions, perhaps, stood in the way of broader adoption and, therefore, in limited integration of models across the full pharmaceutical bioprocess.”
He adds that, “The aims of Biopharma 4.0—monitoring, automated decision making, and advanced control of integrated manufacturing bioprocesses—provide ample incentives and opportunities to establish novel modeling strategies.”
Modeling automation
Jiménez del Val and colleagues outlined just such a strategy in a paper this year, presenting an algorithm designed to process CHO cell growth, nutrient consumption, and protein production data, quantify correlations, and automatically produce predictive models.
“Our intention with the proposed algorithm is that users simply input cell culture data and retrieve a fully functioning dynamic hybrid model that describes the bioprocess.
“From the work’s inception, its overarching objective was to automate all aspects of model development and, thereby, incentivize the uptake of modeling platforms across the biopharmaceutical sector,” he says, adding, “The core advantages of our proposed strategy are its data processing strategy and, of course, its automated model assembly.”
And the potential benefits of automating process development are significant, according to Jiménez del Val.
“I believe that establishing automated model assembly strategies—where the user inputs bioprocess data to obtain a fully defined hybrid model that describes the dynamics of cell culture processes—will increase the uptake of such strategies across the sector, incentivize the development of novel data processing and model assembly strategies and, thereby, advance the realization of Biopharma 4.0.”
The team used the Matlab language to create the algorithm which, Jiménez del Val says, facilitates interaction between different systems and is a good fit for the biopharmaceutical manufacturing environment where multiple technologies are combined.
“For the sake of computational efficiency, our Matlab code outputs the hybrid model in gPROMS ModelBuilder v7.0 for simulations.
“The code can be adapted to output the model in any other language—open source or otherwise—that is capable of solving systems of differential algebraic equations. We can facilitate our code with a transfer agreement in place and the company can simply input experimental data and run the resulting hybrid model.”
