Acceleration of Bioprocess Development Using DoE and Physiological Control
with Christoph Herwig
Understanding the interdependencies among process variables, product quality attributes, and process parameters is key to achieving quality by design (QbD). Developers need to gather as much meaningful information as possible from their experiments. In a BPI “Ask the Expert” webinar on 7 August 2016, Professor Christoph Herwig of the Vienna University of Technology demonstrated how experiments based on physiological key parameters can be linked to design of experiment (DoE) approaches to accelerate bioprocess development. Herwig recommended an all-in-one tool for such work: the Lucullus process information and management system from Applikon.
Herwig’s Presentation As a unique tool for bioprocess development, piloting, and manufacturing, Lucullus software has four main features for planning, preparation, execution, and evaluation of experiments. Herwig focused on the latter. “These elements can be done not only on historical data, but also fed back to process control and run in real time.”
One basic Lucullus tool is the Metabolic Balance Analyzer, which is based on first-principle balances (elemental and mass balances). It is a generic “soft sensor” that estimates unmeasured components and requires no strain-specific information. User inputs are variables such as off-gas analysis, feed flow rates, bioreactor volume, feed concentration, and so on.
Herwig showed an example for robust estimation of biomass in induced conditions. “When you calculate biomass over time, then you can calculate from there very important entities for bioprocess development: scalable entities like specific growth rates or specific substrate uptake rates.” Those are important to developing a scalable process without surprises in scale-up.
Once those entities are calculated, they help users understand the relation between their cells and product quality attributes, as well as productivity and titer. Herwig offered examples here as well, including a DoE for determining the maximum specific growth rate depending on temperature, with results available overnight. He also demonstrated a physiological approach to determining feed profiles for optimizing productivity. Dynamic experiments simplify the process, further streamlining the time savings of multivariate analysis.
Herwig showed how the balancing effect can improve data quality in real time. Hard sensors create “noisy data,” which can be adjusted for by the balancing tool. Herwig said the tool also applies to kinetic models. “That helps you predict process events: When is the batch finishing? When is the substrate limiting? When did I reach the maximum productivity to trigger an action?” Tools in the Lucullus system can be used for acceleration and optimization of such processes.
The professor explained how dynamics, real-time approaches, and “soft sensors” can cut process development times, ensure scalability, increase productivity, and create synergies with future products and processes by helping companies develop platform knowledge. In an example, he showed how applying these methods to increasing titers, clone selection, and scale-up/scale-down could cut down development time here by 50%.
Questions and Answers
You mainly showed data on microbial processes. Can this be equivalently transferred to mammalian processes? Yes. Mammalian processes are more complex and involve different measurements. But we have implemented elements like that. We have tools to tailor, for example, the sampling plan and the number of measurements you need. You may need some kinetic models in the background, but you can apply the same thing to mammalian processes.
Are these tools available in real-time or only in offline mode? You can do both. With IPC data exchange, you can include offline data for a more detailed analysis of your process. However, here we emphasize real-time implementation and execution. Both are possible with the Lucullus system.
How much effort does it take to adapt the tool for specific tasks? Normally we are analyzing the process environment. So we first look at what kind of data sources are available in your standard environment. Then we assess whether they are enough for proper balancing and estimation. We are not emphasizing just measuring more to have more data. Rather, we are mainly considering what kind of data you have already and then analyzing in a second step which kind of additional data actually provides additional knowledge that you can apply to accelerating bioprocess development.
The full presentation of this “Ask the Expert” webcast can be found on the BioProcess International website at www.bioprocessintl.com/webinars/acceleration-of-bioprocess-development-using-DoE-and-physiological-control. And you can visit www.applikon-bio.com for more information.