Insights from the ESACT Bioprocess Modeling Course 2023

Authored by Mike Khor

In October, I had the privilege to represent Culture Biosciences at the first-ever "Metabolic Bioprocess Modeling" course organized by ESACT (European Society for Animal Cell Technology) in the picturesque Llafranc, Spain. In my work at Culture, I have built mechanistic and hybrid bioprocess models, which is why this was a great opportunity to interact with experts on the topics, while learning about other tangential modeling realms.

ESACT and the Course

ESACT was founded in 1976 to serve as a platform for sharing ideas on biological and engineering techniques. Their goal? To promote the knowledge and use of human and animal cells for manufacturing products. Every year, ESACT curates short, intensive courses on topics critical to professionals in the animal cell culture, and this year’s new introduction of “Metabolic Bioprocess Modeling” course is a sign of growing relevance of modeling applications in bioprocessing.

Over five days, professors and industry experts from around the world presented lectures and workshops covering a spectrum of relevant topics:

• Flux Balance Analysis (FBA) and genome-scale metabolic modeling
• Kinetic modeling
• Hybrid modeling
• Computational fluid dynamics (CFD)
• Chemometrics in their support of Process Analytical Technologies (PAT)
• Multiomics analysis

Group photo of course coordinators, lecturers, and participants, by the beach of Llafranc.

Key Takeaways

Even as a practitioner of modeling as a data scientist at Culture Biosciences, I was reminded in this course on how much breadth there is in bioprocess modeling, beyond what I know today. Here are some insights I’ve learned at the course:

• Most of metabolic modeling today comes in some form of Flux Balance Analysis (FBA). FBA is derived from metabolite balancing, based on a set of known cellular reaction pathways and measured metabolite fluxes. FBAs are commonly used to optimize an objective (e.g. product flux), and can act as a guide for gene knockouts, nutrient restrictions, or the identification of nutrient bottlenecks. For a holistic view of reaction pathways, a modeler can leverage genome-scale metabolic models (GEMs) developed by the modeling community in academia (such as CHO-GEMs developed since 2016), and modify pathways based on internal genetic engineering efforts to conduct FBA that is accurate to their cell line.
• The course presented a systematic method to develop mechanistic kinetic models, using observed correlations between metabolites and reaction rates to guide model equation construction. The presenters dove deep into decisions surrounding parameters and the significance of global sensitivity analysis. This mirrors Culture's own work on the topic in CHO cell lines. The course also covered topics in hybrid modeling, particularly the effectiveness of data-driven expressions (Gaussian processes, partial least squares, neural networks) in predicting specific reaction rates. The presenter also introduced the prospect of an “automated model assembly pipeline,” which I think is the holy grail of data-driven kinetic modeling.
• The adoption of NIR and Raman spectroscopy requires robust chemometrics processing techniques. The data preprocessing and the configuration of partial least squares (PLS) models may vary based on the process, but the procedure is becoming increasingly standardized. While chemometrics technologies have shown difficulties in generalizing across different cell culture conditions (making it less useful in frequently changing conditions such as process development), researchers today are aiming to develop generic algorithms such as generic Raman that can be generalized to varying environments.
• Other advanced topics (that are frankly much further outside of my domain of expertise) included multiomics and computational fluid dynamics (CFD). A common multiomics application is differential expression, which leverages RNA sequencing to spot differences in gene activity across different conditions, providing insights into the metabolic processes at play when cells are exposed to different environments. CFD remains pivotal for deciphering hard-to-measure variables in reactors. With Ansys Fluent software, the course demonstrated the impact of turbulent eddies on cell death.
• The bioprocess modeling community has an ultimate vision of combining multiple modeling techniques together into giant integrated models (sometimes called “digital twin” but I think that term is quite ambiguous). For example, integrating FBA with hybrid kinetic models, or using multiomics data to refine genome scale metabolic models, or using CFD-CRD (computational reaction dynamics) which combines CFD and metabolic models to assess an organism’s response to heterogenous environments. This increasing interest can be seen in recent publications (here’s a good review article) in the form of collaborations between academia and industry experts of different domains.

FBA workshop material showing results on flux optimization in a simplified CHO metabolic model. Optimization results by COBRA are visualized in Escher. Material credit: Dong-Hyuk Choi and Yi Qing Lee from Sungkyunkwan University.

Advance Your Bioprocess Modeling with Culture

At Culture Biosciences, we're harnessing the power of our 250-mL and 5-L cloud-based bioreactors to drive forward bioprocess modeling. Our team is deeply engaged in mechanistic kinetic and hybrid models, and we're actively integrating these approaches for our clients.

We're exploring new territories in Flux Balance Analysis (FBA), Genome-Scale Metabolic Models (GEMs), multiomics, and Computational Fluid Dynamics (CFD). Recognizing that there's always more to learn, we're open to collaborations with academic and industry partners who specialize in these areas. Our strength lies in generating high-quality bioprocess data and providing software engineering expertise.

If you're interested in these fields or looking for a collaborative partner in bioprocess modeling, consider joining forces with Culture Biosciences and reach out to us. Let's work together to push the boundaries of bioprocess technology.