Using Hybrid Modeling to Generate In-Silico Bioreactor Simulations for Faster Upstream Process Development

Combining in-vitro experiments with in-silico simulations conducted in a virtual bioreactor can help speed upstream process development for any cell-based biopharmaceutical product. Cell Insights by Umetrics^® Studio is an advanced data analytics application that uses hybrid models to generate bioreactor simulations for a variety of cell-culture development uses.

^{This article is posted on our Science Snippets Blog}

Developing new cell-based therapies relies on cultivating living cells using biological processes that are extremely sensitive and can be impacted by countless variables. Every bit of variation in the environment can affect the cell’s growth and can even cause unexpected changes in productivity or quality.  Gaining early awareness of potential obstacles to efficient process development and being able to predict process outcomes in-silico is critical for making informed decisions.

Current upstream process development methods require teams to conduct numerous experiments, which is both time-consuming and costly. An alternative approach is to create a virtual bioreactor that uses simulations to predict different experimental outcomes. Advanced data analytics methods like in-silico simulation can not only help create a more efficient and error-resistant process, but also reduce the time it takes to develop a robust process.

Bioreactor Simulations Incorporate Hybrid Models

The latest developments in virtual bioreactor simulations¹ rely on a hybrid modeling approach that combines historical process data with mechanistic knowledge about cell metabolism and growth. These types of in-silico simulations can be used to explore “what-if” scenarios for different upstream process development uses.

The term in-silico stems from silicon chips, which were used for computing at the time when the term was coined. In-silico is often translated to "in the computer". In this way, in-silico methods refer to methods or predictions that use computational or statistical approaches, rather than analyzing live data. In-silico simulations allow scientists to explore different designs without performing each individual experiment, which results in a quicker and more affordable process development.

For example, factors like temperature and pH levels can significantly impact cell growth within a reactor. Creating a mathematical model of the cell biology and combining that with experimental data provides a way to quantify the effect that certain Critical Process Parameters (CPPs) have on the outcome of an experiment.

These types of strategies are found within the Quality by Design (QbD) framework  and can only be carried out if Process Analytical Technology (PAT) tools are in use. It ensures that both safety and therapeutic efficacy are optimized when looking at the final product.

Virtual Bioreactor as Key Enabler of Biopharma 4.0

In-silico simulations are a significant improvement over current methods of process development and represent an important step towards Biopharma 4.0 . By incorporating data analytics and machine learning into bioreactor simulations, you are able to increase the quality and yield of bioprocesses and reduce development timeframes.

Bioreactor simulations enhance decision-making by coupling in-vitro cell culture experiments with machine learning and mechanistic knowledge of cell behavior and metabolic pathways.

Hybrid Modeling Creates Robust Processes

Hybrid modeling combines the advantages of physical modeling, which relies on correlation patterns, and computational modeling, which assumes the “ideal scenario”, to create a virtual bioreactor that mimics the real-world environment.

Hybrid models combine first-principal knowledge of cell dynamics and metabolism with data-driven information to generate statistically accurate predictions about how cells perform under various circumstances or events. Both the mechanistic models and in-vitro data are combined into one workflow for more accurate predictions.

Numerous studies  show a growing interest in the application of digital twins  that rely on the use of hybrid models (in-silico combined with in-vitro data) in biopharma as part of a move toward Biopharma 4.0².

“Recently, there is a growing interest in industrial applications of the digital twins (DT) which integrate physical and virtual systems via real-time data monitoring, thus enabling their interactive communications for the enhanced operational efficiency towards advanced biomanufacturing.”³

However, one of the challenges in employing model-based methods is ensuring the accuracy of the model. In-silico models can function as digital twins, which are very helpful in the design of experiments or process monitoring. Using reliable models that have been validated, such as Cell Insights by Umetrics^® Studio , saves you time.

Digital Twin vs Simulation

What’s the difference between a digital twin  and a bioreactor simulation?

A digital twin is a virtual environment that serves as the real-time digital counterpart of a physical bioreactor process. Within a digital twin all types of simulations can be run and are typically represented in 2D or 3D.

A bioreactor simulation, or virtual bioreactor, is an offline model-based representation of a particular bioreactor operation or process event. Simulations can represent multiple or single events that occur during a bioreactor process and are typically used for design and optimization projects.

Typically, you will use a digital twin in real-time to proactively control and maintain the process, whereas a bioreactor simulation is more likely to be used for investigation. So, a digital twin is used in manufacturing, while a simulation is more often used in process development.

Advanced Data Analytics Applications Use Hybrid Approach

Cell Insights by Umetrics^® Studio  is an advanced data analytics application that uses hybrid models to generate bioreactor simulations. These simulations can actually be generated using process data from just one run.
Cell Insights supports a variety of cell-culture development use cases, including:

1. Perfusion cell line selection.  A simulation which demonstrates the impact of different cell culture methods, whether it's fed batch perfusion or N-1 on cell growth and productivity.
2. Conditions configuration. Perform “what-if” investigations on how changes to process parameters (i.e., temperature shifts, pH shifts, etc.) will affect cell growth, metabolism, and productivity.
3. Perfusion protocol configuration. In-silico simulation which provides insights about how media composition and exchange rates impact cell growth so you can design optimal protocols.
4. Seed train configuration. Experiment with different seed durations, number of passages, and culture methods to predict how cell growth will be impacted.

In-Silico Simulations. Actionable Insights.

Cell Insights by Umetrics^® Studio, an advanced data analytics application, expands on in-vitro experiments with in-silico simulations in a virtual bioreactor, accelerating timelines and decreasing costs.

Find Out More

_{References:
¹ Tsopanoglou, A. and  Jiménez del Val, I. (2021), Moving towards an era of hybrid modelling: advantages and challenges of coupling mechanistic and data-driven models for upstream pharmaceutical bioprocesses. Current Opinion in Chemical Engineering, 32:100691.
² Sokolov, M. et al., (2021), Hybrid modeling — a key enabler towards realizing digital twins in biopharma? Current Opinion in Chemical Engineering, 34:100715.
³ Park, S. et al., (2021), Bioprocess digital twins of mammalian cell culture for advanced biomanufacturing. Current Opinion in Chemical Engineering, 33: 100702.}