OR WAIT 15 SECS
Feliza Mirasol is the science editor for BioPharm International.
Process modeling offers an opportunity to troubleshoot for and anticipate difficult aspects of a bioprocess.
Choosing a bioprocess model should be done early in the process design/development stage for a new biotherapeutic as it provides an opportunity to troubleshoot difficult aspects of a bioprocess and determine the conditions under which process monitoring tools should be used and process controls applied.
Artur Arsenio, head of Product Management PAT & Automation, Sartorius, and Doug Miller, managing director, BDO Bioprocess Technology Group (BPTG), shared insights into bioprocess modeling and how to avoid potential pitfalls with BioPharm International.
BioPharm: What criteria should be considered when choosing a bioprocess model?
Arsenio (Sartorius): Manufacturers should consider, first of all, how precise the model is. Can it accurately predict the variable of interest? What is the sensitivity-to-noise ratio? Another important consideration is robustness. When transferring the process from process development (PD) to commercial manufacturing, can the model be scaled to larger, potentially different, bioreactors? Finally, they should ask whether the model can be applied immediately, or whether it will require further calibration, and, if it does, how much experimentation and adaptation will be required.
Miller (BPTG): The process model or “scale-down” model used to develop and optimize a biotherapeutic manufacturing process should meet a number of criteria. First, the process model should have the same number, type, and duration of process steps as the actual process at production scale. In addition, product yields (e.g., volumetric productivity or percent recovery) should be similar to product yields at actual production scale. It is also important to consider that the critical quality attributes (i.e., key structural and functional aspects) for product manufactured using the process model should be similar to those of product manufactured at actual production scale, as determined by analytical characterization.
BioPharm: What would be the benefits and challenges of outsourcing/partnering process modeling vs. doing it in-house?
Miller (BPTG): Outsourcing process modeling eliminates the need to build and fund internal staff, infrastructure, and lab space, which can reduce fixed costs. However, it also carries with it a host of challenges. When a company outsources process modeling, it has less control over data, experimental outcomes, and intellectual property. Additionally, it has less flexibility to make experimental changes ‘on the fly’ or to react to data. Essentially, one trades greater control and flexibility for more convenience and lower costs.
Keeping process modeling in-house allows for quicker turnaround time between experiments and more flexibility to make experimental changes in response to data. It offers more control over intellectual property and provides the opportunity to ‘learn by doing’. Of course, the trade-off here is that there are costs to building the staff and infrastructure and securing laboratory space.
Arsenio (Sartorius): The benefits of outsourcing include having access to a wider pool of suppliers and, potentially, accelerated innovation. Outsourcing offers the possibility of rapidly adopting the most advanced technologies. However, with outsourcing, core competencies remain external to the developer’s organization, leaving the manufacturer dependent on external expertise for model building, calibration, application, and maintenance.
BioPharm: Are current process models adequately equipped to deal with increasingly complex therapeutic molecules?
Arsenio (Sartorius): A current challenge, especially for the development and production of more complex therapeutic molecules, concerns better process understanding, modeling, and control. Current bioprocess instruments are composed of a multitude of univariate feedback loops to control critical process parameters (CPPs), such as pH. However, this approach neglects the interdependence and interaction among variables that affect the system and makes it difficult to reliably infer the controlled system’s overall robustness, stability, and sensitivity to disturbances. This triggers the adoption of control strategies that are strongly reliant on the system design, with small margins for system uncertainty and disturbances.
This issue can be addressed by better multivariate models and multivariate control strategies for CQAs, for instance, the continuous automation of the adjustment of setpoints based on simulation outputs, which would close the loop of adaptive and automated model predictive control. Combined with state-of-the-art automated process batch management and batch recipe-control functionalities that meet American National Standards Institute (ANSI)/International Society of Automation) ISA-88 standards, as well as the use of electronic batch records and multivariate models, these control strategies might pave the way to real-time release testing based on batch fingerprinting.
Miller (BPTG): Cell and gene therapy products are much more complex than mAbs or protein therapeutics, and the process models for manufacturing them can be more difficult to construct. For example, many process steps may be difficult to perform at a smaller scale while still being representative of actual production scales. New models are needed to facilitate process optimization while reducing development costs and timelines. Improvements in process monitoring for cell and gene therapy manufacturing are also needed to enable learning how manufacturing process steps may affect the critical quality attributes (CQAs) of the products.
BioPharm: What process model(s) have been successful so far in meeting the biopharma industry’s current needs?
Miller (BPTG): Process models for mAbs and recombinant proteins produced in mammalian or microbial cell lines have become very successful tools for enabling manufacturing process optimization and/or troubleshooting. Scaling up production from small-scale models to actual production scale is generally straightforward. However, scale-down models for cell and gene therapy products are currently less reliable than for mAbs or protein therapeutics, and data from these models must be supported by actual production-scale data during process development.
Arsenio (Sartorius): The increasing complexity of therapeutic molecules being developed in the biopharmaceutical pipeline requires innovative process analytical technology (PAT) approaches, namely the capability for inline or online CQA monitoring and control. Technology vendors are further evolving process controls and analytics by optimizing the full potential of Raman spectroscopy in bioreactor systems. For example, Sartorius has launched a new spectroscopy platform that offers enhanced process control of the company’s ambr high throughput multiparallel bioreactor systems. Using this platform, a connected, online Raman spectroscopy sensor analyzes multiple parameters and further integrates sensor data with cell culture and process information to build robust models. Such models can be employed not only for monitoring but also to design feedback-control strategies (e.g., automated feed and bleed controls) for controlling several CPPs or CQAs, such as glucose, viable cell density, or titer, and reduce the risks associated with manual sampling and control.
To remain competitive, biopharmaceutical companies are also continually looking at strategies to reduce costs (e.g., increasing the flexibility of their biologics’ production and improving speed-to-market.) Leveraging PAT and associated software technologies, allows process parameters to be predicted, accurately, and automated process control strategies to be applied so that both performance and product quality may be improved. This approach also speeds up and simplifies process development and transferability for commercial manufacturing.
Vol. 33, No. 8
When referring to this article, please cite it as F. Mirasol, “Using Process Modeling to Troubleshoot a Bioprocess,” BioPharm International 33 (8) 20–21 (2020).