Jeanne Linke Northrop: Hello, I'm Jeanne Linke Northrop with BioPharm international. I'm joined today by Nora Ketterer, Manager of Modeling Services, Cytiva. Nora, welcome. Thank you so much for joining me today. I’m looking forward to talking with you about in silico process development and the application of mechanistic models. Quickly, can you recap some of the main benefits of in silico process development? And then in a little bit more detail, can you shed some light on using mechanistic models that simulate and support chromatography process development.
Nora Ketterer: Yes, sure. Enhancing quality, capacity, speed and flexibility throughout the process are what makes it difference when developing a biopharmaceutical for successful market entry. By simulating the process to gain a deeper understanding, you create the opportunity to meet all of those objectives. One main benefit of mechanistic modeling to support chromatography process development is that a high level of process understanding is created with only a limited set of experiments. Once a mechanistic model is created, the model can be used to run thousands of in silico experiments overnight. This is a huge benefit, for example, for process characterization in late-stage PD. Due to the mechanistic nature of the models, it's even possible to extrapolate from the space investigated during model calibration. Mechanistic modeling can also be used to speed up early-stage PD for off-platform molecules where conventional PD might be very time consuming and tedious. In general, having a mechanistic model in place that can be used to run in silico predictions anytime, anywhere without the need of lab equipment is a big advantage and proved especially valuable during the pandemic.
Jeanne Linke Northrop: Thanks, Nora. What are the typical challenges process developers face in their daily work?
Nora Ketterer: Our customers are typically under a lot of pressure to develop robust chromatography processes with a limited amount of material available, especially with off-platform molecules. Gaining adequate process understanding typically requires a large number of experiments with a valid calibrated model in place. The number of needed experiments can be reduced, which saves resources and time.
Jeanne Linke Northrop: I also want to talk a little bit with you about simulations and process development. Digitalization is changing the industry. In terms of using simulations and process development, is there a risk that PD scientists won’t be working in the lab anymore?
Nora Ketterer: No. So mechanistic modeling is a tool that helps customers with their daily work challenges. Regarding the lab work, there are still experiments required to calibrate and also validate the model. The experimental planning might differ slightly, but not necessarily in the complexity of the experiments that are needed but with a different focus for the experiments. So experiments that are performed for modeling aim to capture the dynamics of the interaction between the biomolecules and the ligand. This can be realized by running, for example, linear gradient experiments at low density, and also step elution experiments at different conditions. So once calibrated, the model can be used, for example, to find an optimal process condition or to assess the process robustness by simulating different conditions. There will still be live experiments needed to validate the model's predictions. But those could be focused, for example, on the previously identified parameters where the model predicted that the process will perform well. Or, that this might be the edge of failure of the process.
Jeanne Linke Northrop: Great point. Thank you so much. I want to talk more about utilizing a mechanistic modeling approach. What are some typical use cases?
Nora Ketterer: Once a model is calibrated, it can be used to support the entire PD lifecycle. Recently, we worked with scientists at Takeda. There, we used modeling for process optimization. This resulted in improvements in purity and yield of a mixed mode chromatography purification step, and doubled productivity. Mechanistic modeling can also support early-phase development. The technology can help shorten timelines, especially for off-platform molecules. And using the generated process understanding to define robust process operating ranges to mitigate risk is a great tool to be applied in process characterization and validation studies, and can be used for filing as well. So moreover, mechanistic models can be of great help when scaling up. A lot of the effects of the thermodynamics that are happening inside the chromatography column are scale independent. Therefore, it's really easy to also predict what happens at a larger or smaller scale. And mechanistic models can also be used to predict what happens in the manufacturing column and can be used for troubleshooting.
Jeanne Linke Northrop: What should be kept in mind when someone is starting to work with mechanistic models, and, in general, in silico process development techniques?
Nora Ketterer: For new users who do not have prior experience with mechanistic modeling, it's really, really important to keep in mind that getting familiar with this technology will take some time and practice. The GoSilico software is very user-friendly and intuitive to use. This makes it a lot easier. But still, of course the technology is new to scientists and, therefore, takes time and experience. The good thing is that with the GoSilico software, there are no requirements to have any prior background experience in programming or mathematics. All of the equations are basically hidden in the user interface. But of course, some expertise in knowing how mechanistic modelling is done, how the workflows work, how the process and how the project should be conducted are important. This all develops over time, and only when working on different types of projects. As resources are sometimes limited, an alternative strategy that is taken by some customers is actually to outsource modeling activities to Cytiva and to use the expertise that we have here at Cytiva instead of building the capabilities in-house.