Mechanistic modeling – become an expert or engage an expert?

ChromX software user in front of an Äkta system.

As pioneers of in silico process development, GoSilico is committed to enabling the digital transformation of bioprocessing by computer simulation. Currently seven of the top ten biopharma companies worldwide employ GoSilico’s ChromX simulation technology to improve their bioprocesses. This software has been designed to be as intuitive as possible whilst maintaining the versatility that makes mechanistic modeling such a powerful tool for bioprocess development and process understanding. Ensuring the technology is user-friendly has reduced the barrier to entry significantly over the last few years, however, the application of mechanistic modeling still requires expertise for effective use. When planning to implement this ground-breaking technology one essential questions must be addressed beforehand: When is it most appropriate to build this expertise in-house, or when to contract in established experts?

Do I need an interdisciplinary team of experts to pursue modeling?

Modeling is an interdisciplinary and complex affair, though the barrier of entry has fallen massively in the past few years. In its early years, the application of mechanistic modeling was reserved to academia, as it required profound expertise and knowledge in mathematics, computer science and engineering.

Over the past decades, this situation has changed: The increasing need for mechanistic process understanding, along with the growing time-to-market pressure and process economic drivers has driven the need of mechanistic modeling. This, in turn, has spurred the development of intuitive and user-friendly simulation tools such as ChromX, that are far more powerful, versatile and accessible than ancient academic command line tools and alike. Our in silico tools shroud the complex mathematics behind accessible graphical user interfaces and make simulation feasible to process engineers and scientists without prior expertise in mathematics and computer sciences, with the sophisticated mathematical solution procedures already preset.

However, as mechanistic chromatography models rely on natural laws describing fluid- and thermo-dynamic phenomena, implementing an effective model still requires expertise and understanding of the driving thermodynamic and fluid-dynamic effects inside a chromatography column and how they are represented in the solution framework. It is mandatory to understand the significance and impact of the model equations in gaining insights into the physical processes, such as discerning physically reasonable parameter values for the applied isotherm equations. As intuitive as mechanistic modeling tools are today, their application still needs some expertise that must be built internally, or outsourced.

Build the expertise in-house or collaborate with experts?

Collaborative workers analyzing data, become or engage an expert

One of the most essential questions before implementing mechanistic modeling in your company or department is the intended field of application. The use may cover the whole product life cycle, from early process development to late stage process characterization, scale-up or root cause investigations, but may also comprise one specific subset only, with each application subject to varying complexity and therefore required know-how.

Modeling expertise is best developed stepwise, an increasing stage of complexity should be modeled starting with simple processes and applications and evolving these models over the life of the development. If timelines or personnel constraints impede this or the intended application is simply too complex for novices, outsourcing the modeling activities partially or completely may be favorable.

For the implementation of mechanistic modeling, it all comes down to two options: becoming an expert or engaging an expert. But how can you decide which option is the right choice for one’s company or team?

GoSilico has been successfully implementing mechanistic models in the biopharmaceutical industry since 2016, having customers amongst the biopharma giants, CMDOs and smaller biotech companies. Tailored service offerings provide modeling solutions applicable across the entire industry, thus providing several routes to gaining mechanistic modeling expertise. Which options are available? Let’s have a more detailed look.

Option 1: Build the expertise

For many companies becoming an expert is the option of choice, as they plan to capitalize on mechanistic modeling extensively in their future process development activities. Digitalization is proceeding at great speed and more and more companies are generating inhouse modeling expertise to keep up with industry standards and develop better and more robust processes in shorter timelines, while meeting regulatory demands of Quality-by-Design and associated process understanding.

When it comes to the decision of becoming an expert or contracting one, timelines and resources are the key questions. To become an expert in modeling and to develop the required skills in-house, it is necessary to practice modeling frequently. If that is not possible because of time resources, the progress may be slower. For this reason, most companies are not only training one person but empower teams of scientists to become proficient users of ChromX. Within the team it is easier to discuss challenges, speed up the knowledge acquirement and to adapt company internal standard operating procedures and workflows to the needs of modeling. If team members are not solely dedicated to modeling but also to other tasks such as wet-work, as it is the case for most companies, this modeling team approach is highly beneficial. There are also several companies where individual people are almost exclusively occupied with modeling tasks after developing their modeling skills, consequently, they are highly proficient experts, often developing processes in silico across multiple sites from the comfort of their desks.

If a company or department is completely new to modeling it may be fortunate to hire one of those experts – but supply of these skilled personnel rarely meets demand. For practical reasons, it may therefore be the only option to build up the expertise in-house within the future modeling team.
To get started, new customers planning to do modeling in-house, participate in an intensive, interactive and tailored ChromX training covering the theoretical background, modeling workflows and hands-on practice to get to know mechanistic modeling. As ChromX is still an expert tool, additional practical experience is required for proficient usage, also dependent on the prior modeling experience of the new ChromX users.

Flowchart of ChromX roll-out and implementation

This additional experience can best be acquired by modeling own projects, much like any technical skill, and it makes sense to start with the simpler modeling cases to bootstrap expertise. The nature of the chromatographic separation, process used and quality of data impact model complexity and mean some modes of interaction are easier to model than others. A process which is modeled frequently is a monoclonal antibody (mAb) in bind and elute mode on a cation exchange resin (CEX), as the required equations are especially well-understood and consequently their application is straight forward. Experience shows that modeling novices need about two or three times longer to finish a modeling project as an expert would consume on the same tasks, even with project overheads such as data transfer and limitations on working remotely across time zones.

Therefore, almost all new customers rely on GoSilico’s consultancy to be guided through their first projects, enabling first success cases to be easily realized within 3-6 months. Afterwards, the processes to be modeled can increase gradually, typically through modeling more complex modes of interaction such as mixed-mode chromatography with varying pH. Following this approach, the modeling capabilities will be built up stepwise, organically, through first-hand experience.

Option 2: Collaborate with experts

The previously outlined example does not cover all routes: What is the appropriate course if timelines, resource or required complexity are restrictive? Is modeling still a solution?

The answer to that question is clearly “yes”. Modeling is a powerful tool to reduce the experimental workload and to design safe and efficient processes within the given timelines, but as the timelines are sometimes too challenging, it may not be possible to build the required expertise in-house and in time. Alternatively, the initial modeling activities can be outsourced by engaging GoSilico’s modeling experts, which reduces the timeline for first projects from months to weeks.

The benefit of this is that GoSilico uses its years of modeling experience, proven bioprocess expertise and resources to develop digital twins of your process. GoSilico can perform all modeling activities, from experimental design, model selection, data review, model calibration, model interrogation and documentation, all from our office in Germany. The customer team is then only responsible with either providing historical data, performing a few relatively simple calibration and validation experiments with always-available consultation and guidance from GoSilico.

Developer at GoSilico looking at screen

Fast troubleshooting of projects often requires deep know-how and proficient knowledge in modeling, considering deviations and unexpected performance is typically due to a combination of different, often seemingly unconnected, process parameters and material attributes. Root cause analysis is a typical example in which immediate access to mechanistic process understanding often means the consultation and contract modeling workflow is appropriate. Additionally, complex process modalities, such as continuous chromatography, pH dependent mixed-mode separations and developing adaptive control schemes require a greater level of modeling understanding.

These processes, when combined with the typical drivers of time-to-market, and lean process development, often mean one cannot wait for expertise to be built, but may rely on GoSilico to develop these models, provide updates, documentation and further training such that the knowledge can be managed and disseminated between sites. Even established modeling teams may require GoSilico’s world leading expertise for rapid digital twin synthesis.

A solution tailored to your needs

To summarize, initially it may be time consuming to develop in-house expertise, these 3-6 months are an investment in future capabilities, both with process and personnel. If timelines or resources are limiting, often customers contract GoSilico to share their know-how, to manage all aspects within a project or perform the modeling activities in-house. There is never a single solution for the entire industry, but GoSilico works to tailor an offering considering all factors impacting a prospective modeling project to accelerate the adoption of mechanistic models, better bioprocesses and therefore more accessible medicines.