Advanced therapy medicinal products (ATMPs) are a focus for biopharma. The ability to treat disease in a targeted manner, the reduced likelihood of competition, and regulatory support make such products attractive R&D targets.
The revenue potential of ATMPs is another factor. Estimates suggest the global market for ATMPs could grow to $9.6 billion by 2026.
But for the market to reach its full potential, industry must find efficient ways of making ATMPs, according to Maria Papathanasiou, PhD, assistant professor, Imperial College London’s department of chemical engineering, who says mathematical modeling is key.
“Mathematical models and/or computer modeling tools can assist decisions throughout drug development from discovery all the way to therapy distribution,” she says. “In manufacturing, such tools can help with decision making related to selection of units, optimization of processes, identification of optimal conditions of operation, and they can also be used as soft sensors when measurements are not readily available.”
For Papathanasiou, the key benefit of building a mathematic model is that it allows for the systematic analysis of the system considering multiple factors at the same time.
“The impact of synergetic/antagonistic effects can be better studied. Also, they provide a cost-efficient basis for in silico experimentation, decreasing the time and labor required for wet lab experiments,” she tells GEN. “Such tools give us the ability to run a very high number of experiments on the computer and then decide which of those we want to validate in the lab. They have a great potential to assist with initiatives such as Quality-by-Design and Design Space Identification as well.”
Models can also be used on the factory floor, Papathanasiou says, explaining “they prove to be a very powerful tool for in-process monitoring and online control as validated models can be used as soft sensors when online measurements are not readily available.”
Role of models in process development
Papathanasiou, who co-authored a recent study examining the role modeling can play in process development for ATMPs and next-generation vaccines, says choosing the correct type of model on which to base process development is a critical step.
“Such models can be mechanistic, data-driven, or hybrid. Each class of models is different with mechanistic models being the most detailed in terms of translating the physicochemical characteristics of the system into mathematical equations (each variable and parameter reflects an actual entity of the system at hand),” she explains. “Data-driven models sit at the other end where they use input/output datasets to capture the system dynamics, without the need to know what is exactly going on in terms of physicochemical properties, reactions, etc.”
Choice of model is important as each require different types of data and process monitoring technology required.
“Depending on the class of models one uses the amount of data differs. In our collaborations with companies, we use primarily existing or offline data for the development of such tools and the current analytics tie in nicely,” she continues. “For an online implementation of such tools that are also known as digital twins advanced PAT would really advance the usability.”