(10a) Digital Twin for Continuous Lyophilization of Biotherapeutics

Lyophilization (also known as freeze drying) is a crucial process in pharmaceutical manufacturing by which the liquid solvent is removed from drug product via sublimation under vacuum. Recent studies have shown that lyophilization can provide long-term stability for mRNA vaccines, which eliminates the need for cold chains during vaccine distribution and hence holds promise for the future of mRNA and biotherapeutics manufacturing [1,2]. While extensive efforts have been dedicated to shifting the pharmaceutical industry toward continuous manufacturing, the majority of industrial-scale lyophilization is still being operated in batch mode [3]. Recently, a novel continuous lyophilization technology was proposed by [4], in which the vials are suspended and move continuously through the process. This suspended-vial technology has several benefits over other existing continuous lyophilization schemes, in particular its simple configuration allowing the product quality control to be done more conveniently and rigorously [3,4].

This work presents a digital twin for continuous lyophilization considering the suspended-vial technology. Establishing a two-way flow of information, our digital twin receives and processes data from its physical counterpart, performs several complex computations in real-time, including multiphysics simulation, state estimation, dynamic optimization, and uncertainty quantification, and provides instructions/insights back to the physical system. The digital twin serves as a virtual representation of the physical system for guiding and enhancing the manufacturing system.

In this work, four key components of our digital twin are discussed. First, we present a novel mechanistic model for continuous lyophilization [5]. The model describes key transport phenomena in all three steps of lyophilization, namely freezing, primary drying, and secondary drying. The validated model can accurately predict the evolution of critical process parameters, including the product temperature, ice/water fraction, sublimation front position, and amount of residual moisture (bound water), for the entire process. Second, we showcase a state observer that estimates the amount of bound water in the product by using only temperature measurement as an input [6]. This state observer allows for real-time tracking of the residual moisture, which is typically present in trace amounts and difficult to measure online, offering more data to the digital twin. Third, we describe a new, efficient way for solving optimal control problems by reformulation as a hybrid discrete/continuous system of differential-algebraic equations [7,8]. Our proposed algorithm is more than an order of magnitude faster than the traditional approaches while maintaining the same level of accuracy. The algorithm is also demonstrated for finding the optimal control policy for continuous lyophilization. Finally, we develop a highly efficient framework that accounts for the probabilistic uncertainty in continuous lyophilization via polynomial chaos theory (PCT) [9]. Our PCT-based model is applied for fast uncertainty quantification and stochastic control of continuous lyophilization. All the frameworks presented in this work enable highly efficient computation for various applications, which can be adapted and extended further to build digital twins for other manufacturing systems.

Acknowledgements:

RDB and PS were supported by the U.S. Food and Drug Administration under the FDA BAA-22-00123 program, Award Number 75F40122C00200.

References

[1] H. Muramatsu, K. Lam, C. Bajusz, D. Laczkό, K. Karikό, P. Schreiner, A. Martin, P. Lutwyche, J. Heyes, N. Pardi, Lyophilization provides long-term stability for a lipid nanoparticle-formulated, nucleoside-modified mRNA vaccine, Molecular Therapy 30 (5) (2022) 1941–1951.

[2] R. Pisano, A. Arsiccio, L. C. Capozzi, B. L. Trout, Achieving continuous manufacturing in lyophilization: Technologies and approaches, European Journal of Pharmaceutics and Biopharmaceutics 142 (2019) 265–279.

[3] R. Pisano, A. Arsiccio, L. C. Capozzi, B. L. Trout, Achieving continuous manufacturing in lyophilization: Technologies and approaches, European Journal of Pharmaceutics and Biopharmaceutics 142 (2019) 265–279.

[4] L. C. Capozzi, B. L. Trout, R. Pisano, From batch to continuous: Freeze-drying of suspended vials for pharmaceuticals in unit-doses, Industrial & Engineering Chemistry Research 58 (4) (2019) 1635–1649.

[5] P. Srisuma, G. Barbastathis, R. D. Braatz, Mechanistic modeling and analysis of transport phenomena for continuous lyophilization (2024). arXiv:2409.06251.

[6] P. Srisuma, G. Barbastathis, R. D. Braatz, Real-time estimation of bound water concentration during lyophilization with temperature-based state observers, International Journal of Pharmaceutics 665 (2024) 124693.

[7] P. Srisuma, G. Barbastathis, R. D. Braatz, Simulation-based approach for optimal control of a Stefan problem, in: American Control Conference, 2024, pp. 3031–3036.

[8] P. Srisuma, G. Barbastathis, R. D. Braatz, Simulation-based approach for fast optimal control of a Stefan problem with application to cell therapy (2024). arXiv:2412.18272.

[9] P. Srisuma, G. Barbastathis, R. D. Braatz, Probabilistically robust uncertainty analysis and optimal control of continuous lyophilization via polynomial chaos theory, in: American Control Conference, 2025. (in press)