(360e) Dynamic Optimization of an Ultrafiltration System for the Concentration of Monoclonal Antibody Solutions Under Uncertainty

Monoclonal antibodies (mAbs) are a very important type of therapeutic agent with myriad applications in several therapeutic areas, such as immunology, oncology and infectious disease. These immunoglobins owe their success to their higher efficacy, specificity and safety compared to conventional drugs, which is a direct consequence of their mechanism of action. Unfortunately, all these benefits come at a price, namely, mAbs are much more expensive than conventional drugs. In fact, they may cost between 1,000 and 50,000 $/g.

One of the reasons why monoclonal antibodies are extremely expensive is their complex and low-yield production process, which includes several challenging manufacturing steps, namely, an initial fed-batch fermentation followed by centrifugation, depth filtration, several chromatographic purifications, virus filtration and, finally, concentration by ultrafiltration. Therefore, one of the most direct ways to reduce the average cost of monoclonal antibodies is to optimize their production process, in terms of both processing sequence and operating conditions at which the individual manufacturing steps are carried out.

Several authors have proposed optimal design configurations for mAb production processes (Liu et al., 2016) and computed the optimal operating conditions at which the fermentation and chromatographic separations should be carried out (Alhuthali et al., 2019; Papathanasiou et al., 2017). Some of these contributions have also analyzed the impact of model uncertainty on the results of their design and on the associated dynamic optimization calculations (Liu and Gunawan, 2017). Surprisingly, the ultrafiltration step has received very little attention so far, even though the typical yield of this operation may be as low as 90 % (as the monoclonal antibody is being concentrated, a significant amount of it forms a hydrogel that sticks onto the ultrafiltration membrane). Therefore, this contribution investigates the extent to which the systematic use of stochastic dynamic optimization techniques can help increase the efficacy/yield of the ultrafiltration step and reduce its duration.

More specifically, we first develop a hybrid mechanistic-statistical model of the ultrafiltration device, which is comprised of a simplified (yet first-principles inspired) model of the physical system, augmented with a Bayesian neural network (this type of model retains all the benefits of first-principles models, learns from experimental/process data as data-driven models, and provides point estimates of the process variables of interest as well as estimates of their degree of uncertainty). Then, we train this hybrid model on experimental data, collected at several different operating conditions, using advanced optimization-driven uncertainty quantification algorithms. Finally, we make use of the trained hybrid model and of a stochastic dynamic optimization algorithm to optimize the operating conditions of the ultrafiltration system under uncertainty with two principal goals, namely, maximization of the expected process efficacy/yield, measured as final mAb concentration/recovery, and minimization of the ultrafiltration time.

The results of our calculations are extremely promising, in that we can both improve the final mAb concentration/recovery and reduce the ultrafiltration time without violating any operational constraints, even in the presence of significant model uncertainty.

References

Alhuthali, S., Fadda, S., & Kontoravdi, C. (2019). Constrained Optimisation of Cell Culture Feeding Strategy and Temperature Shift Duration to Enhance Monoclonal Antibody Titre and Purity. 2019 AIChE Annual Meeting, Orlando (FL).

Liu, Y., & Gunawan, R. (2017). Bioprocess optimization under uncertainty using ensemble modeling. Journal of biotechnology, 244, 34-44.

Liu, S., Farid, S. S., & Papageorgiou, L. G. (2016). Integrated optimization of upstream and downstream processing in biopharmaceutical manufacturing under uncertainty: a chance constrained programming approach. Industrial & Engineering Chemistry Research, 55(16), 4599-4612.

Papathanasiou, M. M., Quiroga‐Campano, A. L., Steinebach, F., Elviro, M., Mantalaris, A., & Pistikopoulos, E. N. (2017). Advanced model‐based control strategies for the intensification of upstream and downstream processing in mAb production. Biotechnology progress, 33(4), 966-988.