(540d) Efficient Experimental Design for Bioprocesses across Scales Using Bayesian Optimization

To address this challenge, we introduce a probabilistic decision-making framework that integrates bioprocess modeling and Bayesian optimization to design more efficient experimental strategies. Our approach systematically incorporates information from multiple scales, reducing experimental costs while improving bioprocess performance.

Bioprocess model for scale- and clone-dependent analysis

A key component of our approach is a customized bioprocess model for Chinese hamster ovary (CHO) cells, adapted from Craven et al. [2]. The model captures scale-dependent process dynamics, enabling us to evaluate different scale-up scenarios and the limitations of small-scale experiments in representing pilot-scale performance. By incorporating realistic feeding strategies, experimental noise, and cell line variability (e.g. differences in lactate metabolism), our model generates in silico data that mimics real-world scale-up challenges.

This model allows us to compare traditional bioprocess development strategies with a more systematic, probabilistic approach, highlighting when and why a sequential screening might be leading so suboptimal experiments compared to early on performing few experiments at lager scale.

Bayesian optimization for experimental designs across scales

Bioprocess development from early-stage cell line screening to optimization of process condition is costly and time-consuming, often requiring days to weeks per experiment. This leads to a typical exploration-exploitation trade-off: new clones and conditions need to be screened to identify the highest-performing candidates, while at the same time, the most promising ones must be tested repeatedly to ensure robust statistical evaluation. Conventional experimental strategies often follow a linear, sequential screening approach, where clones are discarded early based on small-scale results, and scale-up is performed only after a fixed number of experiments. This can result in inefficient experimental allocation and suboptimal process optimization.

In this work, we introduce a multi-fidelity batch Bayesian optimization (BO) framework for experimental planning in bioprocess development. BO is well-suited for optimization problems where evaluations, such as bioprocess experiments, are costly, as it balances exploration and exploitation based on probabilistic models. Our approach extends BO by explicitly considering multiple experimental scales (multi-fidelity), allowing for dynamic selection of the scale at which experiments should be conducted (Figure 1A). In addition, batch BO allows to consider different batch sizes, which is important to integrate parallelized screening stages such as microtiter plates vs. individual bioreactors at later stages.

Using in silico data generated from our customized bioprocess model (Figure 1B), we evaluate how the multi-fidelity batch BO framework compares to conventional heuristic-driven screening.

Case studies and performance evaluation

Throughout the work, different industry-inspired case studies are considered, including the influence of different batch sizes, the number of clones and the assumed changes in process behavior between different scales. Specifically, we analyze the impact of the multi-fidelity batch BO approach on the titer of a simulated pharmaceutical product, evaluating the number of experiments that are required to reach certain titers and confidence levels in the prediction.

Figure 1 shows one such case, in which multi-fidelity batch BO for experiments at three scales (microtiter plate, medium-scale reactor system, and a pilot scale) is compared to a classical sequential screening strategy. Specifically, we constructed this scenario to evaluate how a suboptimal scale-down (yellow), e.g. due to unfavorable oxygen supplementation, can impact the viable cell density and product formation in small scale and thus the information gained from this screening stage.

Our results demonstrate that integrating information from all scales early in bioprocess development significantly reduces costs and improves product titers with fewer experiments compared to classical sequential screening. Our findings thus suggest that early-stage multi-scale experimentation can streamline decision-making and accelerate bioprocess optimization.

Industrial relevance and future applications

The combination of cell line screening and process optimization across scales remains a challenge with high industrial relevance, which is currently not sufficiently addressed in research. Current heuristic-driven approaches often lead to inefficient experimental allocation and suboptimal scale-up decisions. This work provides a new approach that takes various features of real-world experiments into account: investigating both cell lines and process conditions, different confidence in experimental set-ups across scales and batch-wise experimentation.

By reducing the number of required experiments while maintaining or improving product titers, our approach offers significant potential for industrial application. Future work could further explore integrating additional information about cell lines, including their genetic characteristics, to enhance predictive accuracy. This study marks a step towards more efficient, systematic screening strategies, moving beyond the status quo in bioprocess development.