The fed-batch cultivation is in many ways a benchmark for fermentation processes, and it has been an attractive choice for the biotechnological production of various products in the past decades. The majority of biopharmaceuticals that are presently undergoing clinical trials or are available on the market are manufactured through fed-batch fermentations. A crucial process parameter in fed-batch cultivations is the substrate feed rate, which directly influences the overall process productivity, product quality and process repeatability; henceforth, effective control of this parameter is imperative for a successful fed-batch fermentation process. Two distinct control strategies can be distinguished—open-loop and closed-loop (feedback) control. Each of these methods has its own set of benefits, limitations and suitability for specific bioprocesses. This article surveys and compares the most popular open- and closed-loop methods for substrate feed rate control in fed-batch fermentations. Emphasis is placed on model-predictive feed rate control (MPC)—a stand out among other methods that offers a promising application perspective. The authors also demonstrate a practical example of the implementation of a robust, flexible MPC solution that is suitable for various cultures and runs on standard computer hardware, thus overcoming one of the main reported MPC drawbacks—high computational requirements.
A model predictive control (MPC) method was investigated as a route to optimize and control the growth of E. coli BL21 (DE3) and biosynthesis of two different recombinant proteins (nerve growth factor NGF and coat protein of bacteriophage Qβ (Qβ-CP)). To determine the target trajectory for the E. coli cultivation process and estimate the model parameters, the off-line run-to-run optimization method was used. The proven method allowed us to successfully control the growth of microbial biomass, with a deviation of 6–12% from the target trajectory. It was proven that it is possible to obtain a “Golden Batch” profile for the implementation of MPC using datasets from only four to eight fermentation runs. The method showed its robustness when the cultivation of E. coli was carried out with two different titrant supply control systems—volumetric and gravimetric. Furthermore, the MPC method exhibited high adaptability, reliability, and resistance to various types of disturbances. MPC proved to be a reliable and effective method for controlling the cultivation and recombinant protein biosynthesis of fast-growing microorganisms such as E. coli.
Soy leghemoglobin (LegH) has been gaining interest over the last years as an efficient flavor and aroma compound in plant-based meat substitutes. Hence, in the following article, we demonstrate the methods for LegH production using a recombinant Komagataella phaffii strain. Multiple fed-batch fermentation with an alternative to a BSM medium, where glucose was used as the main carbon source, was implemented and the growth kinetics, e.g., a maximal specific biomass growth of 0.239 g·g−1·h−1, a biomass yield from the substrate of 0.298 g·g−1, and a maximal specific substrate consumption rate of 0.81 g·g−1·h−1 were identified. Leghemoglobin production resulted in a yield of 0.513 mg·gDCW−1, while the highest biomass density achieved in this study was 121.80 gDCW·L−1. The applied medium that showed potential for additional optimization studies, which, in contrast to BSM, made it possible to separate pH control from nitrogen supply, does not affect medium turbidity measurements and does not induce metabolite synthesis during yeast biomass growth.
Hybrid modeling techniques are increasingly important for improving predictive accuracy and control in biomanufacturing, particularly in data-limited conditions. This study develops and experimentally validates a hybrid deep learning model predictive control (MPC) framework for recombinant P. pastoris fed-batch fermentations. Bayesian optimization and grid search techniques were employed to identify the best-performing hybrid model architecture: an LSTM layer with 2 hidden units followed by a fully connected layer with 8 nodes and ReLU activation. This design balanced accuracy (NRMSE 4.93%) and computational efficiency (AICc 998). This architecture was adapted to a new, smaller dataset of bacteriophage Qβ coat protein production using transfer learning, yielding strong predictive performance with low validation (3.53%) and test (5.61%) losses. Finally, the hybrid model was integrated into a novel MPC system and experimentally validated, demonstrating robust real-time substrate feed control in a way that allows it to maintain specific target growth rates. The system achieved predictive accuracies of 6.51% for biomass and 14.65% for product estimation, with an average tracking error of 10.64%. In summary, this work establishes a robust, adaptable, and efficient hybrid modeling framework for MPC in P. pastoris bioprocesses. By integrating automated architecture searching, transfer learning, and MPC, the approach offers a practical and generalizable solution for real-time control and supports scalable digital twin deployment in industrial biotechnology.
Crypthecodinium cohnii is a marine microalga that can accumulate high amounts of polyunsaturated fatty acids (PUFAs) and thus replace conventional routes of fish oil production. They are associated with the destruction of marine resources and multiple downstream/purification complications. The major drawbacks of using C. cohnii for industrial-scale production are associated with low PUFA productivity. One of the means of increasing the PUFA synthesis rate is to maintain the medium component concentrations at optimal values throughout cultivation, thus increasing PUFA production efficiency, which can result in the successful transfer of the process to pilot and/or industrial scale. The goal of the present research was to develop techniques for increasing the efficiency of PUFA production via C. cohnii cultivation. Multiple experiments were carried out to test and fine-tune the cultivation medium composition and oxygen transfer factors. The biomass yields from individual components, yeast extract, sea salts, and glucose amounted to 5.5, 0.65, and 0.61 g·g−1, respectively. C. cohnii cell susceptibility to mechanical damage was experimentally evaluated. Power inputs of <276.5 W/m3 did not seem to promote cell destruction when Pitched-blade impellers were used. The obtained cultivation conditions were shown to be efficient in terms of increasing the biomass productivity and the omega-3 fatty acid content in C. cohnii. By using the applied methods, the maximal biomass productivity reached 8.0 g·L−1·day−1, while the highest obtained biomass concentration reached 110 g·L−1. A steady increase in the concentration of PUFAs during cultivation was observed from the FTIR data.
There are several ways in which mathematical modeling is used in fermentation control, but mechanistic mathematical genome-scale models of metabolism within the cell have not been applied or implemented so far. As part of the metabolic engineering task setting, we propose that metabolite fluxes and/or biomass growth rate be used to search for a fermentation steady state marker rule. During fermentation, the bioreactor control system can automatically detect the desired steady state using a logical marker rule. The marker rule identification can be also integrated with the production growth coupling approach, as presented in this study. A design of strain with marker rule is demonstrated on genome scale metabolic model iML1515 of Escherichia coli MG1655 proposing two gene deletions enabling a measurable marker rule for succinate production using glucose as a substrate. The marker rule example at glucose consumption 10.0 is: IF (specific growth rate μ is above 0.060 h−1, AND CO2 production under 1.0, AND ethanol production above 5.5), THEN succinate production is within the range 8.2–10, where all metabolic fluxes units are mmol ∗ gDW−1 ∗ h−1. An objective function for application in metabolic engineering, including productivity features and rule detecting sensor set characterizing parameters, is proposed. Two-phase approach to implementing marker rules in the cultivation control system is presented to avoid the need for a modeler during production.