Machine learning in bioprocess development: from promise to practice

Advances and opportunities in process analytical technologies for viral vector manufacturing

Viral vectors are an emerging, exciting class of biologics whose application in vaccines, oncology, and gene therapy has grown exponentially in recent years. Following first regulatory approval, this class of therapeutics has been vigorously pursued to treat monogenic disorders including orphan diseases, entering hundreds of new products into pipelines. Viral vector manufacturing supporting clinical efforts has spurred the introduction of a broad swath of analytical techniques dedicated to assessing the diverse and evolving panel of Critical Quality Attributes (CQAs) of these products. Herein, we provide an overview of the current state of analytics enabling measurement of CQAs such as capsid and vector identities, product titer, transduction efficiency, impurity clearance etc. We highlight orthogonal methods and discuss the advantages and limitations of these techniques while evaluating their adaptation as process analytical technologies. Finally, we identify gaps and propose opportunities in enabling existing technologies for real-time monitoring from hardware, software, and data analysis viewpoints for technology development within viral vector biomanufacturing.

Optimization of culture condition for Spodoptera frugiperda by design of experiment approach and evaluation of its effect on the expression of hemagglutinin protein of influenza virus

The baculovirus expression vector system (BEVS) is a powerful tool in pharmaceutical biotechnology to infect insect cells and produce the recombinant proteins of interest. It has been well documented that optimizing the culture condition and its supplementation through designed experiments is critical for maximum protein production. In this study, besides physicochemical parameters including incubation temperature, cell count of infection, multiplicity of infection, and feeding percentage, potential supplementary factors such as cholesterol, polyamine, galactose, pluronic-F68, glucose, L-glutamine, and ZnSO4 were screened for Spodoptera frugiperda (Sf9) cell culture and expression of hemagglutinin (HA) protein of Influenza virus via Placket-Burman design and then optimized through Box-Behnken approach. The optimized conditions were then applied for scale-up culture and the expressed r-HA protein was characterized. Optimization of selected parameters via the Box-Behnken approach indicated that feed percentage, cell count, and multiplicity of infection are the main parameters affecting r-HA expression level and potency compared to the previously established culture condition. This study demonstrated the effectiveness of designing experiments to select and optimize important parameters that potentially affect Sf9 cell culture, r-HA expression, and its potency in the BEVS system.

Hybrid machine learning model based predictions for properties of poly(2-hydroxyethyl methacrylate)-poly(vinyl alcohol) composite cryogels embedded with bacterial cellulose

Supermacroporous composite cryogels with enhanced adjustable functionality have received extensive interest in bioseparation, tissue engineering, and drug delivery. However, the variations in their components significantly impactfinal properties. This study presents a two-step hybrid machine learning approach for predicting the properties of innovative poly(2-hydroxyethyl methacrylate)-poly(vinyl alcohol) composite cryogels embedded with bacterial cellulose (pHEMA-PVA-BC) based on their compositions. By considering the ratios of HEMA (1.0-22.0 wt%), PVA (0.2-4.0 wt%), poly(ethylene glycol) diacrylate (1.0-4.5 wt%), BC (0.1-1.5 wt%), and water (68.0-96.0 wt%) as investigational variables, overlay sampling uniform design (OSUD) was employed to construct a high-quality dataset for model development. The random forest (RF) model was used to classify the preparation conditions. Then four models of artificial neural network, RF, gradient boosted regression trees (GBRT), and XGBoost were developed to predict the basic properties of the composite cryogels. The results showed that the RF model achieved an accurate three-class classification of preparation conditions. Among the four models, the GBRT model exhibited the best predictive performance of the basic properties, with the mean absolute percentage error of 16.04 %, 0.85 %, and 2.44 % for permeability, effective porosity, and height of theoretical plate (1.0 cm/min), respectively. Characterization results of the representative pHEMA-PVA-BC composite cryogel showed an effective porosity of 81.01 %, a permeability of 1.20 × 10-12 m2, and a range of height of theoretical plate between 0.40-0.49 cm at flow velocities of 0.5-3.0 cm/min. These indicate that the pHEMA-PVA-BC cryogel was an excellent material with supermacropores, low flow resistance and high mass transfer efficiency. Furthermore, the model output demonstrates that the alteration of the proportions of PVA (0.2-3.5 wt%) and BC (0.1-1.5 wt%) components in composite cryogels resulted in significant changes in the material basic properties. This work represents an attempt to efficiently design and prepare target composite cryogels using machine learning and providing valuable insights for the efficient development of polymers.

LowTempGAL: a highly responsive low temperature-inducible GAL system in Saccharomyces cerevisiae

Temperature is an important control factor for biologics biomanufacturing in precision fermentation. Here, we explored a highly responsive low temperature-inducible genetic system (LowTempGAL) in the model yeast Saccharomyces cerevisiae. Two temperature biosensors, a heat-inducible degron and a heat-inducible protein aggregation domain, were used to regulate the GAL activator Gal4p, rendering the leaky LowTempGAL systems. Boolean-type induction was achieved by implementing a second-layer control through low-temperature-mediated repression on GAL repressor gene GAL80, but suffered delayed response to low-temperature triggers and a weak response at 30°C. Application potentials were validated for protein and small molecule production. Proteomics analysis suggested that residual Gal80p and Gal4p insufficiency caused suboptimal induction. 'Turbo' mechanisms were engineered through incorporating a basal Gal4p expression and a galactose-independent Gal80p-supressing Gal3p mutant (Gal3Cp). Varying Gal3Cp configurations, we deployed the LowTempGAL systems capable for a rapid stringent high-level induction upon the shift from a high temperature (37-33°C) to a low temperature (≤30°C). Overall, we present a synthetic biology procedure that leverages 'leaky' biosensors to deploy highly responsive Boolean-type genetic circuits. The key lies in optimisation of the intricate layout of the multi-factor system. The LowTempGAL systems may be applicable in non-conventional yeast platforms for precision biomanufacturing.

Rapid total sialic acid monitoring during cell culture process using a machine learning model based soft sensor

Total sialic acid content (TSA) in biotherapeutic proteins is often a critical quality attribute as it impacts the drug efficacy. Traditional wet chemical assays to quantify TSA in biotherapeutic proteins during cell culture typically takes several hours or longer due to the complexity of the assay which involves isolation of sialic acid from the protein of interest, followed by sample preparation and chromatographic based separation for analysis. Here, we developed a machine learning model-based technology to rapidly predict TSA during cell culture by using typically measured process parameters. The technology features a user interface, where the users only have to upload cell culture process parameters as input variables and TSA values are instantly displayed on a dashboard platform based on the model predictions. In this study, multiple machine learning algorithms were assessed on our dataset, with the Random Forest model being identified as the most promising model. Feature importance analysis from the Random Forest model revealed that attributes like viable cell density (VCD), glutamate, ammonium, phosphate, and basal medium type are critical for predictions. Notably, while the model demonstrated strong predictability by Day 14 of observation, challenges remain in forecasting TSA values at the edges of the calibration range. This research not only emphasizes the transformative power of machine learning and soft sensors in bioprocessing but also introduces a rapid and efficient tool for sialic acid prediction, signaling significant advancements in bioprocessing. Future endeavors may focus on data augmentation to further enhance model precision and exploration of process control capabilities.

The Development and Opportunities of Predictive Biotechnology

Recent advances in bioeconomy allow a holistic view of existing and new process chains and enable novel production routines continuously advanced by academia and industry. All this progress benefits from a growing number of prediction tools that have found their way into the field. For example, automated genome annotations, tools for building model structures of proteins, and structural protein prediction methods such as AlphaFold2TM or RoseTTAFold have gained popularity in recent years. Recently, it has become apparent that more and more AI-based tools are being developed and used for biocatalysis and biotechnology. This is an excellent opportunity for academia and industry to accelerate advancements in the field further. Biotechnology, as a rapidly growing interdisciplinary field, stands to benefit greatly from these developments.

A review on microbes mediated resource recovery and bioplastic (polyhydroxyalkanoates) production from wastewater

BACKGROUND

Plastic is widely utilized in packaging, frameworks, and as coverings material. Its overconsumption and slow degradation, pose threats to ecosystems due to its toxic effects. While polyhydroxyalkanoates (PHA) offer a sustainable alternative to petroleum-based plastics, their production costs present significant obstacles to global adoption. On the other side, a multitude of household and industrial activities generate substantial volumes of wastewater containing both organic and inorganic contaminants. This not only poses a threat to ecosystems but also presents opportunities to get benefits from the circular economy. Production of bioplastics may be improved by using the nutrients and minerals in wastewater as a feedstock for microbial fermentation. Strategies like feast-famine culture, mixed-consortia culture, and integrated processes have been developed for PHA production from highly polluted wastewater with high organic loads. Various process parameters like organic loading rate, organic content (volatile fatty acids), dissolved oxygen, operating pH, and temperature also have critical roles in PHA accumulation in microbial biomass. Research advances are also going on in downstream and recovery of PHA utilizing a combination of physical and chemical (halogenated solvents, surfactants, green solvents) methods. This review highlights recent developments in upcycling wastewater resources into PHA, encompassing various production strategies, downstream processing methodologies, and techno-economic analyses.

SHORT CONCLUSION

Organic carbon and nitrogen present in wastewater offer a promising, cost-effective source for producing bioplastic. Previous attempts have focused on enhancing productivity through optimizing culture systems and growth conditions. However, despite technological progress, significant challenges persist, such as low productivity, intricate downstream processing, scalability issues, and the properties of resulting PHA.

A machine learning-based approach for improving plasmid DNA production in Escherichia coli fed-batch fermentations

Artificial Intelligence (AI) technology is spearheading a new industrial revolution, which provides ample opportunities for the transformational development of traditional fermentation processes. During plasmid fermentation, traditional subjective process control leads to highly unstable plasmid yields. In this study, a multi-parameter correlation analysis was first performed to discover a dynamic metabolic balance among the oxygen uptake rate, temperature, and plasmid yield, whilst revealing the heating rate and timing as the most important optimization factor for balanced cell growth and plasmid production. Then, based on the acquired on-line parameters as well as outputs of kinetic models constructed for describing process dynamics of biomass concentration, plasmid yield, and substrate concentration, a machine learning (ML) model with Random Forest (RF) as the best machine learning algorithm was established to predict the optimal heating strategy. Finally, the highest plasmid yield and specific productivity of 1167.74 mg L-1 and 8.87 mg L-1/OD600 were achieved with the optimal heating strategy predicted by the RF model in the 50 L bioreactor, respectively, which was 71% and 21% higher than those obtained in the control cultures where a traditional one-step temperature upshift strategy was applied. In addition, this study transformed empirical fermentation process optimization into a more efficient and rational self-optimization method. The methodology employed in this study is equally applicable to predict the regulation of process dynamics for other products, thereby facilitating the potential for furthering the intelligent automation of fermentation processes.

Deep hybrid modeling of a HEK293 process: Combining long short-term memory networks with first principles equations

The combination of physical equations with deep learning is becoming a promising methodology for bioprocess digitalization. In this paper, we investigate for the first time the combination of long short-term memory (LSTM) networks with first principles equations in a hybrid workflow to describe human embryonic kidney 293 (HEK293) culture dynamics. Experimental data of 27 extracellular state variables in 20 fed-batch HEK293 cultures were collected in a parallel high throughput 250 mL cultivation system in an industrial process development setting. The adaptive moment estimation method with stochastic regularization and cross-validation were employed for deep learning. A total of 784 hybrid models with varying deep neural network architectures, depths, layers sizes and node activation functions were compared. In most scenarios, hybrid LSTM models outperformed classical hybrid Feedforward Neural Network (FFNN) models in terms of training and testing error. Hybrid LSTM models revealed to be less sensitive to data resampling than FFNN hybrid models. As disadvantages, Hybrid LSTM models are in general more complex (higher number of parameters) and have a higher computation cost than FFNN hybrid models. The hybrid model with the highest prediction accuracy consisted in a LSTM network with seven internal states connected in series with dynamic material balance equations. This hybrid model correctly predicted the dynamics of the 27 state variables (R2 = 0.93 in the test data set), including biomass, key substrates, amino acids and metabolic by-products for around 10 cultivation days.

Pushing the boundaries of phosphorylase cascade reaction for cellobiose production I: Kinetic model development

One-pot cascade reactions of coupled disaccharide phosphorylases enable an efficient transglycosylation via intermediary α-d-glucose 1-phosphate (G1P). Such transformations have promising applications in the production of carbohydrate commodities, including the disaccharide cellobiose for food and feed use. Several studies have shown sucrose and cellobiose phosphorylase for cellobiose synthesis from sucrose, but the boundaries on transformation efficiency that result from kinetic and thermodynamic characteristics of the individual enzyme reactions are not known. Here, we assessed in a step-by-step systematic fashion the practical requirements of a kinetic model to describe cellobiose production at industrially relevant substrate concentrations of up to 600 mM sucrose and glucose each. Mechanistic initial-rate models of the two-substrate reactions of sucrose phosphorylase (sucrose + phosphate → G1P + fructose) and cellobiose phosphorylase (G1P + glucose → cellobiose + phosphate) were needed and additionally required expansion by terms of glucose inhibition, in particular a distinctive two-site glucose substrate inhibition of the cellobiose phosphorylase (from Cellulumonas uda). Combined with mass action terms accounting for the approach to equilibrium, the kinetic model gave an excellent fit and a robust prediction of the full reaction time courses for a wide range of enzyme activities as well as substrate concentrations, including the variable substoichiometric concentration of phosphate. The model thus provides the essential engineering tool to disentangle the highly interrelated factors of conversion efficiency in the coupled enzyme reaction; and it establishes the necessary basis of window of operation calculations for targeted optimizations toward different process tasks.

Pushing the boundaries of phosphorylase cascade reaction for cellobiose production II: Model-based multiobjective optimization

The inherent complexity of coupled biocatalytic reactions presents a major challenge for process development with one-pot multienzyme cascade transformations. Kinetic models are powerful engineering tools to guide the optimization of cascade reactions towards a performance suitable for scale up to an actual production. Here, we report kinetic model-based window of operation analysis for cellobiose production (≥100 g/L) from sucrose and glucose by indirect transglycosylation via glucose 1-phosphate as intermediate. The two-step cascade transformation is catalyzed by sucrose and cellobiose phosphorylase in the presence of substoichiometric amounts of phosphate (≤27 mol% of substrate). Kinetic modeling was instrumental to uncover the hidden effect of bulk microviscosity due to high sugar concentrations on decreasing the rate of cellobiose phosphorylase specifically. The mechanistic-empirical hybrid model thus developed gives a comprehensive description of the cascade reaction at industrially relevant substrate conditions. Model simulations serve to unravel opposed relationships between efficient utilization of the enzymes and maximized concentration (or yield) of the product within a given process time, in dependence of the initial concentrations of substrate and phosphate used. Optimum balance of these competing key metrics of process performance is suggested from the model-calculated window of operation and is verified experimentally. The evidence shown highlights the important use of kinetic modeling for the characterization and optimization of cascade reactions in ways that appear to be inaccessible to purely data-driven approaches.

Current challenges and recent advances on the path towards continuous biomanufacturing

Continuous biopharmaceutical manufacturing is currently a field of intense research due to its potential to make the entire production process more optimal for the modern, ever-evolving biopharmaceutical market. Compared to traditional batch manufacturing, continuous bioprocessing is more efficient, adjustable, and sustainable and has reduced capital costs. However, despite its clear advantages, continuous bioprocessing is yet to be widely adopted in commercial manufacturing. This article provides an overview of the technological roadblocks for extensive adoptions and points out the recent advances that could help overcome them. In total, three key areas for improvement are identified: Quality by Design (QbD) implementation, integration of upstream and downstream technologies, and data and knowledge management. First, the challenges to QbD implementation are explored. Specifically, process control, process analytical technology (PAT), critical process parameter (CPP) identification, and mathematical models for bioprocess control and design are recognized as crucial for successful QbD realizations. Next, the difficulties of end-to-end process integration are examined, with a particular emphasis on downstream processing. Finally, the problem of data and knowledge management and its potential solutions are outlined where ontologies and data standards are pointed out as key drivers of progress.

Advancing 2D fluorescence online monitoring in microtiter plates by separating scattered light and fluorescence measurement, using a tunable emission monochromator

Online fluorescence monitoring has become a key technology in modern bioprocess development, as it provides in-depth process knowledge at comparably low costs. In particular, the technology is widely established for high-throughput microbioreactor cultivation systems, due to its noninvasive character. For microtiter plates, previously also multi-wavelength 2D fluorescence monitoring was developed. To overcome an observed limitation of fluorescence sensitivity, this study presents a modified spectroscopic setup, including a tunable emission monochromator. The new optical component enables the separation of the scattered and fluorescent light measurements, which allows for the adjustment of integration times of the charge-coupled device detector. The resulting increased fluorescence sensitivity positively affected the performance of principal component analysis for spectral data of Escherichia coli batch cultivation experiments with varying sorbitol concentration supplementation. In direct comparison with spectral data recorded at short integration times, more biologically consistent signal dynamics were calculated. Furthermore, during partial least square regression for E. coli cultivation experiments with varying glucose concentrations, improved modeling performance was observed. Especially, for the growth-uncoupled acetate concentration, a considerable improvement of the root-mean-square error from 0.25 to 0.17 g/L was achieved. In conclusion, the modified setup represents another important step in advancing 2D fluorescence monitoring in microtiter plates.

High-throughput process development from gene cloning to protein production

In the post-genomic era, the demand for faster and more efficient protein production has increased, both in public laboratories and industry. In addition, with the expansion of protein sequences in databases, the range of possible enzymes of interest for a given application is also increasing. Faced with peer competition, budgetary, and time constraints, companies and laboratories must find ways to develop a robust manufacturing process for recombinant protein production. In this review, we explore high-throughput technologies for recombinant protein expression and present a holistic high-throughput process development strategy that spans from genes to proteins. We discuss the challenges that come with this task, the limitations of previous studies, and future research directions.

Hybrid deep modeling of a CHO-K1 fed-batch process: combining first-principles with deep neural networks

Introduction: Hybrid modeling combining First-Principles with machine learning is becoming a pivotal methodology for Biopharma 4.0 enactment. Chinese Hamster Ovary (CHO) cells, being the workhorse for industrial glycoproteins production, have been the object of several hybrid modeling studies. Most previous studies pursued a shallow hybrid modeling approach based on three-layered Feedforward Neural Networks (FFNNs) combined with macroscopic material balance equations. Only recently, the hybrid modeling field is incorporating deep learning into its framework with significant gains in descriptive and predictive power. Methods: This study compares, for the first time, deep and shallow hybrid modeling in a CHO process development context. Data of 24 fed-batch cultivations of a CHO-K1 cell line expressing a target glycoprotein, comprising 30 measured state variables over time, were used to compare both methodologies. Hybrid models with varying FFNN depths (3-5 layers) were systematically compared using two training methodologies. The classical training is based on the Levenberg-Marquardt algorithm, indirect sensitivity equations and cross-validation. The deep learning is based on the Adaptive Moment Estimation Method (ADAM), stochastic regularization and semidirect sensitivity equations. Results and conclusion: The results point to a systematic generalization improvement of deep hybrid models over shallow hybrid models. Overall, the training and testing errors decreased by 14.0% and 23.6% respectively when applying the deep methodology. The Central Processing Unit (CPU) time for training the deep hybrid model increased by 31.6% mainly due to the higher FFNN complexity. The final deep hybrid model is shown to predict the dynamics of the 30 state variables within the error bounds in every test experiment. Notably, the deep hybrid model could predict the metabolic shifts in key metabolites (e.g., lactate, ammonium, glutamine and glutamate) in the test experiments. We expect deep hybrid modeling to accelerate the deployment of high-fidelity digital twins in the biopharma sector in the near future.

Data-driven prediction models for forecasting multistep ahead profiles of mammalian cell culture toward bioprocess digital twins

Recently, the advancement in process analytical technology and artificial intelligence (AI) has enabled the generation of enormous culture data sets from biomanufacturing processes that produce various recombinant therapeutic proteins (RTPs), such as monoclonal antibodies (mAbs). Thus, now it is very important to exploit them for the enhanced reliability, efficiency, and consistency of the RTP-producing culture processes and for the reduced incipient or abrupt faults. It is achievable by AI-based data-driven models (DDMs), which allow us to correlate biological and process conditions and cell culture states. In this work, we provide practical guidelines for choosing the best combination of model elements to design and implement successful DDMs for given hypothetical in-line data sets during mAb-producing Chinese hamster ovary cell culture, as such enabling us to forecast dynamic behaviors of culture performance such as viable cell density, mAb titer as well as glucose, lactate and ammonia concentrations. To do so, we created DDMs that balance computational load with model accuracy and reliability by identifying the best combination of multistep ahead forecasting strategies, input features, and AI algorithms, which is potentially applicable to implementation of interactive DDM within bioprocess digital twins. We believe this systematic study can help bioprocess engineers start developing predictive DDMs with their own data sets and learn how their cell cultures behave in near future, thereby rendering proactive decision possible.

From genotype to phenotype: computational approaches for inferring microbial traits relevant to the food industry

When selecting microbial strains for the production of fermented foods, various microbial phenotypes need to be taken into account to achieve target product characteristics, such as biosafety, flavor, texture, and health-promoting effects. Through continuous advances in sequencing technologies, microbial whole-genome sequences of increasing quality can now be obtained both cheaper and faster, which increases the relevance of genome-based characterization of microbial phenotypes. Prediction of microbial phenotypes from genome sequences makes it possible to quickly screen large strain collections in silico to identify candidates with desirable traits. Several microbial phenotypes relevant to the production of fermented foods can be predicted using knowledge-based approaches, leveraging our existing understanding of the genetic and molecular mechanisms underlying those phenotypes. In the absence of this knowledge, data-driven approaches can be applied to estimate genotype-phenotype relationships based on large experimental datasets. Here, we review computational methods that implement knowledge- and data-driven approaches for phenotype prediction, as well as methods that combine elements from both approaches. Furthermore, we provide examples of how these methods have been applied in industrial biotechnology, with special focus on the fermented food industry.

Big data and machine learning driven bioprocessing - Recent trends and critical analysis

Given the potential of machine learning algorithms in revolutionizing the bioengineering field, this paper examined and summarized the literature related to artificial intelligence (AI) in the bioprocessing field. Natural language processing (NLP) was employed to explore the direction of the research domain. All the papers from 2013 to 2022 with specific keywords of bioprocessing using AI were extracted from Scopus and grouped into two five-year periods of 2013-to-2017 and 2018-to-2022, where the past and recent research directions were compared. Based on this procedure, selected sample papers from recent five years were subjected to further review and analysis. The result shows that 50% of the publications in the past five-year focused on topics related to hybrid models, ANN, biopharmaceutical manufacturing, and biorefinery. The summarization and analysis of the outcome indicated that implementing AI could improve the design and process engineering strategies in bioprocessing fields.

Machine Learning Models Using Data Mining for Biomass Production from Yarrowia lipolytica Fermentation

In this paper, a database of biomass production from Yarrowia lipolytica fermentation is prepared and constructed using machine learning and data mining approaches. The database is curated from 15 publications and consists of 301 rows of data with 25 predictors and 1 label. The predictors include inoculum size, temperature, pH, and time, while the label is the corresponding biomass production. The database is then divided into training, validation, and test datasets and analyzed as a supervised machine learning task for regression. Twenty-six regression models are employed and compared for their performance in predicting biomass production. The best-performing model is the Matern 5/2 Gaussian process regression model, which has the lowest root-mean-squared error of 0.75 g/L, the highest R squared of 0.90, and the lowest mean absolute error of 0.52 g/L. The t-test is used to identify the most important predictors, and 14 predictors are sufficient for creating an accurate model. These 14 predictors are fermentation time, peptone, temperature, total Kjeldahl nitrogen, shaking rate, total nitrogen, inoculum size, yeast extract, crude glycerol, glucose, oil and grease, media pH, ammonium sulfate, and olive oil. This research demonstrates the application of machine learning and data mining to estimate biomass production and gives insight into which parameters are essential for Yarrowia lipolytica fermentation.

Bayesian Optimization for an ATP-Regenerating In Vitro Enzyme Cascade

Enzyme cascades are an emerging synthetic tool for the synthesis of various molecules, combining the advantages of biocatalysis and of one-pot multi-step reactions. However, the more complex the enzyme cascade is, the more difficult it is to achieve adequate productivities and product concentrations. Therefore, the whole process must be optimized to account for synergistic effects. One way to deal with this challenge involves data-driven models in combination with experimental validation. Here, Bayesian optimization was applied to an ATP-producing and -regenerating enzyme cascade consisting of polyphosphate kinases. The enzyme and co-substrate concentrations were adjusted for an ATP-dependent reaction, catalyzed by mevalonate kinase (MVK). With a total of 16 experiments, we were able to iteratively optimize the initial concentrations of the components used in the one-pot synthesis to improve the specific activity of MVK with 10.2 U mg−1. The specific activity even exceeded the results of the reference reaction with stoichiometrically added ATP amounts, with which a specific activity of 8.8 U mg−1 was reached. At the same time, the product concentrations were also improved so that complete yields were achieved.