See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/382166201

Computer-Aided Chemical Engineering Research Advances in Precision

Fermentation

Article in Current Opinion in Food Science · July 2024

DOI: 10.1016/j.cofs.2024.101196

CITATIONS READS

19 23

4 authors:

Tom Vinestock Michael Short

King's College London University of Surrey
7 PUBLICATIONS 29 CITATIONS 96 PUBLICATIONS 727 CITATIONS

SEE PROFILE SEE PROFILE

Keeran Ward Miao Guo

University of Leeds King's College London
36 PUBLICATIONS 694 CITATIONS 132 PUBLICATIONS 3,551 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Tom Vinestock on 30 July 2025.

The user has requested enhancement of the downloaded file.

 Available online at www.sciencedirect.com

Review
Computer-aided chemical engineering research
advances in precision fermentation
Tom Vinestock 1, Michael Short 2, Keeran Ward 3 and Miao Guo 1 ]]
]]]]]]
]]

Precision fermentation is a promising food production as a substrate, is most commonly a carbohydrate, such as
technology that uses micro-organisms to produce specific glucose, although simpler substrates such as methane can
proteins, fats, and vitamins, offering a more sustainable also be used [3]. While fermentation has a long history,
alternative to animal agriculture. This review explores recent with evidence for wine-making stretching back at least
advances in computer-aided chemical engineering research 7000 years [4], this review will focus on modern, precision
within precision fermentation, focusing on process systems fermentation, rather than traditional fermentation or bio­
engineering (PSE), process control, and artificial intelligence. mass fermentation.
PSE offers important process synthesis and process
optimisation tools for fermentation, helping evaluate Precision fermentation systems are biologically optimised
environmental impacts and economic feasibility during design. to produce specific high-value biomolecules, such as
Advanced control strategies, such as soft sensors, can improve proteins, vitamins, enzymes, natural pigments, and fats
productivity and yield. Artificial intelligence methods, such as using naturally occurring or genetically modified organ­
surrogate modelling, enable rapid experimentation, process isms as microbial ‘factories’ [5]. Because of this, precision
optimisation, and scale-up, accelerating development. These fermentation lends itself to the production of food com­
advances pave the way for precision fermentation to play a ponents that mimic those obtained in traditional animal
greater role in the food production system of the future. agriculture, such as animal fats and proteins [6]. It can be
distinguished from biomass fermentation as the desired
Addresses
1
product is a metabolite, rather than the microbial biomass
Department of Engineering, King’s College London, Strand, itself. Interest in precision fermentation, and the related
London, UK
2
School of Chemistry & Chemical Engineering, University of Surrey,
term ‘cellular agriculture’ has increased exponentially in
Stag Hill, Guildford, UK recent years. This can be seen as being driven by the
3
School of Chemical & Process Engineering, University of Leeds, application of methods developed in the pharmaceutical
Woodhouse, Leeds, UK industry to optimise fermentation-based drug production,
such as metabolic engineering, to the food industry. Ex­
Corresponding author: Guo, Miao ([email protected])
amples of products currently produced by precision fer­
mentation include rennet, an enzyme used in cheese-
Current Opinion in Food Science 2024, 58:101196 making; whey and casein proteins, components of milk;
This review comes from a themed issue on Innovations in Food and soy leghemoglobin, a haem analogue used to improve
Science the taste of plant-based burgers [7].
Edited by Davide Giacalone
Relative to conventional methods of food production,
For complete overview of the section, please refer to the article
collection, “Innovations in Food Science 2024” fermentation-derived food has several advantages as a
source of dietary fats and proteins, particularly when
Available online 10 July 2024
compared with animal agriculture. On a per-unit protein
https://doi.org/10.1016/j.cofs.2024.101196 basis, relative to meat, eggs, and dairy, fermentation-de­
2214–7993/© 2024 The Author(s). Published by Elsevier Ltd. This is rived protein has significantly lower carbon and nitrogen
an open access article under the CC BY license (http:// emissions [8], reduced land use [9], and lower water use
creativecommons.org/licenses/by/4.0/).
[7], largely as a result of microbial fermentation being
more efficient than livestock at converting carbohydrate-
based feedstocks into protein. Fermentation also has ad­
vantages in terms of food and health security by decou­
pling food production from animal husbandry. By
avoiding use of livestock, fermentation-based protein
Introduction production reduces the risk zoonotic infections pose to
Fermentation can be defined as the use of micro-organ­ human health [10] and avoids the problems of antibiotic
isms to produce value-added products in the presence of pollution [11]. Additionally, fermentation can produce
an organic carbon source [1]. It is an important industrial vastly more protein per unit area than animal agriculture
process used in the production of food, biochemicals, and [9], potentially allowing small, densely populated areas to
pharmaceuticals [2]. The carbon source, often referred to be more self-sufficient in protein supply.

www.sciencedirect.com Current Opinion in Food Science 2024, 58:101196

2 Innovations in Food Science

An important distinction in the design, operation, and order to be cost competitive with conventionally pro­
control of fermentation processes is between batch, fed- duced alternatives [12]. Product development, including
batch, and continuous fermentations. In batch fermen­ selection and engineering of novel variants of target
tation, all the nutrients available to the micro-organisms molecules, is also important to allow fermentation-de­
during the course of the fermentation are present in the rived ingredients to compete with a wider set of con­
reactor vessel at the point of inoculation. The system is ventional foods. These challenges may be addressed in
partially closed, with no feeding or discharge [1]. In fed- many different ways, with cost reductions possible
batch operation, the substrate concentration in the re­ through improvements in strain development, feedstock
actor is controlled over the course of the batch fermen­ optimisation, and bioprocess design and operation.
tation by varying the rate at which a substrate solution is
added to the reactor. This avoids substrate inhibition of Currently, the majority of precision fermentations rely
growth and can improve productivity [1]. In continuous on a small number of legacy hosts, fed on a refined sugar
fermentation, the reactor is fed with a substrate solution, substrate, in a continually stirred tank reactor. However,
as in fed-batch, but product is also drained from the strain selection and development could result in higher
reactor on an ongoing basis. Precision fermentation can, titres and yields, greater robustness, and faster growth,
in theory, be run in any of these three configurations. making precision fermentation more cost-efficient.
However, research to date has focused on the operation Strain selection and improvement could be accelerated
and control of batch and fed-batch precision fermenta­ by combining computational modelling, design of ex­
tion, often to produce pharmaceutical products. periments, and high-throughput screening to efficiently
explore the design space. Equally, the use of alternative
Despite widespread adoption of precision fermentation substrates such as agricultural waste products and by-
to produce high-value products such as vitamins and products could significantly reduce feedstock costs, one
antibiotics, for precision fermentation to become a sig­ of the main costs of manufacture [7], but this relies on
nificant part of the food production system, several research and design tools to model the trade-offs of
challenges will need to be addressed. Process improve­ different host–substrate–product combinations. This
ments and scale-up are needed to enable fermentation to cross-disciplinary review provides an overview of recent
be cost competitive with conventional alternatives at research relevant to precision fermentation. Figure 1
scale; food products such as fats and proteins are much shows common keywords in the range of literature re­
less valuable, gram for gram, than pharmaceuticals and viewed. The perspectives of three different research
vitamins, leading to a greater need for cost-efficient disciplines are considered: process systems engineering
operation. Indeed, it has been estimated that food fer­ (PSE), process control, and artificial intelligence (AI).
mentations require production titres several orders of Some of the tools and methodologies associated with
magnitude higher than those for pharmaceuticals in each of these areas are shown in Figure 2. In each

Figure 1

Current Opinion in Food Science

Visualisation of keywords from cited works. The frequency of words and phrases in this corpus is indicated by their size. Created using wordclouds.
ethz.ch.

Current Opinion in Food Science 2024, 58:101196 www.sciencedirect.com

 Engineering Research in Precision Fermentation Vinestock et al. 3

Figure 2

Current Opinion in Food Science

Research methodologies/tools relevant to precision fermentation and associated research disciplines.

section, research frontiers are highlighted that can ulti­ optimisation of the operations required to produce effi­
mately enable a step change in precision fermentation in ciently at scale.
the food industry.
Sustainable design considering multiple objectives, such as
sustainability and profitability, is challenging, as it requires
Process systems engineering comprehensive assessment of the alternative biosynthetic
Bioprocess design and optimisation is key to both de­ pathways and the wider process systems. To address these
velopment of new fermentation products and improve­ challenges, rigorous methodologies that can embed sus­
ment of existing fermentation processes. Traditional tainability metrics into product and process design are
process synthesis and design provides an established set needed. Heuristic process synthesis (HPS) and process
of methodologies founded on heuristics driven by che­ optimisation (PO) are key strategies in conceptualising a
mical engineering expertise. These heuristics often se­ process flowsheet. HPS aims to find a pathway for con­
parate product design from process synthesis. Product verting raw materials into useful products, combining
design focuses on consumer demand, which in the multiple unit operations [13]. It defines the topology of the
context of precision fermentation may consist of identi­ process and outlines the design strategies that solve the
fying specific desirable biomolecules, such as proteins, desired objectives. While HPS is fast, simple, and provides
while process synthesis involves planning and good solutions to process design problems, it requires

www.sciencedirect.com Current Opinion in Food Science 2024, 58:101196

4 Innovations in Food Science

extensive heuristic rules that are sensitive to the emer­ lie ahead for new products and processes, including
gence of new technologies. On the other hand, PO is a precision fermentation. Process synthesis and optimisa­
more systematic tool to deliver optimal solutions to mul­ tion will be important in enabling the use of new sub­
tiobjective problems by using mathematical programming strates and alternative reactor designs, as well as the
in the process and product design [14]. systematic incorporation of economic and environmental
objectives into decision-making [20]. The current drive
Life cycle assessment (LCA) offers a useful tool to for greater sustainability and circularity in design has
quantify the environmental impacts of products and motivated the development of robust and adaptive PS
processes and highlight potential improvement spaces and PO methods that can optimise multiple objectives
for precision fermentation. LCA is now commonplace, throughout the process and product design life cycle
having been standardised in technical standards such as [21]. Furthermore, newer, more sustainable technologies
ISO14040/44 [15], and the systems thinking that it has are being developed that utilise greener supply chains,
brought about has led to reductions in the environmental renewable energy, and low-carbon feedstocks. However,
impacts of many processes. Previous studies have high­ these technologies have their own limitations [20], such
lighted that products produced via precision fermenta­ as increased complexity in design of unit operations and
tion result in environmental benefits due to the separations, low yield and conversion, and difficult
increased yields and increased efficiency particularly scaling-up considerations. The use of modelling tools
when the synergies between fermentation and other that employ both mechanistic [22] and data-driven
parts of the process are considered [16]. Additionally, methods [23] can significantly improve the process de­
new market mechanisms may be able to finance novel sign of these technologies.
fermentation processes that deliver significant carbon
reductions but would not otherwise be financially viable Fermentation process control
[17]; if LCA can be used to demonstrate that fermen­ In recent years, there has been significant innovation in
tation can produce environmentally preferable sub­ fermentation process control, with advances in sensing,
stitutes for foods such as meat, carbon credits could offer process modelling, and machine learning being used to
a way of monetising this advantage. improve productivity, sustainability, and efficiency.
While the bounds for many process operating parameters
To facilitate sustainable design and operation, LCA and are set by the reactor, process design, and the choice of
technoeconomic analysis [18] can be incorporated into host species, within these bounds improved control can
the process optimisation. More sophisticated algorithms result in increased output, higher titres, and improved
have increased the speed and robustness of PO, enabling yields at lower capital costs when compared with mod­
solutions to process design that simultaneously solve ifications to the reactor design.
both the process flowsheet and the operating conditions.
However, PO does come with shortcomings such as The dynamics of fermentation control depend on whe­
oversimplification of complex process designs and high ther the fermentation is run in batch, fed-batch, or
computational cost. Additionally, optimal solutions can continuous operation. Table 1 gives an overview of
only be found if the design space allows for an optimal common state and manipulated variables in each op­
pathway. In some cases, heuristic and mathematical ap­ erational mode, partially based on a literature review
proaches can be hybridised, yielding a reduced process conducted by Chai et al. [1]. Batch operation has tradi­
design space that can be optimised more easily [19]. The tionally been favoured by the pharmaceutical industry,
process can be split into two phases, the first providing as this minimises the risk of contamination. It typically
early stage elimination of unacceptable process designs also has a smaller number of manipulated variables.
and the second solving for an optimal design of the re­ However, batch productivity is much lower than for
sulting superstructure. continuous fermentation, as the time-average biomass
concentration will be much lower in the batch fermen­
While PO and HPS methods have been used extensively tation. In continuous fermentation, the biomass con­
in chemical engineering for decades, several challenges centration can be maintained at the productivity-

Table 1
Common state and manipulated variables in fermentation based partially on works reviewed by Ref. [1]. T: temperature; S: substrate
concentration; X: biomass concentration; P: product concentration; DO: dissolved oxygen concentration; N: nitrogen concentration; V:
broth volume; EtOH: ethanol concentration.

Type Common state variables Common manipulated variables

Batch T, pH, DO Cooling water flow rate acid flow rate base flow rate oxygen-source flow rate
Fed-batch T, pH, S, X, P, DO, N, V, EtOH As above + substrate feed rate nitrogen-source feed rate stir rate
Continuous As above As above + outlet flow rate

Current Opinion in Food Science 2024, 58:101196 www.sciencedirect.com

 Engineering Research in Precision Fermentation Vinestock et al. 5

maximising level. Indeed, Anand and Srivastava [24] hybrid predictive model for fed-batch fermentation
found the productivity of the continuous fermentation of using a neural network trained on plant data to predict
mycophenolic acid was more than four times higher than critical time-varying parameters in a complex kinetic
the productivity of the batch fermentation. Fed-batch model based on the current state and control action. The
offers a compromise, with some additional complexity hybrid model has better predictive accuracy than the
resulting in higher productivity than is possible for fixed mechanistic model and, when used as the model in
batch, while still lower than that obtained in continuous MPC, is found to result in higher product concentra­
fermentation. tions. A similar hybrid predictive approach was devel­
oped by Winz et al. [33] to describe the growth and
Model predictive control (MPC) combines a model and sporulation of B. subtilis, with a neural network trained to
mathematical optimisation to control a plant. It is based predict the specific growth rate as a function of cell
on a receding-time horizon and aims to find the optimum concentration and temperature, with biological knowl­
sequence of control actions to maximise the objective edge incorporated to this function using a penalty term.
function over the chosen time window. While the This model is found to describe the process dynamics
method itself is not new, having originally been devel­ sufficiently well, fitting experimental data well. Time
oped in the 1970s, research on its application to the series–based machine learning methods have also been
control of bioprocesses is ongoing. MPC can either be applied in fermentation control, such as by Wang et al.
used to track a reference or to calculate its own reference [37], who used a recurrent neural network to accurately
based on the cost function and constraints of the pro­ predict glucose and ethanol concentrations from elec­
blem. Jorgensen and Petersen successfully applied MPC tronic nose signals in an ethanol fermentation.
to set the inlet and outlet flow rates in a single-cell
protein fermentation to maximise the value of the Another area of advance is in fermentation sensors and
output [25]. More recently, Wang et al. [26] employed a soft sensors. Increasingly, the line between online and
nonlinear MPC, in combination with a support vector offline sampling is being blurred, with chemical analysis
machine predictor, to control product concentration in a techniques such as High-performance liquid chromato­
lysine fermentation. graphy, Raman spectroscopy, and infrared spectroscopy
used to estimate product and substrate concentrations
Reinforcement learning (RL) is emerging as a control for process control purposes, as well as quality control
approach in fermentation. While RL is, in theory, a [1]. Another emerging alternative to estimate these
model-free technique, often a model-based emulator concentrations is to use soft sensors [2] based on an ac­
will be used to train the RL controller. After this initial curate process model that may be mechanistic, data
training, the controller will be applied to a plant. Li et al. based, or hybrid.
[27] and Pandian et al. [28] explored the application of
RL to computational models of batch and fed-batch Another control approach looks at deriving insights from
fermentations, while other authors have examined the metabolic models. Dynamic flux balance analysis
use of RL to control microbial cocultures [29]. Oh et al. (dFBA) combines a flux balance model that describes
[30] combine RL and MPC to control the substrate feed known intracellular reaction pathways with substrate
rate in a fed-batch fermentation of penicillin, finding uptake kinetics and extracellular mass balances on sub­
that the hybrid outperformed various RL strategies. strates and products [34]. A linear programming model is
Panjapornpon et al. [31] investigated the use of a deep then used to determine the fluxes associated with each
deterministic policy gradient algorithm for pH-con­ reaction. This can be combined with MPC to optimise
trolled processes, including fermentation, but this was fermentation control in real time. Within precision fer­
limited to the control of pH and liquid level. Future mentation, methods combining MPC and dFBA have
work will likely widen the number of variables con­ been demonstrated in the control of fed-batch fermen­
trolled using RL methods, although this is complicated tations of Saccharomyces cerevisiae to produce ethanol [34]
by the curse of dimensionality, which makes learning and Chinese Hamster Ovary cells to produce antibodies
much harder in higher dimensional decision spaces. [35], with both studies finding improved production ti­
tres relative to simpler comparison controllers. An over­
Increasing application of machine learning methods, view of the fermentation process control research
particularly neural networks, to control is another clear reviewed here is provided in Table 2.
trend. Khaleghi et al. [36] provide an excellent review of
machine learning methods for fermentation optimisation
and control. They identify synergies and challenges in Artificial intelligence in precision fermentation
implementing machine learning in combination with The hype generated around AI and digital twins has led
mechanistic modelling for fermentation processes, to renewed industrial interest in use modelling and data-
highlighting the potential for these hybrid methods to driven insights to support modelling new technologies,
improve predictability. Shah et al. [32] developed a pathway selection and bio-process design. Precision

www.sciencedirect.com Current Opinion in Food Science 2024, 58:101196

6 Innovations in Food Science

fermentation has been an early adopter of machine

Summary: recent research on fermentation process control. Variables — V: volume; t: time; S: substrate concentration; Sin: substrate feed concentration; X: biomass concentration; P:
product concentration; T: temperature; DO: dissolved oxygen concentration; I: intermediate product concentration; QN: nitrogen-source feed rate; Q: substrate feed rate; Qout: outlet
flow rate; QO2 : oxygen-source feed rate; QC: catalyst feed rate; QBase: base feed rate; ω: stir rate. Methods — NL-MPC: non-linear model predictive control; SVM: support vector

[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
Ref
learning and advanced statistical methods, with meta­
bolic engineering requiring sequencing, high-

Precision fermentation of penicillin

throughput screening, and ‘omics’ to optimise metabolic
Precision fermentation of lysine
Precision fermentation of lysine
pathways, strains, and yields [5].

Precision fermentation
Biomass fermentation

Machine learning techniques such as Gaussian process

Ethanol fermentation
Antibody production
Microbial coculture
Yeast fermentation
regression, support vector machines, and neural net­

Spore production
works have all been shown to be effective for modelling
pH treatment complex relationships between experimental parameters
Process

and performance [38]. To screen for experimental de­

signs, Pensupa et al. [39] developed machine learning
models by data mining existing literature to find the key
experimental parameters for maximising biomass con­
Manipulated variables

centration in a fermentation of Yarrowia lipolytica, with

Gaussian processes providing the most accurate pre­
dictive capability. Similarly, Packiam et al. [40] devel­
Q, Qout, Sin

oped PERISCOPE-Opt for predicting optimal

Q, QBase
QN , QO2,
Q, Qout

QA, QB

QC , Q S2

Q, DO

fermentation conditions using XGBoost, random forest,

and support vector machines, demonstrating the poten­
Q
Q

tial for these approaches.

Xvegetative, Xspore, S, Xunstable_spore
machine; RL: reinforcement learning; ANN: artificial neural network; DDPG: deep deterministic policy gradient.

Another recent trend is use of AI techniques to speed up

the solving of physics-based models [41]. By using ma­
chine learning and data-driven modelling to obtain faster
X, S1, S2, P, I, DO, V

input–output relations, ML-assisted CFD makes it

State variables

possible to simulate larger models by using computa­

pHout, pHin, V
T, pH, DO, P

Glucose flux
X, S, P, V, t

tional resources more efficiently [42]. These approaches

X, S, P, V

X, S, P, V

have yet to see widespread use in the fermentation

X, S, V

XA, XB

community but could be helpful in addressing current

X, S

challenges around scale-up, in which gas transfer is often

a key problem.
ANN and polynomial regression for parameter estimation

With the rise in importance of renewable feedstocks,

there has been increased interest in low-carbon feed­
stocks and waste valorisation to create circular processes.
NL-MPC w/ metabolic flux balance analysis
NL-MPC w/ metabolic flux balance analysis

A core challenge of modelling these systems is the

complexity of the biological system [43] and the com­
MPC w/ ANN parameter estimation

plexity and variability of the feedstock composition [44].

New approaches, tools, and techniques are required to
address this, as often the traditional approaches are in­
accurate or inapplicable. Data-driven approaches to
modelling, improved experimental design via Bayesian
Economic NL-MPC
NL-MPC w/ SVM

optimisation, and application of tools from synthetic

RL-MPC hybrid
RL-ANN hybrid

biology were suggested to hold the key to unlocking

RL (DDPG)

higher productivity from such systems [45].

Method

In the last 5 years, the major trend in engineering

RL

RL

modelling research has been development of predictive

Fed-batch & continuous

surrogate models, for complex processes that are either

poorly understood with large amounts of experimental
data or are computationally intensive to simulate me­
chanistically. Embedding surrogate models into larger
Continuous
continuous

optimisation problems is now commonplace [46], in­

Fed-batch
Fed-batch
Fed-batch

Fed-batch

Fed-batch
Fed-batch
Table 2

cluding for the modelling of complex physical processes

Batch
Batch
Mode

[47]. A major challenge in this area is the accuracy of

such models, particularly when there is limited data

Current Opinion in Food Science 2024, 58:101196 www.sciencedirect.com

 Engineering Research in Precision Fermentation Vinestock et al. 7

availability, or when extrapolating outside the domain of 3. Linder T: Making the case for edible microorganisms as an
the training data [48]. Recent advances in Bayesian op­ integral part of a more sustainable and resilient food
production system. Food Secur 2019, 11:265-278, https://doi.org/
timisation [49] and physics-informed [50] AI may unlock 10.1007/s12571-019-00912-3
the benefits of both traditional modelling and modern 4. McGovern P, et al.: Early neolithic wine of georgia in the south
data-driven approaches. caucasus. Proc Natl Acad Sci 2017, 114:E10309-E10318, https://
doi.org/10.1073/pnas.1714728114
5. Chai KF, et al.: Precision fermentation to advance fungal food
Conclusion fermentations. Curr Opin Food Sci 2022, 47:100881, https://doi.
In conclusion, recent advances in AI, from optimisation org/10.1016/j.cofs.2022.100881
to machine learning and reinforcement learning, are 6. Augustin MA, et al.: Innovation in precision fermentation for food
being employed widely in precision fermentation, with ingredients. Crit Rev Food Sci Nutr 2023, 64:6218-6238, https://
applications in process synthesis, process optimisation, doi.org/10.1080/10408398.2023.2166014
control and state estimation (see Figure 2). This is ac­ 7. Good Food Institute: Fermentation: State of the Industry Report;
•• (2021).
companied by increased use of predictive surrogate NGO report on the key technologies, business developments, and sci­
models and complex metabolic models, leading to entific advances in the alternative protein fermentation industry.
powerful multiscale models that capture the relation­ 8. Humpenöder F, et al.: Projected environmental benefits of
ships between metabolism and process control. Another • replacing beef with microbial protein. Nature 2022, 605:90-96,
development is the increased role of LCA in process https://doi.org/10.1038/s41586-022-04629-w.
Scenario-based modelling and analysis of environmental impact of mi­
design, driven by the greater importance of sustain­ crobial protein partially displacing meat consumption with emphasis on
ability. Together, these approaches are accelerating the land-use.
development of sustainable fermentation technologies, 9. Pikaar I, et al.: Decoupling livestock from land use through
by facilitating experimentation, process optimisation, • industrial feed production pathways. Environ Sci Technol 2018,
52:7351-7359, https://doi.org/10.1021/acs.est.8b00216.
and scale-up, paving the way for a increased role for Discussion and analysis of different pathways for microbial protein
precision fermentation as part of a more efficient, sus­ production of protein-rich animal feed from organic and gaseous sub­
strates.
tainable, and resilient food system.
10. Larnder-Besner M, Tremblay-Gravel J, Christians A: Funding
pandemic prevention: proposal for a meat and wild animal tax.
Data Availability Sustainability 2020, 12:9016, https://doi.org/10.3390/su12219016
11. Van Boeckel TP, et al.: Global trends in antimicrobial use in food
No data were used for the research described in the ar­ animals. Proc Natl Acad Sci 2015, 112:5649-5654, https://doi.org/
10.1073/pnas.1503141112
ticle.
12. Nielsen MB, Meyer AS, Arnau J: The next food revolution is here:
•• recombinant microbial production of milk and egg proteins by
Declaration of Competing Interest precision fermentation. Annu Rev Food Sci Technol 2023,
15:173-187, https://doi.org/10.1146/annurev-food-072023-
034256.
The authors declare that they have no known competing Overview of commercial, technical and industrial landscape of precision
fermentation for food.
financial interests or personal relationships that could
have appeared to influence the work reported in this 13. Grossmann IE, Guillén-Gosálbez G: Scope for the application of
mathematical programming techniques in the synthesis and
article. planning of sustainable processes. Comput Chem Eng 2010,
34:1365-1376, https://doi.org/10.1016/j.compchemeng.2009.11.
012
Acknowledgements 14. Mencarelli L, et al.: A review on superstructure optimization
Tom Vinestock and Miao Guo gratefully acknowledge iCASE funding from
approaches in process system engineering. Comput Chem Eng
the London Interdisciplinary Doctoral Programme and Marlow Foods 2020, 136:106808.
(formerly Quorn), for the project ‘Optimising microbial protein bio­
technology using machine learning and mathematical optimisation for 15. Environmental Management — Life Cycle Assessment — Principles
protein sustainability’. and Framework; Standard. Geneva, International Organization for
Standardization, 2006.
16. Vanapalli KR, et al.: Life cycle assessment of fermentative
References and recommended reading production of lactic acid from bread waste based on process
Papers of particular interest, published within the period of review, have modelling using pinch technology. Sci Total Environ 2023,
been highlighted as: 905:167051, https://doi.org/10.1016/j.scitotenv.2023.167051
17. Sotiriou C, Zachariadis T: A multi-objective optimisation
•• of special interest
approach to explore decarbonisation pathways in a dynamic
•• of outstanding interest.
policy context. J Clean Prod 2021, 319:128623.

1. Chai WY, et al.: Fermentation process control and optimization. 18. Guillén-Gosálbez G, Grossmann I: A global optimization strategy
•• Chem Eng Technol 2022, 45:1731-1747, https://doi.org/10.1002/ for the environmentally conscious design of chemical supply
ceat.202200029. chains under uncertainty in the damage assessment model.
Excellent review fermentation process control methods, with a thorough Comput Chem Eng 2010, 34:42-58, https://doi.org/10.1016/j.
introduction to the control problem being addressed. compchemeng.2009.09.003
2. Mears L, et al.: A review of control strategies for manipulating 19. Tula AK, et al.: A computer-aided software-tool for sustainable
the feed rate in fed-batch fermentation processes. J Biotechnol process synthesis-intensification. Comput Chem Eng 2017,
2017, 245:34-46. 105:74-95, https://doi.org/10.1016/j.compchemeng.2017.01.001

www.sciencedirect.com Current Opinion in Food Science 2024, 58:101196

 8 Innovations in Food Science

20. Martín M, Adams TA II: Challenges and future directions for 35. Monteiro M, Fadda S, Kontoravdi C: Towards advanced
process and product synthesis and design. Comput Chem Eng bioprocess optimization: a multiscale modelling approach.
2019, 128:421-436, https://doi.org/10.1016/j.compchemeng.2019. Comput Struct Biotechnol J 2023, 21:3639-3655.
06.022
36. Khaleghi MK, et al.: Synergisms of machine learning and
21. Lara CL, et al.: Deterministic electric power infrastructure constraint-based modeling of metabolism for analysis and
planning: mixed-integer programming model and nested optimization of fermentation parameters. Biotechnol J 2021,
decomposition algorithm. Eur J Oper Res 2018, 271:1037-1054. 16:e2100212.
22. Peres JCG, et al.: Analysis of a microreactor for synthesizing 37. Wang B, et al.: Electronic nose signals-based deep learning
nanocrystals by computational fluid dynamics. Can J Chem Eng models to realize high-precision monitoring of simultaneous
2019, 97:594-603. saccharification and fermentation of cassava. Microchem J
2022, 182:107929.
23. Schweidtmann AM, Mitsos A: Deterministic global optimization
with artificial neural networks embedded. J Optim Theory Appl 38. Pandey AK, et al.: Machine learning in fermentative biohydrogen
2019, 180:925-948. production: advantages, challenges, and applications.
24. Anand S, Srivastava P: Comparative study for the production of Bioresour Technol 2023, 370:128502.
mycophenolic acid using penicillium brevicompactum in batch,
39. Pensupa N, Treebuppachartsakul T, Pechprasarn S: Machine
fed-batch and continuous fermentation process. Biointerface
learning models using data mining for biomass production
Res Appl Chem 2021, 12:366-376, https://doi.org/10.33263/
from Yarrowia lipolytica fermentation. Fermentation 2023, 9:239.
BRIAC121.366376
25. Petersen LN, Jorgensen JB: Real-time economic optimization 40. Packiam KAR, et al.: PERISCOPE-Opt: machine learning-based
for a fermentation process using model predictive control. Inst prediction of optimal fermentation conditions and yields of
Electr Electron Eng Inc 2014,1831-1836, https://doi.org/10.1109/ recombinant periplasmic protein expressed in Escherichia coli.
ECC.2014.6862270 Comput Struct Biotechnol J 2022, 20:2909-2920.

26. Wang B, et al.: A non-linear model predictive control based on 41. Kochkov D, et al.: Machine learning-accelerated computational
grey-wolf optimization using least square support vector fluid dynamics. Proc Natl Acad Sci USA 2021, 118:e2101784118.
machine for product concentration control in l-lysine
fermentation. Sensors 2020, 20:3335, https://doi.org/10.3390/ 42. Wang B, Wang J: Application of artificial intelligence in
s20113335 computational fluid dynamics. Ind Eng Chem Res 2021,
60:2772-2790.
27. Li D, et al.: Multi-objective reinforcement learning for fed-batch
fermentation process control. J Process Control 2022, 43. Wang Z, et al.: The role of machine learning to boost the
115:89-99, https://doi.org/10.1016/j.jprocont.2022.05.003 bioenergy and biofuels conversion. Bioresour Technol 2022,
343:126099.
28. Pandian BJ, Noel MM: Control of a bioreactor using a new
partially supervised reinforcement learning algorithm. J 44. Kushwah A, Reina TR, Short M: Modelling approaches for
Process Control 2018, 69:16-29, https://doi.org/10.1016/j.jprocont. biomass gasifiers: a comprehensive overview. Sci Total Environ
2018.07.013 2022, 834:155243.

29. Treloar NJ, et al.: Deep reinforcement learning for the control of 45. Del Rio-Chanona EA, et al.: Comparison of physics-based and
microbial co-cultures in bioreactors. PLoS Comput Biol 2020, 1, data-driven modelling techniques for dynamic optimisation of
https://doi.org/10.1371/journal.pcbi.1007783 fed-batch bioprocesses. Biotechnol Bioeng 2019,
116:2971-2982.
30. Oh TH, et al.: Integration of reinforcement learning and model
predictive control to optimize semibatch bioreactor. AIChE J 46. Misener R, Biegler L: Formulating data-driven surrogate models
2022, 68, https://doi.org/10.1002/aic.17658 for process optimization. Comput Chem Eng 2023, 179:108411.
31. Panjapornpon C, et al.: Reinforcement learning control with 47. Schweidtmann AM, et al.: Deterministic global optimization with
deep deterministic policy gradient algorithm for multivariable Gaussian processes embedded. Math Program Comput 2021,
pH process. Processes 2022, 10, https://doi.org/10.3390/ 13:553-581.
pr10122514
48. Alizadeh R, Allen JK, Mistree F: Managing computational
32. Shah P, et al.: Multi-rate observer design and optimal control to complexity using surrogate models: a critical review. Res Eng
maximize productivity of an industry-scale fermentation Des 2020, 31:275-298.
process. AIChE J 2022, 69, https://doi.org/10.1002/aic.17946
49. Wang K, Dowling AW: Bayesian optimization for chemical
33. Winz J, Assawajaruwan S, Engell S: Dynamic gray-box model of a products and functional materials. Curr Opin Chem Eng (100728)
fermentation process for spore production. Chem Ing Tech 2022, 36:100728.
2023, 95:1154-1164, https://doi.org/10.1002/cite.202200237
34. Chang L, Liu X, Henson MA: Nonlinear model predictive control 50. Bradley W, et al.: Perspectives on the integration between first-
of fed-batch fermentations using dynamic flux balance models. principles and data-driven modeling. Comput Chem Eng 2022,
J Process Control 2016, 42:137-149. 166:107898.

Current Opinion in Food Science 2024, 58:101196 www.sciencedirect.com

View publication stats