Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Available online at www.sciencedirect.

com

ScienceDirect

Challenges in process control for continuous


processing for production of monoclonal antibody
products
Anurag S Rathore1, Saxena Nikita1, Garima Thakur1 and Navnath
Deore2

Continuous processing has been garnering much interest lately 137-$220 billion by 2022 [1]. Development, scale-up, and
for production of biopharmaceutical products. Purported optimization of a bioprocess for commercial manufacturing
benefits include higher productivity (10-15X), significant remain resource intensive tasks [2]. Based on the cost of
shrinkage in facility footprint as well as equipment cost, and goods and sensitivity analysis, it has been estimated that the
improved process control and consistency in product quality. overall capital cost for production of mAb per year for fed
However, industrial implementation remains a non-trivial task batch process is three folds higher than for continuous
due to the significant complexities associated with controlling processing [3]. The continuous mode of operation has
continuous processes. Real time monitoring of critical quality the ability to handle upstream titre variations, to fulfil
attributes followed by real time process control while desired, product quality specifications and to produce the required
may not be feasible for every attribute and for every unit product in smallerunits, thereby making thesetupseasierto
operation. This article presents a perspective on the challenges create and maintain [4]. Development of advanced process
that are faced while developing and implementing a robust and control strategies [5] along with model-based data driven
adaptable automated control system for continuous control is critical for automated operation of the continuous
biopharmaceutical processes. Potential solutions are train and for achieving high productivity. However, model-
presented together with gaps that have been identified. based control needs mechanistic models, which thus far
remain too complex and computationally intensive for
Addresses
1
industrial implementation. In addition, there remains a
Department of Chemical Engineering, Indian Institute of Technology gap in availability of commercial online/real time sensors
Delhi, India
2
Tata Consultancy Services, India
for direct measurement of critical quality attributes such as
glycosylation, charge variants, aggregates, or host cell impu-
Corresponding author: Rathore, Anurag S (asrathore@biotechcmz.com) rities so as to facilitate implementation of process analytical
technology (PAT, see glossary). Process non-linearities that
need to be effectively managed in continuous processing
Current Opinion in Chemical Engineering 2021, 30:xx–yy
include variable titers and flow rates along with a lack of
This review comes from a themed issue on Biotechnology and Bio- operational flexibility in most existing process systems,
process Eng; Biotechnology and Bioengineering
thereby resulting in system conflicts during the process
Edited by Doraiswami Ramkrishna and Jamey Young scheduling. Achieving robust continuous operation
requires an improved level of automation to minimize
human intervention and appropriately handle a wide range
https://doi.org/10.1016/j.coche.2021.100671 of product and process deviations. The appropriate struc-
ture of unit operations with combination of feed-forward/
2211-3398/ã 2021 Elsevier Ltd. All rights reserved.
feed-backward control strategies and suitable PAT tools
with proper automation must be evaluated, designed and
implemented from an end-to-end continuous perspective.
This paper aims to discuss challenges that are faced during
integration, automation and control of continuous processes
Current status of continuous processing (Tables 1 and 2). Design and implementation of integrated
Motivated by the technological advancement in the field of process control using a distributed control system has been
bioprocessing and biotherapeutics, pharmaceutical compa- evaluated. This is crucial for realizing the benefits promised
nies and contract manufacturing organizations are shifting by continuous processing, and to mitigate the risks associ-
from batch to continuous processing to improve productiv- ated with its added complexity and failure modes.
ity, cope with growing demands, and reduce manufacturing
footprint. A recent market survey reported the approval and Modelling of bioprocess unit operations
production of 35 new monoclonal antibodies (mAbs) since Continuous biopharmaceutical manufacturing involves
2013, with an increase in production from 7.2% to 18.3% integration of numerous upstream and downstream unit
each year. The anticipated mAb production growth rate of operations into a single continuous process. Each unit
7%, 10% or 15% per year yields projected global sales of US$ operation by itself is dynamic in nature with multivariable

www.sciencedirect.com Current Opinion in Chemical Engineering 2019, 30:100671


2 Biotechnology and Bioprocess Eng; Biotechnology and Bioengineering

Glossary randomness and variability) and quantity of data


Process Analytical Technology (PAT): Defined by the US FDA as “a
(depends on the process operation time) available [18].
mechanism to design, analyze, and control pharmaceutical
manufacturing processes through the measurement of Critical Process The accuracy of data driven models can be improved by
Parameters (CPP) which affect Critical Quality Attributes (CQA)”. building hybrid models that combine mechanistic and
Distributed Control System (DCS): Control system to supervise statistical approaches. Figure 1a and b summarize some of
continuous processes at manufacturing scale. It is a combination of the significant literature on development and implemen-
sensors, controllers and computers distributed throughout the facility.
Programmable Logic Controller (PLC): Controller to automate the
tation of model-based or data-driven control through
production line or process industrially. It receives, monitors and record PAT.
the real time data and perform specific action based on the pre-defined
programs. Control of continuous processes
Supervisory Control and Data Acquisition (SCADA): High level process
Control architecture for continuous processing must be
control and management architecture comprising of human-machine
interface, networked data communications and computers that interact designed across a range of levels, with separate local and
with the plant via PLCs and Controllers. plant wide control. For example, local control should
Quality-by-Design (QbD): Defined by the ICH Q8 guidelines as “a involve controlling glycosylation in the bioreactor, cell
systematic approach to development that begins with predefined debris clearance in the clarification step, HCP clearance
objectives and emphasizes product and process understanding”, with
an enhanced focus on designing quality in biotherapeutic products
and product capture in the Protein A step, aggregate and
through fundamental understanding and in-process analytics. charge variant clearance in the polishing step, final con-
Quality-by-Testing (QbT): A methodology of ensuring product quality centration in the ultrafiltration step, and excipient con-
by testing raw materials and the final product prior to packaging in each centrations in the diafiltration step.
batch, but without significant fundamental understanding or in-process
analytics.
Downstream Processing (DSP): A series of steps for recovery and In contrast, plant wide control should be for process
purification of the active biotherapeutic product from natural sources, scheduling, management of surge tanks, and handling
cell culture or fermentation broth, including but not limited to of extreme deviations such as equipment breakdown or
clarification, chromatography, filtration, precipitation, and refolding. product loss. The availability of the process models is an
Upstream processing (USP): Process steps related to production of the
product molecule including early cell isolation, cell banking, inoculation,
added advantage for developing model-based control
growth, and cultivation until final harvest and collection of the live cell strategies and thereby ensuring high process productivity.
batch. The models must be developed at a range of levels,
including the smallest level (such as pressure monitoring
of depth filters and pH control of viral inactivation), single
unit operation level (such as column scheduling or pool-
interactions of process parameters affecting product qual- ing in chromatography), multiple unit operation level
ity. In such a scenario, modelling of unit operations can be (such as surge tank flowrate matching with the subse-
quite useful and this has fuelled recent interest in mech- quent unit operation), and plant wide level (such as surge
anistic modelling. Empirical, mechanistic, as well as tank management across the continuous train to ensure
hybrid models have been developed and implemented that none of them overflow or run dry, process scheduling
in real time using on-line measurements of critical process and handling of extreme deviations such as equipment
parameters and quality attributes [6,7,8,9,10]. Mecha- breakdown or product loss). The control loops must have
nistic models are developed from an in-depth under- independent goals and be designed such that they can
standing of the unit operations and the underlying physi- operate at the same level in the control structure hierar-
cal processes. Although predictive control systems based chy without conflicts. This increases the overall flexibility
on mechanistic models do exist for some individual unit and robustness of the continuous train. As each unit
operations such as bioreactor [11,12], chromatography operation is generally designed for a specific purpose,
[13,14] and ultrafiltration [15], plant-wide control strate- which is distinct from the other unit operations, it is
gies derived purely from mechanistic models are yet to be advised to develop control strategies independently. It
demonstrated. Data-driven statistical process control sys- should be clear which CQA are impacted by a given unit
tems are useful at the plant level, but their implementa- operation and thus control of that CQA should occur only
tion is limited by the diverse nature of the data, both in in the relevant control loop associated with that operation
type and time scale. Additionally, after data acquisition, to maintain process stability.
extensive efforts are required to analyse the data for
multivariate interactions. Techniques must be imple- The proposed systematic control framework in Figure 2
mented to reduce the dimensionality of the data to extract has been designed taking all these considerations into
the relevant information, such as hierarchical clustering, account. The system enables control based on real-time
multivariate least squares regression, batch evolution process parameters collected from the equipment and
modelling, artificial neural networks, or other prediction from the in-line/at-line sensors. This system also enables
and optimization algorithms [16,17]. However, a key integration with the PAT layer for developing model
limitation is that the data driven models are inherently based predictive control using multivariate data analysis
dependent on the quality (varies due to the process or empirical models. Regulatory frameworks to promote

Current Opinion in Chemical Engineering 2019, 30:100671 www.sciencedirect.com


Challenges in process control for continuous processing for production of monoclonal antibody products Rathore et al. 3

Table 1

Challenges faced during integration of analytical equipment

Analytical equipment Challenges


At-line high pressure liquid Multi-port valves are required to be integrated into the HPLC to draw samples of the required injection volume from the
chromatography (HPLC) process flow streams and flush the tubing with buffer in between sample collections. Careful positioning of sample
ports with in-line filters is required to prevent air or particles from entering the analytical columns.
Spectroscopic tools (FTIR, Flow cells with minimal hold-up volume are needed to take full advantage of the fast analytics facilitated by
NIR, MIR, Raman, spectroscopic tools, which can measure and quantify spectra on the time scale of a few seconds. However, these flow
fluorescence, multi- cells often have to be customized for biotherapeutic applications in terms of fixed/variable path length, back-pressure
wavelength UV) reduction, and sensitivity to vibrations. The spectrometer must be stable with minimal vibrations which can otherwise
lead to significant noise, especially with Raman spectroscopy. The spectrometer and flow cells/immersion probes
also need to be protected from air bubbles and temperature variations which can lead to spectral shifts. Multi-channel
spectrometers are useful for leveraging the expensive spectrometric detection device at multiple points in the
continuous train
In-line probes (pH, turbidity, Probes must be fully immersed in the flow stream and be resistant to calibration shifts, air bubbles, temperature
conductivity, pressure, changes and vibrations. The probes often do not have a human-machine interface and have to be directly connected
temperature, UV) to the overarching controller via a PLC or other input/output device using appropriate hardware handshaking.
At-line high pressure liquid Handshaking of the proprietary HPLC software with the overarching controller of the continuous train is required to
chromatography (HPLC) trigger the start, stop, and sampling intervals of the HPLC. Advanced synchronization can be required, for example if
the sample collection is at the outlet of a chromatography column and should only be triggered during elutions. Auto-
integration is required to obtain quantitative information from chromatograms without manual intervention, and these
algorithms need to be robust against potential baseline shifts, spikes in pressure or UV, and potentially large variations
in the attributes of the incoming sample. Data from auto-integrated chromatograms need to be shared with the
overarching controller in real time to facilitate control decisions or trigger alarms.
Spectroscopic tools (FTIR, Handshaking of proprietary spectrometry software with the overarching controller of the continuous train is required to
NIR, MIR, Raman, trigger spectral collection events. Automated multivariate data analytics models are required to obtain quantitative
fluorescence, multi- information from spectra in real time. The models need extensive calibration as they typically only work in a small
wavelength UV) operating range and need to be built to cover potential variations in process stream attributes without giving
erroneous data. Automated spectral quality checking and error detection algorithms are required to reduce noise.
Data from the MVDA algorithms needs to be shared with the overarching controller in real time.
In-line probes (pH, turbidity, Simple probes often do not come with proprietary software, so a complete data logging, data historian, and alarm
conductivity, pressure, system needs to be built to store the data trajectory and facilitate control decisions.
temperature, UV)

Table 2

Challenges faced during integration of process equipment

Unit operation Challenges


Bioreactor The stringent design space and process non-linearity poses a challenge in controller tuning. In addition, sensor calibration
and resolution for long duration operations is another issue.
Acoustic Wave The separation efficiency depends on load cell density which varies across batches and cell types. As AWS uses
Separator technology based on generation of standing 3D waves using low frequency acoustic forces, long duration operation
results in overheating of the equipment. Pausing the equipment for cooling down is not a feasible option owing to the
material sedimentation that often leads to hurdles in process scheduling.
Chromatography There is no gradient pump in the BioSMB to facilitate gradient elutions in polishing steps. Also, availability of seven pumps
is insufficient if performing multiple chromatography operations simultaneously.
Viral Inactivation Continuous flow at fixed flowrate is required to ensure equal residence time at low pH for all process material. Hardware
complexity increases in case of insufficient process material volume in the preceding unit operations. In such situations, to
prevent the extra hold time in coiled flow inverter reactor, standby buffer tank and pumps are required to drive the material
through the unit thereby increasing the risk of dilution of process stream.
Ultrafiltration and Flux- or TMP-based controllers are required for ensuring consistency over long continuous campaigns to handle
Diafiltration deviations arising from flux decline or variability in concentration of the incoming feed stream. The controller should be
tuned using a combination of pumps, sensors and control valves. Calibration of pumps is also a key concern as the
pumping of diafiltration buffer is conducted through six different ports simultaneously. In addition, combination of pumps
and valves is required to automate buffer flush and product recovery.
Dead-End Filtration Dead-end depth and sterile filtration are required at various points in the continuous train. A setup of filters connected in
parallel is used for switching the flow when one filter is clogged without pausing the process stream. In-line turbidity
sensors (for depth filtration) and pressure sensors (for sterile filtration) are required to be externally integrated for switching
based on turbidity/pressure breakthrough.
Surge Tanks Surge tanks are required at various points in the continuous train for flow-matching between different unit operations and
for added safety in case of unexpected deviations or equipment errors. The surge tank system must be built from the
ground up with appropriate sizing, material, temperature control and monitoring tools (weighing balances with real time
data provided to the DCS).

www.sciencedirect.com Current Opinion in Chemical Engineering 2019, 30:100671


4 Biotechnology and Bioprocess Eng; Biotechnology and Bioengineering

Figure 1

(a)

(b)

Current Opinion in Chemical Engineering

Implementation of PAT tools (primary y-axis) for multiple applications (secondary y- axis) in upstream mammalian cell culture (1a) and downstream
processing of biotherapeutics (1b) using different strategies (x-axis). For the details of the PAT-based control strategy, literature can be referred
with respect to the numbers on secondary y-axis. Each color is associated with a particular application [19–28,29,30–45,46,47–55,56,57,58,
59,60–64].

PAT implementation and continuous manufacturing for controller ensures data collection while maintaining
the innovation in development and manufacturing have integrity of each unit operation. In addition, keeping a
been provided by FDA and other regulatory agencies process data historian at a central location facilitates
[65,66]. Automating the process using a centralized analytics which involve more than one equipment and

Current Opinion in Chemical Engineering 2019, 30:100671 www.sciencedirect.com


Challenges in process control for continuous processing for production of monoclonal antibody products Rathore et al. 5

Figure 2

Current Opinion in Chemical Engineering

Potential control strategy solution showing control loops, CQA’s and CPP’s for continuous downstream processing.

allows easy implementation of advanced control algo- Control and Data Acquisition (SCADA) are used to
rithms using additional software tools based on algorithms integrate process equipment towards achieving synchro-
in Python, MATLAB, R, and SIMCA. In the case illus- nised control of multiple unit operations. Figure 3
trated here, Distributed Control System (DCS), Pro- illustrates the control system network architecture for
grammable Logic Controller (PLC), and Supervisory integration of the unit operations, all of which are

www.sciencedirect.com Current Opinion in Chemical Engineering 2019, 30:100671


6 Biotechnology and Bioprocess Eng; Biotechnology and Bioengineering

Figure 3

Current Opinion in Chemical Engineering

Control system network architecture.

connected over ethernet protocol and interface with the Challenges faced during equipment
control system via a Python Snap7 interface and the key integration
inputs to DCS and key outputs from DCS are mentioned A number of challenges arise when integrating the myriad
in Table 4. A program created in the PLC would control process and analytical equipment that typically exist in a
the scheduling, while SCADA is responsible for monitor- biopharmaceutical process [71,72,73]. The primary source
ing and control of the operation, logging alarms, interfac- of these challenges is the fact that is that the equipment
ing with control algorithms written in other software, and are typically from different solution providers (analytical
displaying the process trends and trajectories. FDA per- and process equipment vendors) and almost always oper-
spective on the adoption of control strategies for contin- ate on a proprietary software, not open to alteration by the
uous manufacturing and system integration revolves user (Table 3). This is a result of absence of standard
around characterization of incoming material, system integration protocols between various equipment vendors
dynamics and the way the perturbations propagate and pre-configured functions of the equipment, along
through the system [67]. The state of control i.e., differ- with the absence of industry standard interfaces such
ent control levels, perturbation handling, process moni- as open platform communication (OPC) for the lab-scale
toring, fault detection and raw material control comprise equipment. Thus, for integrating the different equip-
of the key elements [68]. In addition, application of ment, non-standard solutions such as overlaying a Python
continuous manufacturing in various field is being layer using a calibrated coordinate system over the exist-
reviewed by FDA [69]. The risk-based and science-based ing system software is required. Further, different equip-
approaches are being promoted. These techniques are ment have different IP addresses with no default gateway
aligned with quality by design paradigm [70]. In general, configured, leading to challenges in integration on a
the devised control strategies focus on regulating the common network. Vendor support for integration is typi-
product quality during uncertainties by monitoring the cally limited due to proprietary controls and proprietary
real time quality attributes (Level 1 control), monitoring software linked to the unit operation hardware. In gen-
the end-product, raw material and process parameters eral, equipment is protected and has custom- built user
within the design space (Level 2) and monitoring the interfaces, making it difficult to read data over standard
material attributes and process parameters within a tightly interface and implement direct control from the DCS.
constrained design space (Level 3). Additional networking switches/firewalls are typically

Current Opinion in Chemical Engineering 2019, 30:100671 www.sciencedirect.com


Challenges in process control for continuous processing for production of monoclonal antibody products Rathore et al. 7

Table 3

Software challenges faced during integration of lab equipment

Process equipment Description Software challenges Potential solutions


Bioreactor Single use bioreactor provided by GE Integrating analytical tools with the PLC based machine learning
Healthcare Life Science ensures the existing interface is a challenge causing techniques can be implemented to
controlled environment for stable difficulty in implementation of model- overcome the challenge.
growth of cell lines and proteins. based control logics, continuous data
acquisition and real time analysis.
Acoustic Wave A novel filter-less technology provided Real-time data logging to an external As OPC protocols are not provided by
Separator by Pall Life Sciences for continuous excel file is supported. However, due to the vendor, an in-house coordinate-
clarification on the basis of acousto- unavailability of OPC, control can only based system can be developed in
phoretic separation. be achieved through clicking in the Python to operate the instrument by
existing interface. The interface has layering over the existing interface. This
multiple tabs which require human layer can be in constant communication
intervention. with the DCS.
BioSMB A multi-column continuous A human-machine interface is provided As OPC protocols are not provided by
chromatography technology provided with the equipment for both continuous vendor to read/write the instrument, an
by Pall Life Sciences capable of method design and manual operation of in-house coordinate-based system can
operating 16 columns simultaneously. pumps and valves. However, the be developed in Python to operate the
Researchers have integrated methods software lacks elements required for instrument by layering over the existing
for two types of chromatography (ex: sophisticated continuous automated interface. This layer can be in constant
Protein A and CEX) such that a single control. For example, it does not allow communication with the DCS. This
unit can be used for two different unit real-time OPC based control of pumps allows control decisions for loading and
operations in the continuous train. and valves, only for switching between elution chromatography steps on the
pre-set recipes which might not be basis of multiple factors including:
sufficient to handle unexpected process product breakthrough, surge tank level
deviations. There is no option for limits, and shifts in aggregate or charge
decision-making based on external variant content. Customized control
scripts or analyzers. Also, there are logic can also be written to schedule at-
limitations in the recipe design software. line HPLC injections, carry out
automated chromatogram integration,
and make pooling decisions based on
inbuilt empirical models.
In-Line Concentration Single-pass tangential flow filtration ILC No commercial software is available for Customized control logic can be written
(ILC) and In-Line and ILD modules provided by Pall Life automated control of this unit operation. to monitor pressure and conductivity
Diafiltration (ILD) Science can achieve concentrations data from integrated in-line sensors and
from 2-10x and diafiltration of 5-10 carry out the following functions:
diavolumes in a single pass. These are pressure monitoring, conductivity
connected in series and used for the monitoring, flux control to achieve
final formulation step in the continuous target concentration factor, TMP
train. control, flow-matching between ILC
and ILD, correct flowrate of formulation
buffer relative to the process stream,
and scheduling of flush, cleaning and
product recovery steps.
Viral Inactivation in A coiled tube with optimized mixing No commercial software is available for Customized control logic can be written
Coiled Flow parameters used to achieve low pH hold automated control of this unit operation. in the DCS to monitor and control pH at
Inversion Reactor for viral inactivation in a continuous inlet and outlet by integrating with acid
(CFIR) manner by lowering the pH at the and base pumps and in-line pH sensors.
entrance and allowing fixed residence
time.
Dead End Depth and Dead end filtration is required prior to No commercial software is available for Customized control logic can be written
Sterile Filtration loading in the Protein A automated control of this unit operation. in the DCS to monitor the pressure and
chromatography columns (depth turbidity sensors and integrate multi-
filtration) and after viral inactivation way valves to switch the flow path to a
(sterile filtration). Continuous operation new filter after pressure or turbidity
is achieved by maintaining in-line breakthrough. Customized logic can
sensors and standby filters to which the also be written to carry out automated
flow is switched when pressure or water flush, buffer flush and product
turbidity breakthrough is detected. recovery.

required to integrate the equipment on one single net- which severely limit the range of recipes or control actions
work of control system, as they have different IP which could be developed, and an entire Python-based
addresses and SubNet Masks. In the case of the BioSMB, layer has to be developed to by-pass the existing software
the vendor-provided software has multiple constraints for implementation of PAT-based controls and integrated

www.sciencedirect.com Current Opinion in Chemical Engineering 2019, 30:100671


8 Biotechnology and Bioprocess Eng; Biotechnology and Bioengineering

Table 4

Key inputs to DCS and key outputs from the DCS

Unit operation Bioreactor AWS Capture Polishing chromatography Ultrafiltration – diafiltration


chromatography
Key inputs to Real time Turbidity Concentration Charge variant/aggregate Pressure
DCS aeration rate Breakthrough of first percentage Conductivity
Real time column Concentration
agitation rate
Cell density
Key outputs Predicted Feed flow rate Valve positions Elution pooling time Solenoid valve position, peristaltic
from DCS aeration rate Recirculation pump flow rate
Predicted flow rate Valve and pump Scheduling time
agitation rate Acoustic power

scheduling with the surge tanks before and after the Excellence scheme (No. BT/COE/34/SP15097/2015). Authors would also
like to acknowledge technical support and funding from Pall Life Sciences
chromatography steps. Similar approach needs to be and Tata Consultancy Services.
undertaken for the AWS and ILC-ILD setups. Without
such customized control systems layered over the existing
unit operation hardware and software, it is not possible to References and recommended reading
implement advanced scheduling strategies such as surge Papers of particular interest, published within the period of review,
have been highlighted as:
tank level management in the case of deviations, or
advanced PAT tools such as NIR monitoring of mAb  of special interest
 of outstanding interest
and excipient concentrations. In addition, integrated
continuous manufacturing is a potential solution to
address the cost, flexibility, standardization, scaling speed
and quality related challenges of the hybrid biomanufac- Declaration of Competing Interest
turing systems [71]. The authors report no declarations of interest.
1. Grilo António L, Mantalaris A: The Increasingly Human and
Profitable Monoclonal Antibody Market. Trends in
Concluding remarks Biotechnology 2019, 37(1):9-16.
Continuous manufacturing in biopharmaceutical inte- 2. Somasundaram B, Pleitt K, Shave E, Baker K, Lua LHL:
grates multiple unit operations and offers significant Progression of continuous downstream processing of
monoclonal antibodies: Current trends and challenges.
advantages with respect to cost reduction, higher produc- Biotechnology and Bioengineering 2018, 115:2893-2907 http://dx.
tivity and smaller footprint in comparison to batch pro- doi.org/10.1002/bit.26812.
cessing. However, the risk of product or process devia- 3. Yang O, Prabhu S, Ierapetritou M: Comparison between Batch
tions in a continuous set-up necessitates use of robust and and Continuous Monoclonal Antibody Production and
Economic Analysis. Ind. Eng. Chem. Res. 2019, 58:5851-5863.
intelligent controllers that are capable of handling them
in an automated manner. Developing such integrated 4. Holzer M: Is Continuous Downstream Processing Becoming a
Reality? BioProcess International 2017, 15:20-28.
controllers has several challenges with respect to devel-
5. Patnaik PR: Applications of neural networks to recovery of
oping model-based control strategies and interfacing biological products. Biotechnol. Adv. 1999, 17:477-488 http://dx.
between the unit operation equipment and control sys- doi.org/10.1016/S0734-9750(99)00013-0.
tem. Once appropriate on-line sensors and PAT tools 6. Zhang D, Del Rio-Chanona EA, Petsagkourakis P, Wagner J:
have been developed and process control models and  Hybrid physics-based and data-driven modeling for
bioprocess online simulation and optimization. Biotechnology
loops have been designed, the biggest challenge is over- and Bioengineering 2019:2919-2930 http://dx.doi.org/10.1002/
coming the closed-source system software. This limits the bit.27120.
ability of the equipment to be interfaced with other 7. Shekhawat LK, Rathore AS: Mechanistic modeling based
equipment. The optimal control system helps in sched-  process analytical technology implementation for pooling in
hydrophobic interaction chromatography. Biotechnology
uling, data archiving, plotting trends and carrying out progress 2019, 35(2).
complex model-based analytics for control and prediction
8. Kumar V, Rathore AS: Mechanistic modeling based PAT
in real time. The proposed control approach supports the implementation for ion-exchange process chromatography of
paradigm shift of manufacturing from conventional QbT- charge variants of monoclonal antibody products.
Biotechnology journal 2017, 12(9) 1700286.
based batch-wise, open-loop production to QbD-based
continuous, closed-loop production. 9. Tiwari A, Kateja N, Chanana S, Rathore AS: Use of HPLC as an
Enabler of Process Analytical Technology in Process
Chromatography. Analytical chemistry 2018, 90(13):7824-7829.
Acknowledgements 10. Hebbi V, Thakur G, Rathore AS: Process analytical technology
Authors would like to acknowledge funding support from Department of  implementation for protein refolding: GCSF as a case study.
Biotechnology, Ministry of Science and Technology through the Centre of Biotechnology and bioengineering 2019, 116(5):1039-1052.

Current Opinion in Chemical Engineering 2019, 30:100671 www.sciencedirect.com


Challenges in process control for continuous processing for production of monoclonal antibody products Rathore et al. 9

11. Kroll P, Hofer Alexandra, Stelzer Ines V, Herwig Christoph: 28. Matthews TE, Smelko JP, Berry B, Romero-Torres S, Hill D,
Workflow to set up substantial target-oriented mechanistic Kshirsagar R, Wiltberger K: Glucose Monitoring and Adaptive
process models in bioprocess engineering. Process Biochem. Feeding of Mammalian Cell Culture in the Presence of Strong
2017, 62:24-36. Autofluorescence by near Infrared Raman Spectroscopy.
Biotechnol. Prog. 2018 http://dx.doi.org/10.1002/btpr.2711.
12. Mears L, Stocks SM, Albaek MO, Sin G, Gernaey KV: Mechanistic
Fermentation Models for Process Design, Monitoring, and 29. Ma F, Zhang A, Chang D, Velev OD, Wiltberger K, Kshirsagar R:
Control. Trends Biotechnol. 2017, 35:914-924.  Real-time monitoring and control of CHO cell apoptosis by in
situ multifrequency scanning dielectric spectroscopy. Process
13. Steinebach F, Angarita M, Karst DJ, Müller-Späth T, Morbidelli M: Biochemistry 2019, 80:138-145 http://dx.doi.org/10.1016/j.
Model based adaptive control of a continuous capture procbio.2019.02.017.
process for monoclonal antibodies production. J Chromatogr A
2016:50-56 http://dx.doi.org/10.1016/j.chroma.2016.03.014. 30. Drieschner TE, Ostertag TE, Boldrini B, Lorenz A, Brecht M,
Rebner K: Direct optical detection of cell density and viability of
14. Feidl F, Garbellini S, Luna MF, Vogg S, Souquet J, Broly H, mammalian cells by means of UV/VIS spectroscopy. Anal
Morbidelli M, Butté A: Combining Mechanistic Modeling and Bioanal Chem. 2020, 412:3359-3371 http://dx.doi.org/10.1007/
Raman Spectroscopy for Monitoring Antibody s00216-019-02322-y.
Chromatographic Purification. Processes 2017, 7:683.
31. Rafferty C, O’Mahony J, Burgoyne B, Rea R, Balss KM, Latshaw
15. Huter MJ, Strube J: Model-Based Design and Process DC II: Raman spectroscopy as a method to replace off-line pH
Optimization of Continuous Single Pass Tangential Flow during mammalian cell culture processes. Biotechnology and
Filtration Focusing on Continuous Bioprocessing. Processes Bioengineering 2017, 117(1):146-156.
2019, 7:317.
32. Kozma B, Salgó A, Gergely S: On-line glucose monitoring by
16. Zhang Z, Schwartz S, Wagner L, Miller W: A greedy algorithm for near infrared spectroscopy during the scale up steps of
aligning DNA sequences. Journal of Computational Biology mammalian cell cultivation process development. Bioprocess
2000, 7:203-214. Biosyst Eng 2019, 42:921-932 http://dx.doi.org/10.1007/s00449-
17. Oyetunde T, Bao FS, Chen JW, Martin HG, Tang YJ: Leveraging 019-02091-z.
knowledge engineering and machine learning for microbial 33. Ulonska S, Kroll P, Fricke J, Clemens C, Voges R, Müller MM,
bio-manufacturing. Biotechnol Adv 2018, 36:1308-1315. Herwig C: Workflow for Target-Oriented Parametrization of an
18. An D, Kim NH, Choi JH: Practical options for selecting data- Enhanced Mechanistic Cell Culture Model. Biotechnology
driven or physics-based prognostics algorithms with reviews. Journal 2017, 13(4) http://dx.doi.org/10.1002/biot.201700395.
Reliability Engineering and System Safety 2015, 133:223-236. 34. Wu CH, Wee S: Micro sequential injection system as the
19. Kuystermans D, Avesh M, Al-Rubeai M: Online flow cytometry interfacing device for process analytical applications.
for monitoring apoptosis in mammalian cell cultures as an Biotechnol. Prog. 2015, 31:607-613 http://dx.doi.org/10.1002/
application for process analytical technology. Cytotechnology btpr.2055.
2016, 68:399-408 http://dx.doi.org/10.1007/s10616-014-9791-
35. Gorrini F, Biagiola S, Figueroa JL, Wouwer AV: Reaction rate
9793.
estimation and model predictive control of hybridoma cell
20. Santos RM, Kaiser P, Menezes JC, Peinado A: Improving cultures. IFAC-PapersOnLine 2019, 52(1):715-720 http://dx.doi.
reliability of Raman Spectroscopy spectroscopy for mAb org/10.1016/j.ifacol.2019.06.147.
production by upstream processes during bioprocess
development stages. Talanta 2019, 199:396-406 http://dx.doi. 36. Nie L, Gao D, Jiang H, Gou J, Li L, Hu F, Guo T, Wang H, Qu H:
org/10.1016/j.talanta.2019.02.088. Development and Qualification of a Scale-Down Mammalian
Cell Culture Model and Application in Design Space
21. Craven S, Whelan J, Glennon B: Glucose concentration control Development by Definitive Screening Design. AAPS
of a fed-batch mammalian cell bioprocess using a nonlinear PharmSciTech 20 2019, 246 http://dx.doi.org/10.1208/s12249-
model predictive controller. Journal of Process Control 2014, 019-1451-7.
24:344-357 http://dx.doi.org/10.1016/j.jprocont.2014.02.007.
37. Kotidis P, Jedrzejewski P, Sou SN, Sellick C, Polizzi K, del Val IJ,
22. André S, Cristau LS, Gaillard S, Devos O, Calvosa E, Duponchel L: Kontoravdi C: Model-based optimization of antibody
In-line and real-time prediction of recombinant antibody titer galactosylation in CHO cell culture. Biotechnology and
by in situ Raman spectroscopy. Analytica Chimica Acta 2015, Bioengineering 2019, 116(7):1612-1626 http://dx.doi.org/10.1002/
892:148-152. bit.26960.

23. Mehdizadeh H, Lauri D, Karry KM, Moshgbar M, Procopio- 38. Patel BA, Gospodarek A, Larkin M, Kenrick SA, Haverick MA,
Melino R, Drapeau D: Generic Raman-based calibration models Tugcu N, Brower MA, Richardson DD: Multi-angle light
enabling real-time monitoring of cell culture bioreactors. scattering as a process analytical technology measuring real-
Biotechnology Progress 2015, 31:1004-1013. time molecular weight for downstream process control. mAbs
2018, 10(7):945-950 http://dx.doi.org/10.1080/
24. Metze S, Ruhl S, Greller G, Grimm C, Scholz J: Monitoring online 19420862.2018.1505178.
biomass with a capacitance sensor during scale-up of
industrially relevant CHO cell culture fed-batch processes in 39. Hansen S, Brestrich N, Staby A, Hubbuch J: Mid-UV Protein
single-use bioreactors. Bioprocess Biosyst. Eng. 2020, 43:193- Absorption Spectra and Partial Least Squares Regression as
205 http://dx.doi.org/10.1007/s00449-019-02216-4. Screening and PAT Tool. Preparative Chromatography for
Separation of Proteins 2017:501-536.
25. Berry B, Moretto J, Matthews T, Smelko J, Wiltberger K: Cross-
scale predictive modeling of CHO cell culture growth and 40. Brestrich N, Sanden A, Kraft A, McCann K, Bertolini J, Hubbuch J:
metabolites using Raman spectroscopy and multivariate Advances in inline quantification of co-eluting proteins in
analysis. Biotechnol. Prog. 2015, 31:566-577. chromatography: Process-data-based model calibration and
application towards real-life separation issues. Biotechnology
26. Berry BN, Dobrowsky TM, Timson RC, Kshirsagar R, Ryll T, and bioengineering 2015, 112(7):1406-1416.
Wiltberger K: Quick generation of Raman spectroscopy based
in-process glucose control to influence biopharmaceutical 41. Brestrich N, Hahn T, Hubbuch J: Application of spectral
protein product quality during mammalian cell culture. deconvolution and inverse mechanistic modelling as a tool for
Biotechnol. Prog. 2016, 32:224-234. root cause investigation in protein chromatography. Journal of
Chromatography A 2016, 1437:158-167.
27. Matthews TE, Berry Brandon N, Smelko J, Moretto J,
Moore Brandon, Wiltberger Kelly: Closed loop control of lactate 42. Großhans S, Rüdt M, Sanden A, Brestrich N, Morgenstern J,
concentration in mammalian cell culture by Raman Heissler S, Hubbuch J: In-line Fourier-transform infrared
spectroscopy leads to improved cell density, viability and spectroscopy as a versatile process analytical technology for
biopharmaceutical protein production. Biotechnol. Bioeng. preparative protein chromatography. Journal of
2016, 113:2416-2424. Chromatography A 2018, 1547:37-44.

www.sciencedirect.com Current Opinion in Chemical Engineering 2019, 30:100671


10 Biotechnology and Bioprocess Eng; Biotechnology and Bioengineering

43. Sanden A, Suhm S, Rüdt M, Hubbuch J: Fourier-transform 59. Rolinger L, Rüdt Matthias, Diehm Juliane, Chow-
infrared spectroscopy as a process analytical technology for  Hubbertz Jessica, Heitmann Martin, Schleper Stefan,
near real time in-line estimation of the degree of PEGylation in Hubbuch Jürgen: Multi-attribute PAT for UF/DF of Proteins—
chromatography. Journal of Chromatography A 2019, 460410. Monitoring Concentration, particle sizes, and Buffer
Exchange. Analytical and Bioanalytical Chemistry 2020, 412:2123-
44. Rüdt M, Brestrich N, Rolinger L, Hubbuch J: Real-time 2136 http://dx.doi.org/10.1007/s00216-019-02318-8.
monitoring and control of the load phase of a protein A capture
step. Biotechnology and bioengineering 2017, 114(2):368-373. 60. Huter M, Strube J: Model-Based Optimization of SPTFF
Ultrafiltration for Integration in Continuous
45. Mendhe R, Thukkaram M, Patil N, Rathore AS: Comparison of Biopharmaceutical Processing. Chemie Ingenieur Technik 2018,
PAT based approaches for making real-time pooling decisions 90 http://dx.doi.org/10.1002/cite.201855263 1251–1251.
for process chromatography–use of feed forward control.
Journal of Chemical Technology & Biotechnology 2015, 90(2):341- 61. Hebbi V, Roy S, Rathore AS, Shukla A: Modeling and prediction
348. of excipient and pH drifts during ultrafiltration/diafiltration of
monoclonal antibody biotherapeutic for high concentration
46. Sauer DG, Melcher M, Mosor M, Walch N, Berkemeyer M, Scharl-
formulations. Separation and Purification Technology 2020, 238
 Hirsch T, Leisch F, Jungbauer A, Dürauer A: Real-time
http://dx.doi.org/10.1016/j.seppur.2019.116392 116392.
monitoring and model-based prediction of purity and quantity
during a chromatographic capture of fibroblast growth factor 62. Rucker-Pezzini J, Arnold L, Hill-Byrne K, Sharp T, Avazhanskiy M,
2. Biotechnology and bioengineering 2019. Forespring C: Single pass diafiltration integrated into a fully
47. Zelger M, Pan S, Jungbauer A, Hahn R: Real-time monitoring of continuous mAb purification process. Biotechnology and
protein precipitation in a tubular reactor for continuous Bioengineering 2018, 115:1949-1957 http://dx.doi.org/10.1002/
bioprocessing. Process Biochemistry 2016, 51(10):1610-1621. bit.26708.

48. Walch N, Scharl T, Felföldi E, Sauer DG, Melcher M, Leisch F, 63. Ghodbane Mehdi, Sripada Sobhana, Bose Sayantan,
Dürauer A, Jungbauer A: Prediction of the Quantity and Purity of Pohlhaus Patrick, Molek Jessica, Bhawnani Rajan: Analytical and
an Antibody Capture Process in Real Time. Biotechnology data strategy for continuous downstream manufacturing. In
journal 2019. 1800521. Integrated Continuous Biomanufacturing IV. Edited by Warikoo ea,
Jungbauer os, Coffman n, Walther sn. 2019 . Roche, USA, BOKU,
49. Zobel-Roos S, Mouellef M, Siemers C, Strube J: Process Austria, Boehringer Ingelheim, USA, Sanofi, USA Eds, ECI
Analytical Approach towards Quality Controlled Process Symposium Series https://dc.engconfintl.org/biomanufact_iv/15.
Automation for the Downstream of Protein Mixtures by Inline
Concentration Measurements Based on Ultraviolet/Visible 64. Kateja N, Kumar D, Sethi S, Rathore AS: Non-protein A
Light (UV/VIS) Spectral Analysis. Antibodies 2017, 6(4):24. purification platform for continuous processing of monoclonal
antibody therapeutics. Journal of Chromatography A 2018,
50. Williams A, Read EK, Agarabi CD, Lute S, Brorson KA: Automated 1579:60-72 http://dx.doi.org/10.1016/j.chroma.2018.10.031.
2D-HPLC method for characterization of protein aggregation
with in-line fraction collection device. Journal of 65. https://www.fda.gov/regulatory-information/search-fda-
Chromatography B 2017, 1046:122-130. guidance-documents/pat-framework-innovative-pharmaceutical-
development-manufacturing-and-quality-assurance" https://
51. Patel BA, Pinto ND, Gospodarek A, Kilgore B, Goswami K, www.fda.gov/regulatory-information/search-fda-guidance-
Napoli WN, Desai J, Heo JH, Panzera D, Pollard D, Richardson D, documents/pat-framework-innovative-pharmaceutical-
Brower M, Richardson D: On-line ion exchange liquid development-manufacturing-and-quality-assurance.
chromatography as a process analytical technology for
monoclonal antibody characterization in continuous 66. https://www.fda.gov/regulatory-information/search-fda-
bioprocessing. Analytical chemistry 2017, 89(21):11357-11365. guidance-documents/quality-considerations-continuous-
manufacturing.
52. Hou Y, Jiang C, Shukla AA, Cramer SM: Improved process
analytical technology for protein a chromatography using 67. Chatterjee S: FDA perspective on continuous manufacturing.
predictive principal component analysis tools. Biotechnology Baltimore, MD: IFPAC Annual Meeting. 2012:34-42.
and bioengineering 2011, 108(1):59-68.
68. Lee SL: Current FDA perspective for continuous
53. Chmielowski RA, Mathiasson L, Blom H, Go D, Ehring H, Khan H, manufacturing. 2nd International Symposium on Continuous
Li H, Cutler C, Lacki K, Tugcu N, Roush D: Definition and dynamic Manufacturing of Pharmaceuticals; Cambridge, Vereinigte Staaten
control of a continuous chromatography process independent von Amerika: 2016.
of cell culture titer and impurities. Journal of Chromatography A
2017, 1526:58-69. 69. Lute Scott, Agarabi Cyrus, Johnson Sarah, Chavez Brittany,
Brorson Kurt: FDA/OBP laboratory research to support
54. Krättli M, Steinebach F, Morbidelli M: Online control of the twin-
continuous bioprocessing. In Integrated Continuous
column countercurrent solvent gradient process for
Biomanufacturing III. Edited by Farid zne, Goudar ea, Warikoo ea.
biochromatography. Journal of chromatography A 2013,
2017 . University College London, United Kingdom, Amgen, USA
1293:51-59.
Paula Alves, IBET, Portugal, Axcella Health, Inc., USA Eds, ECI
55. Papathanasiou MM, Avraamidou S, Oberdieck R, Mantalaris A, Symposium Series https://dc.engconfintl.org/biomanufact_iii/10.
Steinebach F, Morbidelli M, Mueller-Spaeth T, Pistikopoulos EN:
Advanced control strategies for the multicolumn 70. Lee SL, O’Connor TF, Yang X, Cruz CN, Chatterjee S,
countercurrent solvent gradient purification process. AIChE Madurawe RD, Moore CMV, Yu LX, Woodcock J: Modernizing
Journal 2016, 62(7):2341-2357. pharmaceutical manufacturing: from batch to continuous
production. Journal of Pharmaceutical Innovation 2015, 10
56. Papathanasiou MM, Steinebach F, Morbidelli M, Mantalaris A, (3):191-199.
 Pistikopoulos EN: Intelligent, model-based control towards the
intensification of downstream processes. Computers & 71. Konstantinov KB, Cooney CL: White paper on continuous
Chemical Engineering 2017, 105:173-184. bioprocessing. May 20-21, 2014 Continuous Manufacturing
Symposium. Journal of pharmaceutical sciences 2015, 104
57. Baur D, Angarita M, Müller-Späth T, Morbidelli M: Optimal model- (3):813-820 http://dx.doi.org/10.1002/jps.24268.
based design of the twin-column CaptureSMB process
improves capacity utilization and productivity in protein A 72. Halan Babu, Minas Wolfgang: CONTINUOUS BIOPROCESSING
affinity capture. Biotechnol J 2016, 11(1):135-145. PROMISES AND CHALLENGES BioProcess International
eBooks. BioProcess International. 2020.
58. Thakur G, Hebbi V, Rathore AS: An NIR-based PAT approach for
 real-time control of loading in Protein A chromatography in 73. Brower M, Heise C, Moody G, Simpson C: Continuous
continuous manufacturing of monoclonal antibodies. biomanufacturing: a new approach to process scale.
Biotechnology and Bioengineering 2019, 117:673-686 http://dx. Bioprocess international 2020.
doi.org/10.1002/bit.27236.

Current Opinion in Chemical Engineering 2019, 30:100671 www.sciencedirect.com

You might also like