Upload
dinhkhue
View
218
Download
0
Embed Size (px)
Citation preview
Challenges and Opportunities in Biopharmaceutical Manufacturing Control
Moo Sun Hong, Kristen Severson, Mo Jiang, Amos E. Lu, J. Christopher Love, Richard D. Braatz
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
Background: Biopharmaceuticals Manufacturing
• Products derived from biological organisms for treating or preventing diseases
• Hundreds of approved products on the market
• Over 7000 medicines in development
• Usually any off-spec material requires rejection of entire lot
Deloitte (2016). 2016 Global Life Sciences Outlook: Moving Forward with Cautious Optimism. Deloitte LLP, Boston, MA.Informa (2016). Pharmaprojects Pharma R&D Annual Review 2016. Informa PLC, London, United Kingdom.
USD in billions
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
500
400
300
200
100
0
% non-bio drug candidates in pipeline
% bio drug candidates in pipeline95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
100908070605040302010
0
Current State of Biopharmaceutical Manufacturing
• Sequence of batch unit operations
• Upstream: cell culture and harvest
• Downstream: purification using chromatography, filtration, diafiltration
• Bioreactor has T, dissolved oxygen, and maybe pH controls
• Minimal modeling/controls
Cell Culture Centrifugation
Protein A Chromatography
Low pHViral
Inactivation
Cation ExchangeChromatography
Depth and Membrane Filtration
Anion ExchangeChromatography
Viral Filtration
Ultrafiltration/Diafiltration
DOWNSTREAM
UPSTREAM
Kelley, B. (2009). Industrialization of mAb Production Technology: The Bioprocessing Industry at a Crossroad. MAbs, 1, 443-452.Shukla, A. A., Thömmes, J. (2010). Recent Advances in Large-Scale Production of Monoclonal Antibodies and Related Proteins. Trends in Biotechnology, 28, 253-261.
Motivation: Biopharmaceuticals Manufacturing
• Over 7000 medicines in development
• Industry driven to look for new technology to increase flexibility and reduce costs
continuous flow
novel process designs
Deloitte (2016). 2016 Global Life Sciences Outlook: Moving Forward with Cautious Optimism. Deloitte LLP, Boston, MA.Informa (2016). Pharmaprojects Pharma R&D Annual Review 2016. Informa PLC, London, United Kingdom.
USD in billions
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
500
400
300
200
100
0
% non-bio drug candidates in pipeline
% bio drug candidates in pipeline95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16
100908070605040302010
0
Moving Towards Continuous-flow Operations
http://www.drugdevelopment-technology.com/contractors/contract_research/biovian/biovian1.htmlhttp://www.genengnews.com/gen-articles/novasep-morphs-into-full-service-company/3048/
• Move manufacturing from batch to integrated continuous operations with minimal holdup in between
• Continuous flow enables higher process flexibility, consistency,
and volumetric capacity lower equipment cost and operational
complexity tighter specifications on product quality
→ need for research in process modeling/design/controls
Example Limitation of an Existing Process Design: Packed-Bed Chromatography
• Currently the dominant method for protein purification• High resolution but costly for high-dose pharmaceuticals• The resin in a single column can cost >$1 million• Operating cost scales
linearly with throughput• Strongly prefer
sublinear scaling→ see paper for design
alternatives
Inlet Outlet
sorption
convection &dispersion
Porous bead
pore diffusion
film mass transfer
→ need for research in process modeling/design/controls
Technology Directions
• Increased understanding and optimization of each unit operation and component
• Microscale technologies for high-speed continuous process development
• Availability of plug-and-play modular unit operations with integrated process control and monitoring systems to facilitate straightforward deployment in the laboratory
• Dynamic mathematical models for unit operations and entire plant to support process development and plant-wide control
• Easy-to-use model-based control technologies for optimizing startup, changeover, and shutdown
Design of Control Systems Based on “Virtual Plant”
• Constructed from first-principles models wherever possible, grey-box models where necessary
• Highest complexity models used for the invention and optimization of process designs and development
• Lower complexity plant-wide model runs in parallel with process operations, for process control and quality and equipment condition monitoring
Micromixer
pH
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
Towards Biomanufacturing on Demand (BioMOD)
• Enable flexible methodologies for genetic engineering/modification of microbial strains to synthesize multiple and wide-ranging protein-based therapeutics
• Develop flexible & portable device platforms for manufacturing multiple biologics with high purity, efficacy, and potency, at the point-of-care, in short timeframes, when specific needs arise
• Include end-to-end manufacturing chain (including downstream processing) within a microfluidics-based platform
J.C. Love, Towards making biologic drugs on demand, 4th Int. Conf. on Accelerating Biopharmaceutical Development, January 25, 2015
Design Requirements Patient
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forReleaseNot OK
☐
Final Product
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
J.C. Love, Towards making biologic drugs on demand, 4th Int. Conf. on Accelerating Biopharmaceutical Development, January 25, 2015
Plant-wide Control Approach
Characteristics of InSCyT• Multi-product manufacturing plant• Continuous & discrete operations• All of the process characteristics
articulated by Lee/Weekman 1976, Foss 1983, and more
• No SS & must align with regulatory requirements (no off-spec product)
Approach adapted from the chemical industry• Employing systematic & modular design of plantwide control strategies
for large-scale manufacturing facilities (Stephanopoulos/Ng, JPC 2000)• Using algorithms that can handle nonlinearities, uncertainties,
distributed states, time delays, unstable zero dynamics, constraints, and mixed continuous-discrete operations
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forReleaseNot OK
☐
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
Plant-wide Control Approach
• Build first-principles dynamic modelsfor each unit operation (UO)
• Design control system for each UOto meet “local” material attributes
• Evaluate performance in simulationsand propose design modificationsas needed
• Implement and verify the control system for each UO
• Design and verify plantwide control system to ensure that the product quality specifications are met
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forReleaseNot OK
☐
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forReleaseNot OK
☐
Final Product
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
Robust Split-range Control of Bioreactor
• Designed hybrid robust controllers that switch during each manipulated variable change
• Vastly reduces variability by optimized switching
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Downstream
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forReleaseNot OK
☐
Final Product
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
On-line Reactor ControlPerfusion
UF/DF Membrane
Waste
Fill
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Real-TimeProcess
Data
Affinity Chromatography
Waste
Column Tanks 1 2
EluteWaste
Polishing Membrane #1 Tanks
1 2
Polishing Membrane #1
Waste
1 2
Polishing Membrane #2
Polishing Membrane #2 Tanks
Chromatography Modeling
Nonlinear PDEs• Species mass balance
in liquid phase• Adsorbed phase balance
θ∂
−=∂
*( )ii ik qq q
Inlet Outlet
SorptionConvection &
Dispersion
Interstitial volume Porous bead
Pore diffusion
Film mass transfer
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Chromatography Modeling
Nonlinear PDEs• No steady-state, always in transient mode• No measurements in the column, large time delay to exit
measurement• Switch between multiple sets of nonlinear PDEs• Practice is to run open loop or switch on exit concentrations
Inlet Outlet
SorptionConvection &
Dispersion
Interstitial volume Porous bead
Pore diffusion
Film mass transfer
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Hybrid Simulation of Chromatography Operation
• Four-stage bind-and-elute operation mode
# Stage Time Protein Conc. Salt Conc.1 Load tl cp,i cs,l
2 Wash tw 0 cs,w
3 Purge tp 0 cs,e
4 Elute te 0 cs,e
# Specified by upstream units
# #
→ Sensors at entrance and exit only, i.e., feedback limited→ Open-loop optimal control can be used to determine
“optimal” operation based on first-principles modelA.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Multiple Performance Objectives
Performance Objective Description
Purity
Recovery
Exit Concentration
Productivity
0
,0 0
,
,
e
e e
pdt elution
pdt elution cont elution
t
t t
c dt
c dt c dt+
∫∫ ∫
,
,
0
et
pdt elution
pdt loal d
c dt
t c∫
,0
e
pdt elution
t
e
c dt
t∫
,0
e
pdt elution
l w p e d
tQ c dt
t tt t t+ ++ +∫
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Region oflow quality
localextrema
Global optimum
Multiobjective Optimization
• Purity, recovery, exit concentration, and productivity are competing objectives
• Look to maximize exit concentration & productivity while constraining purity to be >99% and recovery to be >90%
• Multiobjective problem solved using stochasticoptimization to avoid being trapped in low-quality local extrema
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Pareto Optimality (Tradeoff Curve)
Dilution Factor
Curves are optimal tradeoff for various feed concentrations
Select “optimal” value based on measured feed concentration using a lookup table
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Nonlinear feedforward control of a hybrid PDE system, with minimal online computational cost
Column Tanks 1 2
Downstream
Affinity Chromatography
Sterile Media
Yeast Inoculum
Crude Product Hold Tanks
Upstream
1 2 1 2
Polishing Membrane #1 Tanks
Polishing Membrane #1
Polishing Membrane #2
Waste Waste Waste
Polishing Membrane #2 Tanks
UF/DF Membrane
Waste
Concentration
Identity
Purity
pH, DO, T
Potency
Safety
Sterility
Analytics
Purified Product for Quality Testing
Multivariate Model
OK forRelease
Not OK
☐
Final Product
On-line Reactor Control
Perfusion
Fill
Elute
Real-TimeProcess
Data
Need buffer makeup unit to avoid carrying all possible buffers on the platform
Integrated and Scalable Cyto-Technology (InSCyT) Biomanufacturing Platform
Batch Buffer Makeup “UO” Modeling
pH Model (Waller, Kravaris, ...)Method of reaction invariants
Mass balances are linear equationspH is a static nonlinear function of the invariants
Multicomponent Mixing RuleRequires estimation of effective ionic radius from data
Single Component Conductivity First-principles Model
Flowrates Fin are the manipulated variables
Semi-batch Tank Reaction Invariant Species Conservation
;
Many Systems and Control Problems Appear in the Buffer “UO”
• Dynamics are highly nonlinear• Slow mixing and uncertain parameters• Model-based control can overcome these issues by predicting the effect of inputs on the process
– Need good parameter values based on process data– Need to develop robust control strategies that operate well
under uncertainty
Design of experiments for conductivitypH control in micromixers
examples
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Example of pH Control in Buffer Mixing “UO”
• Use of in-line micromixers allows for process intensification• pH modeled using the reaction-invariant method• Material balances modeled using tanks-in-series
Micromixer
pH
Fw, wwFb, wbFp, wpFs, ws
w1 w2 wi wn
11
1 1
1 tin
Fddt V V
−= Fw W w
T
w b p sF F F F = F
t w b p sF FF F F+ + += 1 )(ti i
i
id Fdt V −= −w ww
[ ]in w b p s=W w w w w pH ( )nf= w
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Direct PID vs Reaction-invariant PID Control
• Direct PID
• Reaction-invariant PID
Slow response
Marginally stable
Fast response
Slow response
PID based on the reaction invariants improves performance across pH range
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Reaction-invariant Controller:Performance Under pH Model Uncertainty
• No model uncertainty
• 20% error in pKas within pH model
InstabilitySlow response
Caveat: Reaction-invariant control requires accurate models to achieve good
performance
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Model Adaptation of Reaction-invariant Control via Dual Micromixer Configuration
Micromixer
pH
Micromixer
pH
First micromixer used to excite system to enable accurate parameter estimation of pH modelSecond micromixer used to perform corrections to the output of the first micromixer and effect setpoint tracking
Rhinehart & Riggs proposed multi-tankLinear control of n tanks studied by SkogestadOur approach:
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Dual Micromixer Configuration
Input signal to generate data for model adaptation
First micromixer used to excite system to enable accurate parameter estimation in pH model
Second micromixer used to perform corrections to the output of the first micromixer and effect setpoint tracking
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Closed-loop Performance Comparison
• Single micromixer without model adaptation
• Dual micromixer with model adaptation
Model adaptation allows for rapid response in spite of high uncertainty
A.E. Lu, J.A. Paulson, N.J. Mozdzierz, A. Stockdale, A.N. Ford Versypt, K.R. Love, J.C. Love, R.D. Braatz (2015). Control systems technology in the advanced manufacturing of biologic drugs. Proc. of the IEEE Conference on Control Applications, 1505-1515.
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
Comments on the State-of-the-Art in Model-based Control Technology
• The best commercial plant simulation software handles nonlinearities, time delays, unstable zero dynamics, constraints, mixed continuous-discrete operations, and some uncertainty analysis methods (e.g., Si, MC)
• Distributed states facilitated by various math tricks, e.g., moment analysis, transform methods, characteristics
• More sophisticated polynomial series and uncertainty analysis tools can be wrapped around of integrated into such software
• More control research is needed to prove stability, simplify tuning, reduce on-line computational cost, and improve closed-loop performance (e.g., MPC)
Outline
• Background, Current State, Technology Directions
• A Biomanufacturing-on-Demand Platform
• Comments on the State-of-the-Art in Control
• Closing
Summary
• Described some research challenges and opportunities in the manufacturing of biopharmaceuticals
• Discussed some control technology within the context of a specific biopharmaceuticals-on-demand platform
• Model-based control solutions are a mix of nonlinear feedback & feedforward strategies
• Described hybrid controls and a new type of adaptive control strategy motivated by a specific application need
• Described state-of-the-art in control technology and some research needs
Closing Comments
• Small-molecule pharma PSE went from very little in the late 1990s to becoming well established
• We believe that the rise of PSE in biologic manufacturing will mirror its rise in small-molecule pharmaceuticals that occurred over the last 20 years, but will be faster
D2
D1
R2
R1LC LC
M1 M2 S1
C1 C2
W1
LC LC
C3 C4
LC
LC LC
S3S4 W2
M3 M4
M5
S5 S6 E1 CS
CAT
A
B
S2
S1
S3
S1
PU1
PU2
PU4
PU3
S1
S1
S1
C
D
E
S1
EX1
EX2FP
LC
S1
FCsp
FT
FCsp
spsp
sp
FCsp
CC FT
CTFC
FCsp
TC
sp
spsp
FCsp
DCsp
sp
S1RC CC sp
DC
LC
sp
1CATA B I+ ←→
1 2 BPI C I P+ → +
2I E API+ →
Bequette, Comp Chem Eng, 20:S1583–S1588, 1996
Togkalidou+, AIChE J, 47:160-168, 2001
Lakerveld+, Org Process R&D, 19:1088-1100, 2015
Simon+, Org Process R&D, 19:3-62, 2015
AcknowledgmentsHassan Ait HaddouMonica AminPaul BaroneCatie BartlettLisa BradburyDanielle CampJicong CaoDivya ChandraMichael CostaSteve CramerAleksander CvetkovicAmanda Del RosarioSusan DexterJuyai-Ivy DongChaz GoodwineJongyoon HanPeter HanWilliam HancockWilliam HerringtonShan JiangPankaj KarandeSojin KimSung Hee KoRalf KuriyelTaehong KwonRachel LeesonJames LeungLihong LiuKerry LoveAmos LuTimothy Lu
Huaiya MengSylvia MessierNicholas MozdzierzJustin NelsonRussell NewtonWei OuyangSantosh PandeJoel PaulsonPablo Perez-PineraOliver PurcellGK RajuRajeev RamNigel ReuelDaniel SalemKartik ShahDivya ShastryGajendra SinghAnthony SinskeyStacy SpringsAlan StockdaleMichael StranoKyOnese TaylorSteve TimmickJP TrasattiNicholas VecchiarelloAshley Ford VersyptAnnie WangJames WooDi WuAnna Zhang
This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) and SPAWAR Systems Center Pacific (SSC Pacific) under Contract No. N66001-13-C-4025. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Defense Advanced Research Projects Agency (DARPA) and SPAWAR Systems Center Pacific (SSC Pacific).
High-Dimensional and Heterogeneous Data
Heterogeneity in data type and time scale in bioreactor has motivated the use of• Kernel-based methods• Dynamic time warping (DTW)
Very high dimensions in hyperspectral images have motivated the use of• Hierarchical clustering• Greedy algorithms
Charaniya, S., Hu, W. Karypis, G. (2008). Mining Bioprocess Data: Opportunities and Challenges. Trends in Biotechnology, 26, 690-699.
Bioreactor example
Supplementing First-principles with Black-box Models
First-principles model• Based on detailed process
understanding• Conservation equations, reaction
mechanisms, constitutive relations
Black-box model• Based on statistical correlations
observed in experimental data• Principal component analysis
(PCA), partial least squares (PLS)
“Grey-box” model• Combine first-principles and black-box models• Generate more complete and accurate predictions• Parallel approach• Embedded approach• Surrogate output approach
Packed-Bed Chromatography
• Currently the dominant method for protein purification• High resolution but costly for high-dose pharmaceuticals• Operating cost scales linearly with throughput• Increased product titers achieved with modern bioreactor
Non-chromatography separation methods• Aqueous two-phase extraction• Precipitation• Crystallization
Crystallization
• Effective and inexpensive industrial operation to achieve adequate purity and production in large quantities
• Cost scales sub-linearly with throughput• Used for small-molecule drugs and some proteins
• Insulin: treatment of diabetes (room T, pH 8.2, ¼~72 h)• Insulin lispro: fast-acting insulin analogue
5°C, pH 9, NaCl 37.5 mM,
0.75 M acetic acid, 24 h
Insulin lispro 20 g/L
Fermentation Cell Separation
Proteolytic Excision Chromatography Crystallization
Jackson, R. (1973). Process for the Crystallization of the Ammonium and Alkali Metal Salts in Insulin. U.S. Patent 3,719,655.Baker, J. C., Roberts, B. M. (1997). Preparation of Stable Insulin Analog Crystals. U.S. Patent 5,597,893.
Challenges for Therapeutic Proteins
• Hard to find conditions for reproducible crystallization using pharmaceutical-grade buffers and precipitants
• Protein can be denaturated by pH variation, temperature change, precipitant addition, and agitation
→ Phase equilibrium model to characterize multidimensional solubility surfaces
• Slower nucleation & growth rates than small-molecule drugs→ Continuous-flow crystallizers must be designed to operate
at high crystal surface area
Control of Biopharmaceutical Crystallization
First-principles approach• Optimizing objective function
related to the crystal size distribution by using a first-principles model
• Optimal experimental design and data collection
–Parameter estimation–Mechanism selection
• Robust optimization–Worst-case–Weighted
Fujiwara, M., Nagy, Z. K., Chew J. W., Braatz, R. D. (2005). First-Principles and Direct Design Approaches for the Control of Pharmaceutical Crystallization. Journal of Process Control, 15, 493-504.
Control of Biopharmaceutical Crystallization
Direct design approach• Feedback control to follow a
setpoint supersaturation curve in the experimentally determined metastable zone
• Concentration control (C-control)
–Manipulating the temperature (cooling crystallization)
–Solvent addition (antisolventcrystallization)
Fujiwara, M., Nagy, Z. K., Chew J. W., Braatz, R. D. (2005). First-Principles and Direct Design Approaches for the Control of Pharmaceutical Crystallization. Journal of Process Control, 15, 493-504.