QBD Models 2011

Genome Technology in Industrial Cell Bioprocessing
Toward QbD in Cell Culture Biologics
From analysis to design:

Models for metabolism and glycosylation
Ziomara P. Gerdtzen - zgerdtze@ing.uchile.cl
Departament of Chemical Engineering and Biotechnology
University of Chile
October 16, 2011
Introduction
Mathematical model development
Models for Mammalian Cells
Final Remarks
Systems Biology
Mammalian Cells
Systems biology overview
There has been a growing interest in addressing the study of biochemical

networks and cellular processes in a comprehensive way.
A systems approach assumes that the behavior of a system can be
described as a direct consequence of the interactions of its components,
and considers all the components and reactions involved in a particular
cellular process. Those reactions, ranging from DNA, RNA, protein
synthesis and their regulation to biochemical conversions, form complex
interaction networks.
Ziomara P. Gerdtzen
From analysis to design: Models for metabolism and glycosylation
Introduction
Final Remarks
Systems Biology
Mammalian Cells
Modeling, computation and experimental research data from

metabolomics, genomics and proteomics can be combined to generate a
meaningful description of a cellular process or biochemical network.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Systems Biology
Mammalian Cells

Different levels of information
A set of genes, metabolites,

proteins, with a specific copy
number or concentration defines
the physiology an organism (level
1). These components form
genetic- regulatory motifs or
metabolic pathways (level 2),
which are the building blocks of
functional modules (level 3).
Nested modules generate a
hierarchical architecture (level 4).
Ziomara P. Gerdtzen
Introduction
Final Remarks
Systems Biology
Mammalian Cells
Mammalian Cells
In the pharmaceutical industry, nearly 70% of the glycoproteins with

pharmacologic application are produced in mammalian cell cultures using
systems such as
Chinese hamster ovary cells (CHO)
Mouse myeloma (NS0)
Baby hamster kidney (BHK)
Human embryo kidney cells (HEK-293)
since they provide clear advantages over other hosts such as bacteria,
yeast or fungi.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Systems Biology
Mammalian Cells
Summary
The paradigm for the analysis of cellular systems has shifted from a focus
on individual processes to comprehensive descriptions that consider the
interactions of metabolic, genomic and signaling networks.
Different types of models can be used for this purpose, depending on the
type and amount of information available for the specific system.
The development of detailed models for mammalian cell metabolism has
only recently started to grow more strongly, due to the intrinsic
complexities of mammalian systems, and the limited availability of
experimental information and adequate modeling tools.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Systems Biology
Mammalian Cells
The system: Mammalian cells
Complex compartmentalized cellular

systems that perform a large number of
biochemical reactions simultaneously,
subject to a series of thermodynamic
constraints.
Crowded intracellular environment where
many interactions take place.
Internal spatial complexity,
non-homogeneous distribution.
Parameters are a function of growth rate.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Systems Biology
Mammalian Cells
The system: Mammalian cells

Characteristics of animal cells vs. E. coli
Characteristic
Diameter
Volume
Cell generation time
Genome size
Number of genes
RNA content
RNA lifetime
DNA content
Protein content
Proteins/cell
Diffusion time of protein across cell
Carbohydrate content
Lipid content
Compartmentalization
Ziomara P. Gerdtzen
Mammalian cell
18-20 m
10,000 m3
20 h - non dividing
3.0 109 bp
25000-30000
25 g/106 cell
2-5 min
10 g/106 cell
250 g/106 cell
E. coli
0.8-2.0 m
1 m3
30 min to h
4.6 106 bp
3200
20-211 g/109 cell
10 min-10 h
7.6-18 g/109 cell
25-130 g/109 cell
41010
100 sec
150 g/106 cell
120 g/106 cell
4106
0.1 sec
4.4-26 g/109 cell
3-17.3 g/109 cell
Yes
No
Introduction
Final Remarks
Genome scale models

Resources
Physicochemical models
Reconstruction tools
Developing a mathematical model

Identification of components of the system to
be modeled.
Assessment of interactions and directionality.
Sequencing and Annotation
Literature
G A G G C G G T T T G C G T A T T
Phylogenetic Data
Construction of a mathematical description or

model fitting.
Experimental Data
Experimental measurement of parameters, or

use of literature and database sources.
Databases
Model allows organization of information

available at different levels (genomic,
proteomic, metabolomic).
Ziomara P. Gerdtzen
Concentration [mM]
Systems network assembly (data repositories,

automated reconstruction).
Time [h]
Localization
Introduction
Final Remarks
Genome scale models

Resources
Genome scale model reconstruction

A protocol for generating a high-quality genome scale metabolic reconstruction
Thiele and Palsson 2010, Nature Protocols 5(1):93-121
Bottomup metabolic network reconstructions represent structured

knowledge bases that abstract pertinent information on the biochemical
transformations taking place within specific target organisms.
Conversion of a reconstruction into a mathematical format facilitates
computational biological studies, including evaluation of network content,
hypothesis testing and generation, analysis of phenotypic characteristics
and metabolic engineering.
Genome-scale metabolic reconstructions for more than 50 organisms
have been published and this number is expected to increase rapidly.
However, these reconstructions differ in quality and coverage that may
minimize their predictive potential and use as knowledge bases.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Genome scale model reconstruction

A protocol for generating a high-quality genome scale metabolic reconstruction
Thiele and Palsson 2010, Nature Protocols 5(1):93-121
1. Draft reconstruction
Obtain genome annotation
Metabolic functions and reactions
Assemble draft reconstruction
Collect experimental data
2. Renement of reconstruction
- For all genes:
Substrate and cofactor usage
Neutral formula for each metabolite
Charged formula
Reaction stoichiometry and directionality
Add gene and reaction localization
Add subsystems information
Verify gene-protein reaction association
Add metabolite identier
Determine and add condence score
Add references and notes
Flag information from other organisms
Add spontaneous reactions
Add transport and exchange reactions
Draw metabolic map
Determine biomass composition, add biomass reaction
Determine maintenance, requirements and sink reactions
Data assembly and dissemination

Print Matlab model content
Add gap information to reconstruction output
4. Network evaluation
Test mass and charge balance
Identify metabolic dead-ends
Perform gap analysis
Add missing exchange reactions and constraints
Test stoichiometrically balanced cycles
Re-compute gap list
Test precursors in standard/other medium
Test production of secretion products
Check blocked reactions
Compute deletion phenotypes
Test organism in/capabilities
Test growth
3. Conversion to computable format
Initialize COBRA toolbox
Load reconstruction to Matlab
verify S matrix
Set objective function
Set simulation constraints
Hybridoma Comprehensive protocol describing steps for building a high-quality genome-scale metabolic
reconstruction.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Reconstructed models for mammalian cells

Computational prediction of human metabolic pathways from the complete human genome
Romero et al. 2004, Genome Biology 6:R2
Computational pathway analysis of the human genome that assigns

enzymes encoded therein to predicted metabolic pathways.
2,709 human enzymes asigned to 896 bioreactions.
Predicted pathways closely match the known nutritional requirements of
humans.
203 pathway holes on genome annotation.
Predicted metabolic map described by the Pathway/Genome Database
called HumanCyc, available at http://HumanCyc.org/.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources

Computational prediction of human metabolic pathways from the complete human genome
Romero et al. 2004, Genome Biology 6:R2
Elucidated the human metabolic map. The database places many human genes in a pathway context, thereby
facilitating analysis of gene expression, proteomics, and metabolomics datasets through Cellular Overview.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources

Modeling Hybridoma Cell Metabolism Using a Generic Genome-Scale Metabolic Model of Mus
musculus
Sheikh et al. 2005, Biotechnol.Prog.21:112
Reconstructed cellular metabolic network of Mus musculus, based on

annotated genomic data, pathway databases, and currently available
biochemical and physiological information
Collect and characterize the metabolic network of a mammalian cell on
the basis of genomic data.
Generic reaction network. Attempts to capture the carbon, energy, and
nitrogen metabolism of the cell. Compartmentalization and transport
included.
The reaction list consists of 872 internal metabolites involved in a total of
1220 reactions.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Web resources and databases for components and pathways

for mammalian cells
Name
AfCS
APID
BioCyc
BioGRID
BOND
BRENDA
DIP
GeneNet
GenMAPP
Web address and description

www.afcs.org/index.html - Protein-protein interactions and modifications.
bioinfow.dep.usal.es/apid - Open access web application where experimentally validated protein-protein interactions are unified.
biocyc.org - Collection of genome and metabolic pathways for a large number of
organisms, including mammalian.
thebiogrid.org - Public repository with genetic and protein interaction data from
model organisms and humans.
bond.unleashedinformatics.com - Comprehensive information on small molecule
and protein interactions.
www.brenda-enzymes.org - Freely available comprehensive collection of biochemical,molecular and metabolic information on all classified enzymes, based on literature.
dip.doe-mbi.ucla.edu - Experimentally determined protein-protein interactions.
wwwmgs.bionet.nsc.ru - Integrates databases on gene network components and
data processing tools for structure and function of DNA, RNA, and proteins.
www.genmapp.org - Computer application for analysis and visualization of
genome-scale data for rapid interpretation.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources

for mammalian cells
Continued
Name
HPID
IBIS
iHOP
Ii2D
IntAct
KDBI
KEGG BRITE
KEGG PATHWAY
LOCATE

wilab.inha.ac.kr/hpid - Integrates information from other databases on proteins and
their interactions. Predicts potential interactions between human proteins.
www.ncbi.nlm.nih.gov/Structure/ibis/ibis.cgi - Protein interactions, both experimentally determined and inferred by homology.
www.ihop-net.org - Protein association network built by literature mining.
ophid.utoronto.ca - Database with known, experimental and predicted PPIs for five
model organisms and human.
www.ebi.ac.uk/intact - Open source database system and analysis tools for protein
interaction data.
xin.cz3.nus.edu.sg/group/kdbi/kdbi.asp - Kinetic data of interactions between proteins, RNA, DNA and ligands, or reaction events.
www.genome.jp/kegg/brite - Contains functional hierarchies and binary relationships between biological objects.
www.genome.jp/kegg/pathway.html - Collection of pathway maps for a large number of organisms.
locate.imb.uq.edu.au - Curated database describing membrane organization and
sub cellular localization of proteins from mouse and human.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources

for mammalian cells
Continued II
Name
MINT
MIPS
MetaCyc
Predictome
PRIME
PSIbase
Sigmoid
SGMP
SIMPATHWAY
STRING
ResNet

mint.bio.uniroma2.it/mint - Contains experimentally verified protein-protein interactions.
mips.helmholtz-muenchen.de/proj/ppi - Collection of manually curated high-quality
PPI data from the scientific literature.
metacyc.org - Experimentally determined pathways and enzymes.
visant.bu.ed - Predicted functional associations among genes and proteins.
prime.ontology.ims.u-tokyo.ac.jp - Contains integrated gene/protein information automatically extracted.
psibase.kobic.re.kr - Molecular interaction database.
www.sigmoid.org - Cellular signaling and metabolic pathways models.
www.signaling-gateway.org/molecule - Proteins involved in cellular signaling.
www.helios-bioscience.com/technologies-molecular.php - Commercial molecular
interaction database. Includes an integrated intracellular signal transduction network, query and data representation software.
string.embl.de - Database of known and predicted protein interactions.
www.ariadnegenomics.com/products/databases - Commercial database of molecular interactions for human, rat and mouse.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Developing a mathematical model

Continued
There are a number of different representations of metabolism that can be

used, which vary in the level of of detail, the level of available information,
and the type of predictive capacity of the model.
Essential to define beforehand the purpose and scope of the model,
assumptions and restrictions
The modeling strategy to be followed depends on the final objective of the
analysis, the characteristics of the experimental data and the available
details of the structure of the network.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Physicochemical and Statistical models

Physicochemical Models
Describe the different kinds of molecular transformations in biological

systems (association, translocation and modification through reactions).
Each equation in the model can be associated to a specific process in the
system and each parameter assigned a physical interpretation.
For well studied organisms, a high level of detail is available in terms of
pathway components and interactions. In this case the mathematical
model is only a translation of the physical description of the system to
equations.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Physicochemical and Statistical models

Statistical Models
Abundant but intricate data, and a scarce amount of information available

about the structure of the system, data can be used to obtain some
insight about the system using data-driven statistical modeling.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Stages for the development of a mechanistic model

Network
Elements
Pathway Diagram
Reaction List
Dynamics
1
GF
GF
GF
PP
GF
PP
GF
PP
TF
Response
0.75
TF
Time
Approximations
P
0.25
0.5
Equations
K
TF
TF
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
General form of a dynamic mass balance for metabolites j
dCj
dt
= Sr(C,p)
C is the concentrations vector (n 1)

Cj is the concentration of metabolite j
r is the vector of metabolic reaction rates (m 1) which is a function of the metabolites
concentration vector (n 1) and a parameter vector p,
S is the stoichiometric matrix that contains the stoichiometric coefficients that relate the n species
in the m reactions that form the network. This generates a system of n differential equations and
m unknown time-dependent fluxes.
Under steady state assumption
Sr(C,p) = 0
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Model SEED
High-throughput generation, optimization and analysis of genome-scale metabolic models
Henry et al. 2010, Nature Biotechnology 28:977
Genome-scale metabolic models valuable for predicting organism

phenotypes from genotypes.
Model SEED: accelerate new models development with genome
sequencing.
Web-based resource for high-throughput generation, optimization and
analysis of genome-scale metabolic models. Automated from assembled
genome sequence.
Applied to generate 130 genome-scale metabolic models. 22 validated
against available data, average model accuracy of 87% after optimization.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Model SEED
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Model SEED
Research question
What are the essential genes in my
newly sequenced organism?
What defined culture conditions will
my organism grow in?
What are some global trends in microbial metabolic behavior?
How accurate are the annotations for
my organism of interest?
What are the knowledge gaps in
genome annotation in general?
What alternative pathways are
present in an organisms metabolic
reaction network?
How can I identify and fill the gaps in
my genome annotations?
Unique capability of Model SEED

Functioning draft models enable essential genes to
be predicted.
Functioning metabolic models enable culture conditions to be predicted.
Functioning draft models for many diverse microbes
enable the exploration of such trends.
Functioning models convert annotations into predictions of experimentally observable phenotypes.
Recurring annotation gaps can be identified by comparing gaps found in every model.
Comprehensive reaction database, functional role
mappings and updated annotations enable identification of alternative pathways.
Directed searches may be performed for functions
added during model auto-completion and optimization.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Pathway Tools
PathwayTools version 13.0: integrated software for pathway/genome informatics and systems
biology
Karp et al. 209, Briefings in Bioinformatics 11:40
Software environment for creating a Pathway/Genome Database (PGDB).

PGDB such as EcoCyc integrates the understanding of genes, proteins,
metabolic network and regulatory network of an organism.
Multiple computational inferences: prediction of metabolic pathways,
metabolic pathway hole fillers and operons.
Interactive editing of PGDBs by DB curators. It supports web publishing of
PGDBs, and provides query and visualization tools.
More than 800 PGDBs have been created using Pathway Tools, including
curated DBs for model organisms. PGDBs can be exchanged using a
peerto- peer PGDB Registry DB sharing system.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Ziomara P. Gerdtzen
Genome scale models

Resources
Introduction
Final Remarks
Genome scale models

Resources
Modeling resources and databases for systems biology

research in mammalian cells
Name
BioModels

biomodels.net - Model collection in SML format. Includes data resources and publication references. Supports model visualization.
CellDesigner
www.celldesigner.org - Diagram editor for drawing gene-regulatory and biochemical
networks. Models are stored in SBML format.
Cellerator
www.cellerator.org - Mathematica package for automatic generation of differential equations of biochemical networks.
CellNetAnalyzer www.mpi-magdeburg.mpg.de/projects/cna/cna.html - MATLAB based software package for structural and functional analysis of networks based on their topology.
CellWare
www.bii.a-star.edu.sg/achievements/applications/cellware - Grid based tool for modeling, simulation, parameter estimation and optimization.
COBRA
systemsbiology.ucsd.edu/downloads/COBRAToolbox - Matlab package for quantitative
prediction of cellular behavior using a constraint-based approach.
COPASI
www.copasi.org - Supports models in the SBML standard. Uses ODEs or Gillespies
stochastic simulation algorithms.
DOQCS
doqcs.ncbs.res.in - Database of Quantitative Cellular Signaling. Repository of models
of signaling pathways. It includes reaction schemes, concentrations, rate constants, as
well as annotations on the models.
E-Cell
www.e-cell.org - Platform for modeling, simulation and analysis of complex systems.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Genome scale models

Resources
Modeling resources and databases for systems biology

research in mammalian cells
Continued
Name
JDesigner
Metatool
OptFlux
Pathway
Analyser
SBRT
SNA
STOCKS
Virtual Cell
WebCell
YANAsquare

www.sys-bio.org - Visual network design tool for systems biology.
pinguin.biologie.uni-jena.de/bioinformatik/networks - Program for calculating elementary modes, compatible with octave and Matlab. Distributed with CellNetAnalyzer.
www.optflux.org - Tools for in silico metabolic engineering.
sourceforge.net/projects/pathwayanalyser - Flux based analyses and simulations on
SBML Models.
www.ieu.uzh.ch/wagner/software/SBRT - Systems Biology Research Tool. Software
platform for analyzing stoichiometric networks.
www.bioinformatics.org/project/?group_id=546 - Toolbox for steady state analysis of
metabolic networks.
www.sysbio.pl/stocks - Software for stochastic simulation of biochemical processes with
Gillespie algorithm. Supports SBML.
www.vcell.org - Web-based environment for modeling and simulation of cell biology.
webcell.kaist.ac.kr - Integrated simulation environment for the analysis of cellular networks over the web.
yana.bioapps.biozentrum.uni-wuerzburg.de - Software package for the analysis of
metabolic networks. It incorporates database extraction and visualization tools.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Hybridoma Model
musculus
Few works associated with metabolic models of mammalian cellsdue to the

complexity of mammalian cell metabolism, reduced availability of information
about the system and difficulties in measuring in vivo metabolic fluxes in
mammalian cells.
A reconstruction of the cellular metabolic network of Mus musculus was
presented recently by Sheikh et al.
Based on annotated genomic data, pathway databases, and currently
available biochemical and physiological information from KEGG
It captures carbon, energy, and nitrogen metabolism in a
compartmentalized setting, including transport reactions between the
compartments and the extracellular medium.
It considers 872 internal metabolites and 1220 reactions.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Hybridoma Model
musculus
MFA for underdetermined system.

Linear programming considering three objective functions: maximization
of cell growth, minimization of substrate uptake rate, and maximization of
production of monoclonal antibody.
Predicts growth, lactate, and ammonia production given glucose, oxygen,
and glutamine uptake, but it fails to predict alanine production, illustrating
the limitations of the model.
Improved by Selvarasu et al. including biomass and mAb synthesis as
well as updated lipid, amino acids and nucleotide metabolic pathways.
Strategies for increased cell density and mAb productivity identified.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Improved Hybridoma Model

Genome-scale modeling and in silico analysis of mouse cell metabolic network
Selvarasu et al. 2010, Mol.BioSyst. 6:152
Improved by Selvarasu et al. including additional information on

geneprotein-reaction association, and improved network connectivity
through lipid, amino acid, carbohydrate and nucleotide biosynthetic
pathways. Considers biomass and mAb synthesis.
Strategies for increased cell density and mAb productivity identified.
In silico mouse model can be exploited for understanding and
characterizing cellular physiology, identifying potential cell engineering
targets for the enhanced production of recombinant proteins and
developing diseased state models for drug targeting.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Improved Hybridoma Model

Genome-scale modeling and in silico analysis of mouse cell metabolic network
Selvarasu et al. 2010, Mol.BioSyst. 6:152
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
CHO Model
A detailed metabolic flux analysis of an underdetermined network of CHO cells
Zamorano et al. 2010, Journal of Biotechnology 150:497
MFA for CHO cells recently was performed by Zamorano et al., 2010.
Network involving 100 reactions.
Used to assess the efficiency of flux analysis when using a small set of
extracellular measurements (underdetermined mass balance system)
Narrow intervals found for most fluxes.
Confirmed by Goudar et al. (Metabolic flux analysis of CHO cells in perfusion culture by metabolite balancing
). The fluxes at the pyruvate
branch point were almost equally distributed between lactate and the TCA
cycle (55% and 45%, respectively).
and 2D [13 C, 1 H] COSY NMR spectroscopy, 2010 Metabolic Engineering 12:138
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
CHO Model
Metabolic Flux Analysis of CHO Cell Metabolism in the Late Non-Growth Phase
Sengupta et al. 2011, Biotechnology and Bioengineering 108:82
Stoichiometric model for CHO cells used in combination with steady-state

isotopomer balancing to evaluate flux distribution in the late non-growth
phase.
Almost all of the consumed glucose is diverted towards PPP with a high
NADPH production. Almost all of the pyruvate produced from glycolysis
entered the TCA cycle with little or no lactate production.
Similar conclusion from Ahn et al. (Metabolic flux analysis of CHO cells at growth and non-growth phases
using isotopic tracers and mass spectrometry, 2011 Metabolic Engineering 13:598) for CHO cells at the growth
phase and early stationary phase in fed-batch culture, using a isotopic
tracers and a compartmentalized metabolic network model of CHO cell
metabolism. Changes in metabolic fluxes were identified.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
CHO Model
Comparative Metabolic Analysis of Lactate for CHO Cells in Glucose and Galactose
Wilkens et al. 2011, Biotechnology and Bioprocess Engineering 16:714
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Dynamic Model for Hybridoma

Modeling Amino Acid Metabolism in Mammalian Cells-Toward the Development of a Model Library
Kontoravdi et al. 2007, Biotechnol.Prog.23:1261
Single unstructured model structure for describing the cell growth kinetics
and metabolism of HEK-293 and CHO cells.
The network considered is consistent with the information available in
literature sources and the pathways available in KEGG.
ATP
Glutamine Histidine
Proline Arginine
Glucose
-KGT
Glutamate
NAD+
NADH
Lactate
NAD+
Lysine
Leucine
-KGT
-KGT
-ketoglutarate
Threonine
Acetyl CoA
-KGT
Glu
Oxaloacetate
TCA
cycle
Valine
Glu
Cysteine
-KGT
Phenilalanine
Serine
Glycine
-KGT
Methionine
Glu
Pyruvate
Glu
Isoleucine
Glu
Succinyl CoA
Glu
-KGT
-KGT
Glu
NADH
Alanine
Biomass
Glu
Asparagine
Aspartate
Malate
Fumarate
Glu
Ziomara P. Gerdtzen
Tyrosine
-KGT
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models

Dynamic model of CHO cell metabolism
Nolan et al. 2011, Metabolic Engineering 13:108
Detailed model for CHO cells and a framework for simulating the
dynamics of metabolic and biosynthetic pathways of CHO cellsin
fed-batch.
It Considers the effects of temperature shift, seed density, specific
productivity, and metabolite concentrations on viable cell density (VCD),
antibody, lactate, asparagine, and the redox state.
The model defines a subset of intracellular reactions with kinetic rate
expressions , which are used to calculate pseudo-steady state flux
distributions and extracellular metabolite concentrations at discrete time
points.
The model provides time profiles for all metabolites in the reactor and
successfully predicts the effects of several process perturbations on cell
growth and product titer.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models

Dynamic model of CHO cell metabolism
Nolan et al. 2011, Metabolic Engineering 13:108
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Protein Glycosilation
Protein glycosylation is a post-translational modification of high

importance towards the function, immunogenicity, and efficacy of
recombinant glycoprotein therapeutics. Affects pharmacokinetics and
protein physiochemical characteristics.
Obtaining a consistent glycoform profile in production is desired due to
regulatory concerns.
Protein glycosylation control needs to be studied on a case-by-case basis
since there are often conflicting results with respect to a control variable
and the resulting glycosylation.
Gene expression analysis and systems biology have been used for a
multivariate interpretation of potential control variables.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Glycosilation
Optimal and consistent protein glycosylation in mammalian cell culture
Hossler et al. 2009, Glycobiology 19:936
Studies have shown some degree of protein glycosylation control in

mammalian cell culture, through cellular, media, and process effects.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Protein Glycosilation
Gene expression analysis and systems biology
Gene expression analysis has provided information as to how

glycosylation pathway genes both respond to culture environment and
facilitate changes in the glycoform profile.
Systems biology has allowed researchers to model the glycosilation
pathway as well defined, inter-connected systems, allowing for the in
silico testing of pathway parameters that would be difficult to test
experimentally.
Important tools for the production of optimal glycoform profiles on a
consistent basis.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Glycosilation Model
Systems Analysis of N-Glycan Processing in Mammalian Cells
Hossler et al. 2007, PLoS ONE 2(8):e713
Glycosilation pathway: network in which a relatively small number of

enzymes give rise to a large number of N-glycans as the reaction
intermediates and terminal products.
Mathematical model of glycan biosynthesis in the Golgi and analysis of
reaction variables on the resulting glycan distribution.
Four continuous mixing-tanks (4CSTR) and four plug-flow reactors
(4PFR) in series.
A sufficient holding time is needed to produce terminally-processed
glycans. Altering enzyme concentrations has a complex effect on the final
glycan distribution, as many reaction steps in the network are affected.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Glycosilation model
Systems Analysis of N-Glycan Processing in Mammalian Cells
Hossler et al. 2007, PLoS ONE 2(8):e713
Protein transport
Cytoplasm
cis
medial
trans
TGN
Model was used to assess whether a homogeneous glycan profile can be

created through metabolic engineering.
Results indicate that by the spatial localization of enzymes to specific
compartments, all terminally processed N-glycans can be synthesized as
homogeneous products with a sufficient holding time in the Golgi
compartments.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
CHO GT
Platform
ImprovingMammalianCellLineProteinProductionusingaMetabolic
ModelbasedApproach
AdamFeist,Ph.D.,JeffreyRosenbloom,RenataUsaiteBlack,Ph.D.,ImanFamili,Ph.D.
Abstract
ModelDrivenDesignandDiscovery:
ModelDrivenMediaOptimizationResults:
Duplicateshakeflasks,batch&fedbatchexperimentswereperformed:
Baseline: cellculturewithanalreadyoptimizedmediaformulation,i.e.
theCDOptiCHObasalandEfficientFeedA(Invitrogen ,Corp.)
Depletion:
l i
d l d
depletednutrientsaddedbackinusingspentmediaanalysis
i
dd d b k i
i
di
l i
Designs13: modelbasedformulations
Recent advancements in experimental and bioinformatics

technologies have contributed immensely to the expansion of the
knowledge base for cell components and their interactions. This
information can be leveraged to study networks at a wholecell level
using in silico network reconstruction and computational modeling.
Computational methods for analyzing cell metabolism and physiology
have proven successful in discovering biological components and their
interactions, predicting physiological responses to genetic and
environmental changes, and improving product formation for
industrial and medical applications. Using a validated computational
metabolic modeling platform and advanced experimental
technologies, we have simulated and studied metabolism in NS0
murine myeloma and Chinese Hamster Ovary (CHO) cell lines and
identified nutritional and process modifications that result in
significant growth and titer improvements and byproduct reduction.
Experimental validation of our modelbased media formulations and
process design has proven to be effective in improving the final
product titers and reducing byproduct accumulation when compared
to traditional media optimization methods. We have also used this
approach to discover and develop novel selection systems for stable
expression of therapeutic proteins in mammalian cell lines (see other
poster). The results demonstrate that metabolic modeling combined
with advanced experimental technologies can significantly improve
and accelerate discovery and development of therapeutic proteins in
mammalian cell lines.
ModeldrivenMediaOptimizationAdvantages:
Mechanisticunderstandingandcalculationofgrowthandproduction
requirements
Rapidhypothesesandexperimentalimplementation
Integratedcomputationalandexperimentalplatform
GTLifeSciencesComputational&Experimental
PlatformanditsApplications:
ModelDrivenMediaOptimization:
MediaFormulationandDesign
ExperimentalResults BatchMediaOptimization
Baseline
Value
GenomescaleModelofCHO:
Genome
709reactions,626metabolites,1100+metabolic genes
Fullycompartmentalized(mitochondria,ER,golgi,etc.)
Elementallyandchargebalanced reactions
Developedusingwholetranscriptome sequencingdata
g
g
q
g
Tobeaugmentedsoonwithwholegenomesequencing
dataincollaborationwithBeijingGenomicsInstitute(BGI)
ExperimentalResults FedBatchMediaOpt
Baseline
Value
Goal:Toincreaseproteinproductionandgrowth,andtodecrease
byproductformationinahighproducingCHOcelllineusingamodel
drivenmediaoptimization
FormulatingtheSimulation:
g
ALinearProgrammingOptimizationProblem
Conclusions
Sample Media Design:

SampleMediaDesign:
Anexampleofhowtwomodel
basedformulations(outofa
totalof5examinedforbasal
andfeedcompositions)differed
fromadepletionbasedformula
Nutrients
Metab A
Metab B
MetabC
MetabD
Metab E
Metab F
Metab G
Metab H
Depletion Model
Depletion
Modelbased
based
based Formula
Design1
+
+
+
+
+
+
+
+
Model based
Modelbased
Design3
+
+
+
+
+
+
Themodelbasedmediaformulationssignificantlyoutperformedthe
control,aswellastheindustrystandarddepletionanalysisinone
iteration
Th
Theformulationsweredesignedandtestedrapidly
f
l ti
d i
d dt t d
idl usingagenome
i
scalemodelofCHOmetabolism
Theformulationswerenovel andnonintuitive
TheresultsinCHOareconsistentwiththosemadeinNS0 cellline
Theapproachisbeingappliestobioreactor studies,alongwith
processoptimizationandproductquality(i.e.glycosylation)
GTLifeSciences,Inc. 10520WateridgeCircle SanDiego,CA92121USA www.gtlifesciences.com info@gtlifesciences.com
Ziomara P. Gerdtzen
Introduction
Final Remarks
Steady state models

Dynamic models
Glycosilation
Commercial models
Virtual Liver
Home | Contact us | Careers
Products
Technology
Community
Company
Home > Technology > Virtual Liver
Virtual Liver
TECHNOLOGY
Virtual Liver
AVADIS
Liver Injury is the number one cause of withdrawal of drugs post marketing. Such withdrawal results in enormous damages to
pharmaceutical and biotechnological companies, in terms of finances as well as reputation. Traditional animal toxicity testing
methods have proved to be insufficient when it comes to predicting toxicity observed in the clinic.
What the industry needs today is a robust predictive method that integrates data and insights from multiple in vitro methods. This
alone will result in a deeper understanding of the impact of a drug on the liver. To this end, Strand's Virtual Liver' is a unique systems
approach based on mathematical modelling of the kinetics of essential biochemical pathways involved in liver homeostasis.
Strand's Virtual Liver allows researchers to understand the evolution of DILD, i.e. what are the pathways impacted and how this
impact translates over the short and long term into biological changes. It also enables early identification of markers of a specific form
of toxicity and to use the same platform to predict the impact of both small molecule and biological agents.
Virtual Liver features include:

Scientifically Advanced Hepatotoxicity prediction
25 man-years of R&D have already been performed in developing the platform. The platform can predict toxicity of several
known drugs and toxins well
This includes prediction of hepatotoxicity due to necrosis, cholestasis and steatosis, which are the major forms of liver
toxicity observed in the clinic
Combines in silico and wet-lab techniques
Support for Toxicogenomic Data
Modelling Methodology Expertise
Flexible Business Model
Ziomara P. Gerdtzen
Introduction
Final Remarks
Final Remarks
Development of dynamic and genome based models for mammalian cells

require abundant high quality experimental data, software, algorithms and
visualization techniques.
Mathematical models can assist on strain design and complement
metabolic methods in order to increase productivity and improve process
and product quality. Available models are limited due to the scope of the
solutions, and availability and accuracy of the data used on their
construction.
The metabolic networks modeling field will continue to grow. Important
applications in predicting mammalian cell behavior for application in
productive processes as well as on the medical field.
Ziomara P. Gerdtzen
Introduction
Final Remarks
Questions
Ziomara P. Gerdtzen

QBD Models 2011

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

QBD Models 2011

Uploaded by

Copyright:

Available Formats

Genome Technology in Industrial Cell Bioprocessing

Toward QbD in Cell Culture Biologics

From analysis to design:

Systems biology overview

There has been a growing interest in addressing the study of biochemical

From analysis to design: Models for metabolism and glycosylation

Systems biology overview

Modeling, computation and experimental research data from

From analysis to design: Models for metabolism and glycosylation

Systems biology overview

A set of genes, metabolites,

From analysis to design: Models for metabolism and glycosylation

In the pharmaceutical industry, nearly 70% of the glycoproteins with

From analysis to design: Models for metabolism and glycosylation

From analysis to design: Models for metabolism and glycosylation

The system: Mammalian cells

Complex compartmentalized cellular

From analysis to design: Models for metabolism and glycosylation

The system: Mammalian cells

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Developing a mathematical model

Sequencing and Annotation

Construction of a mathematical description or

Experimental measurement of parameters, or

Model allows organization of information

Systems network assembly (data repositories,

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Genome scale model reconstruction

Bottomup metabolic network reconstructions represent structured

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Genome scale model reconstruction

Data assembly and dissemination

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Reconstructed models for mammalian cells

Computational pathway analysis of the human genome that assigns

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Reconstructed models for mammalian cells

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Reconstructed models for mammalian cells

Reconstructed cellular metabolic network of Mus musculus, based on

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Web resources and databases for components and pathways

Web address and description

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Web resources and databases for components and pathways

Web address and description

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Web resources and databases for components and pathways

Web address and description

From analysis to design: Models for metabolism and glycosylation

Genome scale models

Developing a mathematical model

There are a number of different representations of metabolism that can be

From analysis to design: Models for metabolism and glycosylation