Toward whole-cell models for science and engineering

Jonathan Karr March 9, 2015

Positions available research.mssm.edu/karr

karr@mssm.edu

Jayodita Sanghvi

Markus Covert

Jared Jacobs Derek

Macklin

Acknowledgements

Chemicals & fuels

Optimize yield Minimize cost

Food

Optimize yield Resist drought

Prevent infection

Medicine

Predict prognoses Optimize therapy

Maximize quality of life

Central challenge: predict phenotype from genotype

Example: drug biosynthesis

Example: drug biosynthesis

Example: drug biosynthesis

Example: drug biosynthesis

Example: drug biosynthesis

Example: drug biosynthesis

Example: drug biosynthesis

Example: drug biosynthesis

Predicting phenotype from genotype requires “whole-cell” models

Integrated

Comprehensive Dynamic

Gene-complete

Whole-cell modeling principles

Biological data is readily available

?

Data Knowledge

Whole-cell model goals

Whole-cell modeling

A grand challenge of the 21st century – Masaru Tomita

Biology urgently needs a theoretical basis to unify it – Sydney Brenner

The ultimate test of understanding a simple cell, more than being able to build one, would be to build a computer model of the cell – Clyde Hutchison

Single-cell variation Microscopy

Transcription RNA-seq

Protein expression Mass-spec, Western blot

Modeling challenge: heterogeneous data

Modeling challenge: sparse data

Metabolic Signaling

Transcriptional regulatory

Modelling challenge: heterogenous networks

Time

Len

gth

Replication

Growth

Transcription

Metabolism

Modeling challenge: multiple time and length scales

0

25

50

75

100

1970's

Coarse-grained

ODEs

1990's

FBA

2000's

Boolean

models

2008

iFBA

2012

Whole-cell

model

% a

nnota

ted g

enes

Whole-cell modeling progress

v v v v v

Predictive modeling methodologies

Granularity

Sco

pe

ODE

SDE

FBA

Boolean

Bayesian

Gillespie

PDE

Whole-cell model

Uptake FBA

Composition

Metabolism FBA

Composition

Transcription Stochastic binding Gene expression

Translation Stochastic binding Gene expression

Replication Chemical kinetics DNA sequence

Solution: integrated models

0

25

50

75

100

1970's

Coarse-grained

ODEs

1990's

FBA

2000's

Boolean

models

2008

iFBA

2012

Whole-cell

model

% a

nnota

ted g

enes

Whole-cell modeling progress

v v

Model Validate

Engineer

Whole-cell modeling

Validate

Engineer

Model

Whole-cell modeling

Model construction

1. Define system

2. Define scope

3. Curate data

4. Choose representation

5. Identify parameters

6. Test predictions

E. coli M. genitalium

Genome 4700 kb 580 kb

Genes 4461 525

Size 2 μm × 0.5 μm 0.2-0.3 μm

1. Select a tractable model organism

Comparative genomics Fraiser et. al, 1995

Genome-wide essentiality Glass et. al, 1999

M. genitalium is well-characterized

Genomic-scale data Kühner et. al, 2009

M. genitalium is well-characterized

Genomic transplantation Lartigue et. al, 2009

Genomic synthesis Gibson et. al, 2009

M. genitalium has unique engineering tools

2. Choose model scope

2. Choose model scope

• Explicitly represent each metabolite, gene, RNA, and protein species

• Explicitly model the function of every characterized gene product

•Account for the metabolic cost of every uncharacterized gene product

•Represent important, well-characterized molecules individually

3. Broadly curate experimental data

Karr et al., 2013

Uptake FBA

Composition

Metabolism FBA

Composition

Transcription Stochastic events Gene expression

Translation Stochastic events Gene expression

Replication Chemical kinetics DNA sequence

Sub-models States

4. Select a flexible mathematical representation

Mass, shape

Metabolite, RNA, protein counts

Mammalian host

Transcript, polypeptide sequences

DNA polymerization, proteins, modifications

FtsZ ring

1 s

Simulation algorithm

Uptake

Metabolism

Transcription

Translation

Replication

Cel

l st

ates

Cel

l st

ates

Uptake

Metabolism

Transcription

Translation

Replication

Cel

l st

ates

Uptake

Metabolism

Transcription

Translation

Replication

DNA RNA Protein Other

Rep

licat

ion

Rep

Initia

tion

Super

coiling

Conden

sation

Seg

regat

ion

Dam

age

Rep

air

Tra

ns

Reg

T

ransc

ription

Pro

cess

ing

Modific

atio

n

Am

inoac

ylat

ion

Deg

radat

ion

Tra

nslat

ion

Pro

cess

ing I

Tra

nsloca

tion

Pro

cess

ing II

Fold

ing

Modific

atio

n

Com

ple

xation

Rib

oso

me

Ter

m O

rg

Act

ivat

ion

Deg

radat

ion

Met

abolis

m

Shap

e

Fts

Z

Cyt

oki

nes

is

DN

A

Replication

Rep Initiation Supercoiling

Condensation Segregation

Damage Repair

Trans Reg

RN

A Transcription

Processing

Modification

Aminoacylation Degradation l,

Pro

tein

Translation

Processing I Translocation Processing II

Folding

Modification Complexation

Ribosome

Term Org # Activation

Degradation l, #

Oth

er Metabolism

Shape FtsZ

Cytokinesis

Many resources are shared

Many resources are shared

1 s

Uptake

Metabolism

Transcription

Translation

Replication

Cel

l st

ates

Cel

l st

ates

Uptake

Metabolism

Transcription

Translation

Replication

Cel

l st

ates

Uptake

Metabolism

Transcription

Translation

Replication

Div

ide

stat

e

Div

ide

stat

e

Div

ide

stat

e

Simulation algorithm

Mycoplasma model contains 28 sub-models

Karr et al., 2012

Course expertise

Modeling • Frank Bergmann •Marcus Krantz •Wolfgang Liebermeister •Pedro Mendes •Chris Myers •Pnar Pir •Kieran Smallbone

Curation •Vijayalakshmi Chelliah

Standards •Michael Hucka • Falk Schreiber •Dagmar Waltemath

Karr et al., 2012

Example sub-model: Transcription

Example sub-model: Transcription

Karr et al., 2012

Free

Bound

Promoter Bound

Active

1. Update RNA polymerase states

3. Bind RNA polymerase

2. Calculate promoter affinities

4. Elongate and terminate transcripts

AUGAUCCGUCUCUAAUGUCUAC

UTCAACGUGAGGUAAUAAAGUC

UCCACGAUGCUACUGUAUC

GCCUCAUACUGCGGAU

UUACGUAUCAGUGAUCAGUACU

Sequence

Tra

nsc

ript

HcrA Spx Fur GntR LuxR

glpF dnaJ dnaK gntR trxB polC

Example sub-model: Transcription

•Compare the model’s predictions to data, 𝑦𝑖

•Define an error metric

∑ 𝐸 𝑓𝑖(𝑥; 𝑝) cells,time − 𝑦𝑖2

•Numerically minimize error •Gradient descent

• Scatter search

• Simulated annealing

•Genetic algorithms

5. Identify parameters

•Large parameter space

•Stochastic model

•Large computational cost

•Heterogeneous data

•Little dynamic, single cell data

5. Identify parameters

Model reduction enables parameter identification

3. Manually tune parameters using full model

1. Reduce model

Time

Model Experiment

Mole

cule

Mole

cule

2. Identify reduced model parameters using traditional methods

Software: wholecell.org

•ODE models • COPASI: copasi.org

• V-Cell: nrcam.uchc.edu

• Systems biology toolbox

•Boolean models • CellNOpt

• Flux-balance analysis • openCOBRA: opencobra.sourceforge.net

• RAVEN

• Integrative models • E-Cell: e-cell.org

• Whole-cell: wholecell.org

• Standards • SBML: sbml.org

• CellML: cellml.org

Software

Cellular composition

Metabolite concentrations

mRNA, protein copy numbers

RNA synthesis rates

Karr et al., 2012

DNA binding protein collisions

Karr et al., 2012

DNA binding

Replication

Translation

60 m mol ATP / gDCW 80 a mol ATP / cell

Energy consumption

v

v

Karr et al., 2012

Energy consumption

Model Validate

Engineer

Whole-cell modeling

Validate

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Model reproduces observed metabolomics

Karr et al., 2012

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Model validated by experiments and theory Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Model validated by experiments and theory Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Colorimetric growth assay Model predictions

Model reproduces measured growth rate

Karr et al., 2012

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Model validated by experiments and theory Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Matches training data Cell mass, volume

Biomass composition

RNA, protein expression, half-lives

Superhelicity

Matches published data Metabolite concentrations

DNA-bound protein density

Gene essentiality

Matches new data Wild-type growth rate

Disruption strain growth rates

Matches theory Mass conservation

Central dogma

Cell theory

Evolution

No obvious errors Plot model predictions

Manually inspect data

Compare to known biology

Software stable Simulation code is stable

Tests passing

Validate model against experiments and theory

Model

Engineer

Whole-cell modeling

Validate

Engineer

What genomic modifications maximize growth?

Time M

ass

Example: growth optimization

M. genitalium

M. mycoides

M. pneumoniae

Optimal gene expression

Optimal architecture retains robustness Optimal gene expression retains robustness

Graphical design tool Clotho, TinkerCell, GenoCAD

High-level language BioCompiler

Biophysical model Whole-cell models, SCHEMA, MD

Physical implementation Gibson assembly, TALENs, ZFNs, CRISPR

Transplantation Transplantation

(if (nutrients) (grow) (sporulate))

Directed evolution

Mutate Select

Synthetic design landscape

Karr lab: expanding whole-cell models

M. pneumoniae • Expand scope: regulation

• Improve accuracy: species-specific data

• Enable rational genome engineering

•Cell-based drug therapy

Human cancer •Colorectal cancer •Personalized models •Precision medicine

Karr lab: solving important problems

Biological discovery

Synthetic networks

Biological design

Drug repositioning

Drug toxicity

Karr lab: developing modeling tools

Reconstruction: WholeCellKB

Parallelized simulator

Parameter estimation

Simulation storage: WholeCellSimDB

Visualization: WholeCellViz

wholecell.org

??

•How can we model more complex physiology? • Transcriptional regulation • Translational regulation • Stochastic death, failure modes • Higher-order meta-stable states • Resource distribution • Aging • Evolution • Populations

•How can we model more complex organisms? • Larger bacteria • Eukaryotes •Multicellularity • Humans

•How can we use models to direct engineering?

Open challenges

Whole-cell modeling course

1. Teach whole-cell modeling •Model biological systems •Construct dynamical models • Integrate models

2. Improve implementation •Reusable • Standard •Open

3. Improve methodology

Data

?

Whole-cell models

Broadly predicts cell physiology

Integrates heterogeneous data and models

Guides bioengineering and medicine

Knowledge

• Karr JR et al. (2012) A Whole-Cell Computational Model Predicts Phenotype from Genotype. Cell, 150, 389-401.

• Macklin DN, Ruggero NA, Covert MW (2014) The future of whole-cell modeling. Curr Opin Biotechnol, 28C, 111-115.

• Shuler ML, Foley P, Atlas J (2012). Modeling a minimal cell. Methods Mol Biol, 881, 573-610.

• Joyce AR, Palsson BØ (2007). Toward whole cell modeling and simulation: comprehensive functional genomics through the constraint-based approach. Prog Drug Res 64, 267-309.

• Tomita M (2001). Whole-cell simulation: a grand challenge of the 21st century. Trends Biotechnol 6, 205-10.

• Surovtsev IV et al. (2009) Mathematical modeling of a minimal protocell with coordinated growth and division. J Theor Biol, 260, 422-9.

Recommended reading

• Thiele I et al. (2009). Genome-scale reconstruction of Escherichia coli's transcriptional and translational machinery: a knowledge base, its mathematical formulation, and its functional characterization. PLoS Comput Biol. 5, e1000312.

• Orth JD, Thiele I, Palsson BØ (2010). What is flux balance analysis? Nat Biotechnol, 28, 245-8.

• Covert MW et al (2008). Integrated Flux Balance Analysis Model of Escherichia coli. Bioinformatics 24, 2044–50.

• Covert MW et al (2004). Integrating high-throughput and computational data elucidates bacterial networks. Nature, 429, 92-6.

Recommended reading: FBA