Transcript Constraint-based modelling of bacterial metabolic networks

Constraint-based
modelling of
bacterial metabolic
networks
– where are we in 2011?
What I aim to do in 30 minutes...
• Give you a brief intro into our
system of study.
• Recap the things we talked about in
York in York in 2009.
• Think about what we could do in
Edinburgh in 2011?
A bit about our system (see Sandy’s talk on Friday)
The pea aphid,
Acyrthosiphon
pisum.
• About 5000 different species
• Major crop pests.
• Restricted diet of phloem sap.
• All contain an obligate primary symbiont.
The γ-proteobacterium Buchnera aphidicola sp. APS is
the primary symbiont of the pea aphid
• Located in specialised insect cells called
bacteriocytes in their body cavity.
• They are surrounded by an aphid-derived
bacteriocycte membrane.
• TEM of bacteriocyte cytoplasm, showing coccoid
Buchnera.
• The Buchnera are unculturable so not tractable to
traditional microbiological methods.
• Vertically transmitted to aphid offspring via the
ovary.
• The function of the symbiosis is nutritional.
• Phloem sap poor in essential amino acids [EAAs] (His, Iso, Leu, Lys, Met, Phe,
Thr, Trp and Val).
• There is experimental evidence that EAAs are provided by the symbiont.
The Buchnera APS genome
• Small - 0.64 Mb
• 607 genes (569 protein coding genes) that are a subset of the
E. coli K-12 genome.
• Almost 90% of the genes have known functions in E. coli K-12.
• Specific retention of pathways for biosynthesis of EAAs.
• Virtually no transcription regulation.
Carbon-skeleton based manual visualisation of iGT196
196 gene products
240 compounds (39% of iJR904)
263 reactions (27% of iJR904)
35% of reactions for EAA
biosynthesis.
Nework visualisation
• Initially used Cytoscape (picture only really).
• Now use CellDesigner to draw model that can export SBML for
Cobra (or Surrey FBA).
Key
Red hexagon – high flux precursor
Red circle – low flux precursor
Grey triangle – inferred reaction
Thomas et al., (2009) BMC Systems Biology 3:24.
Blue square - EAA
Blue circle – biomass component
Building a whole genome model (of a bacterium)
Taken from Durot, Bourguignon and Schachter (2009) FEMS Microbiology Reviews 33:164-190.
Modelling construction
Orthology
mapping to
known model
KASS
PRIAMS
Assigning transporters
- making specific GPRs is difficult
- need more experimental data.
EFICAz
Assignment of E.C.
Numbers?
Problems with
using E.C. Numbers
-Better ontology?
- more coverage
Input for a Cyc-type
reconstruction
Manual curation in
CellDesigner
Value of automated methods?
Network visualisation
- best tools?
- overlay fluxes?
Model exchange
- SBML – strict enough?
- BioPAX
- MIRIAM
Getting the model to ‘fire’
Tools for simple linear programming
- COBRA – version 2 (Feb 2011 – look at this on Friday? )
- SurreyFBA
- Scrumpy
Sanity checking
- Sympheny (if rich)
- critical so that don’t get nonsense out
- check for production of all biomass
Biomass reaction
components
- Base it on E. coli or figure it out yourself
- check major fluxes are in the ‘right’
- Different biomass reactions for different
direction
growth conditions
- check network ‘quality’ – FBA aims to
- Cofactor constraints
minimise total number of fluxes
- how valid is it to reverse a reaction?
Maintenance energy
- growth and non growth related
- ATP yield in respiration
- Redox balancing
What to do when it doesn’t work?
- Iterative step-wise model building
Objective function
- Need tools to ‘debug the bug’
- Is biomass production always suitable?
- Dual objectives?
Getting more from FBA
Understanding the output
- solution space of the optimisation
- FVA
- how to reduce this further?
Integration of ‘omics data
Transcriptomic data
What does it mean for enzyme fluxes?
Proteomics data
How to use it to constrain the model?
- regulatory FBA – Boolean filter
- mixed integer linear programming (MILP).
Metabolomic data
Constraining internal fluxes
- Flux splits (NDH1 versus NDH2)
- Dual objective functions
- Thermodynamics
- Allosteric regulation
Dynamic FBA
- time dependence
- consider kinetics and concentrations
- integrated FBA (iFBA)
Applications of a working model
Feist definitions...
Metabolic
engineering
Using iAF1260
Lycopene
L-valine
L-threonine
MOMA
OptKnock
Model-directed
discovery
Informing on
the biological
function of
metabolism.
Orphan
enzymes and
transporters.
Interpretation
of phenotypic
screens
Analysis of
network
properties
Compare KO
strains and/or
Biolog data to
model
predictions
Network
analysis – how
much value?
Evolution of
reduced
networks
What are the
inputs and
output?
Pan genomes.
- Improves
model.
Buchnera has
some highvalue waste
products.
Missing
biology?
Studies of
evolutionary
processes
What kind of systems
have we been
building or
analysing?
M. tuberculosis
B. aphidicola
Thomas
Zucker
Wood
Macdonald
Price
McFadden
Ebenhoeh
Pérez-Castillo
Kierzek
E. coli K-12
iAF1260
Brietling
Cornish Bowden
De Jong
Streptomyces sp.
Fell
Poolman
Westerhoff
Velasco
A. thaliana
S. cerevisiae
What next for an E. coli model?
48 % of all CDS included in iAF1260. Not much more to add!
....but still some reactions not assigned to genes
Pan-genome models probably more useful – define the core metabolic network for
the species – removes K-12 specific components.
What I’d like to get out of this meeting
• How can I usefully use my transcriptomic and proteomic data
to add to the FBA (not just with Boolean on and off)?
• I want to decouple growth from amino acid overproduction.
How can I use dual objective functions?
• How can model my transporters more effectively? Both at the
level of functional annotation and kinetics.
• Where are we with kinetic models? Can we usefully integrate
them into our FBA modelling?
• What’s new in terms of methods and software that I can use
to improve my analysis.
Gavin Thomas
Buchnera/aphids
FBA
Andrej Kierzek
Johnjoe McFadden
Streptomyces
Mycobacterium
FBA and kinetic
Isaac Perez Castillo
Kings College London
E. coli
Metabolic optimisation principle
Sergio Bordel Velasco Chalmers, Sweden (Nielson lab) Metabolic models in industrial
microbiology . Overlaying transcription. Random sampling of flux distributions .
Hans Westerhoff
Athel Cornish Bowden
Mark Poolman
David Fell
Thomas Forth
Hidde de Jong
Oliver Ebenhoeh
Nathan Price
metabolism
Grenoble, France
Aberdeen
Illinois, USA
E. coli
metabolic regulatory networks
E. coli
model building tools
E. coli and TB Probabilistic regulation of
Rainer Breitling
Groningen
engineering (with Erico Takano)
Streptomyces and parasites
Jeremy Zucker
BioPAX Buchnera TB, E-Flux.
Broad, USA
Metabolic