From annotated genomes to metabolic flux models

From annotated genomes to metabolic flux models

Jeremy ZuckerBroad Institute of MIT & Harvard

August 25, 2009

Outline

• Metabolic flux models– Tuberculosis

• Annotating genomes– Rhodococcus opacus– Neurospora crassa

E-flux• Goal: To Predict the effect of drugs on

growth using expression data and flux models

• Resources: – Boshoff compendium– Mycolic acid pathway

• Solution: use differential gene expression to differentially constrain flux limits

E-flux results

• Our method successfully identifies 7 of the 8 known mycolic acid inhibitors in a compendium of 235 conditions,

• identifies the top anti-TB drugs in this dataset .

Future Tuberculosis Goals

To model hypoxia-induced persistence using: Proteomics, Metabolomics, Transcriptomics Fluxomics Glycomics Lipidomics

TB Resources

• 3 FBA models, • Chemostat experiments• 27 genomes sequenced in TBDB• On-site TBDB curator. • Systems Biology of TB omics data

Solution: One Database to rule them all

MtbrvCyc13.0

GSMN-TB

MtbrvCyc 11.0

iNJ661

MAP

Omics Viewer

Pathway models

rFBA models

Comparative analysis of Mtb metabolic models

GSMN-TB

iNJ661 MAP

Citations 141 99 23Metabolites

739 740 197

Reactions 849 939 219Genes 726 661 28Enzymes 587 543 18

Genes

235

GSMN-TB

iNJ661 MAP

166 2

19472 3

4

Compounds

440

GSMN-TB

iNJ661 MAP

440 178

18281 0

1

Citations

118

GSMN-TB

iNJ661 MAP

78 21

021 2

0

Reactions

555

GSMN-TB

iNJ661 MAP

646 209

7285 2

1

Reconstructing Metabolic models with Pathway-tools

• EC predictions from sequence• PGDB from Flux model• Automatically refining flux models based

on phenotype data• Applying expression data to Flux

models for Omics analysis

EFICAz

• Goal: Predict EC numbers for protein sequences with known confidence.

• Resources: ENZYME, PFAM, PROSITE

• Solution: homofunctional and heterofunctional MSA, FDR, SVM, SIT-based precision.

sbml2biocyc

• Goal: Generate PGDB from FBA model • Resources: SBML model • Solution:

– sbml2biocyc code to transform SBML data to generate

• reactions, • metabolites, • gene associations, • citations for PGDB.

Biohacker

• Goal: search the space of metabolic models to find the ones that are most consistent with the phenotype data

• Resources: – KO data. – Initial metabolic model. – EC confidence predictions

• Solution: MILP algorithm.

Omics viewer

• Goal: Googlemaps-like interface for cellular overview that enables pasting flux, RNA expression, etc

• Resources: – Pathway-tools source code– OpenLayers, – Flash,– Googlemaps API

Rhodococcus opacus:Goals

• To model lipid storage mechanism for biofuels.

R. opacus: Resources

• Sinsky lab• Biolog data• Expression data• Genome sequence• EC Predictor

R. Opacus solution

• Use EFICaz to make EC predictions• Use reachability analysis to guide

outside-in model reconstruction• Use pathway curation to guide inside-

out model reconstruction• Can we do better?

Neurospora crassa:Goals

• Predict phenotype KO experiments

N. crassa: Resources

• Systems biology of Neurospora grant• Extensive literature• very dedicated community • Genome sequence• Ptools pipeline

N. crassa: Solution

• Inside-out method with Heather Hood• Outside-in method with MILP algorithm