Genomes to Life Contractor-Grantee Workshop III
February 6-9, 2005, Washington, D.C.
Genomics:GTL Program Projects
Harvard Medical School
1
Metabolic Network Modeling of Prochlorococcus marinus
George M. Church* (g1m1c1@arep.med.harvard.edu), Xiaoxia Lin, Daniel Segrè, Aaron Brandes, and Jeremy Zucker
Harvard Medical School, Boston, MA
The marine cyanobacterium Prochlorococcus marinus dominates the phytoplankton in the tropical and subtropical oceans and contributes to a significant fraction of the global photosynthesis. Several strains of Prochlorococcus have been sequenced, which provides us a promising starting point for investigating the relationship between genotype and phenotype at a genome scale and with a comparative approach. To achieve the ultimate goal of understanding the metabolism at a systems level, we are developing and utilizing new metabolic network models in several directions.
Comparison and connection of day-night metabolisms
Day-night cycles are known to play a central role in the metabolism of Prochlorococcus. We are exploring two approaches to model the difference and connection between day and night. One is to take the full metabolic network and formulate two separate models assuming different nutrient conditions and optimality criteria. Then the flux predictions can be compared to mRNA and protein expression data. In the other approach, we make use of the protein expression data, which helps to reduce the feasible flux space and leads to finer flux predictions.
Construction of metabolic networks
One major challenge in constructing complete and accurate in silico metabolic networks for quantitative analysis such as flux balance analysis (FBA) is to identify reactions that are “missed” in the annotation. We have been mainly using Pathway Tools software suite developed by SRI to identify metabolic reactions and are developing new algorithms to construct the “functional” metabolic network from a network perspective. Biochemical reactions with identified enzymes are included and then an “optimal” set of reactions are added such that the network produces the specified growth phenotype given corresponding nutrient conditions. Identification of the missing links will also help to refine the genome annotation. Another problem is that there exist “orphaned enzymes” — experimentally elucidated biochemical reactions whose enzyme has never been sequenced. To address this problem, we are utilizing a pathway hole-filling algorithm developed by SRI and developing bioinformatics techniques to identify candidate genes for these orphaned enzymes.
Analysis of metabolic networks with mass balance and energy balance
Conventional flux balance analysis (FBA) only considers mass balance. We are incorporating constraints representing the second law of thermodynamics, which eliminates thermodynamically infeasible fluxes. A subset of the additional constraints exhibits non-convexity, giving rise to substantial difficulty in the solution of the resulting optimization problem. We are developing new methods to overcome this challenge to make full use of combined FBA and EBA (energy balance analysis).
Construction and comparative study of whole-cell metabolic networks of MED4 and other strains
By combining a bioinformatics pipeline for generating metabolic network models from genome annotations and manual inspection/modification, we have constructed the in silico metabolic network of central carbon metabolism and amino acid biosynthesis for Prochlorococcus MED4, a high-light-adapted strain. We are extending it towards the genome-wide network. In addition, we will construct metabolic network models for the other sequenced strains, including the low-light-adapted MIT9313. Comparison of the structures of their metabolic networks and the calculated flux distributions under varying conditions will enable us to understand at a systems level how these different strains adapt their metabolisms to the different environments.
Project Web site: http://arep.med.harvard.edu/DOEGTL/
* Presenting author
