Metabolic Reconstruction and Modeling From Sequenced Genomes

Abstract

We present new developments in the bioinformatics algorithms that reconstruct the metabolic network of an organism from its sequenced genome, and that convert that metabolic network into a quantitative metabolic model. These algorithms are implemented in the Pathway Tools software. The Pathway Tools metabolic reconstruction process begins with an annotated genome; its first step infers the subset of metabolic reactions in the MetaCyc database that are catalyzed by enzymes within the genome. Reactome inference is based on the enzyme name and/or EC numbers within the genome annotation. The second step uses the inferred reactome to infer the subset of MetaCyc metabolic pathways present in the organism using a pathway scoring algorithm. Recent improvements to these algorithms are based on a large increase in the number of EC numbers in recent years, growth in the reaction (+30%) and pathway (+22%) content of the MetaCyc database, and development of a new pathway scoring algorithm that makes use of key reactions that must be either present or absent, and pathway taxonomic range information. The resulting metabolic reconstruction is available in the form of a Pathway/Genome Database (PGDB) that can be queried and visualized via an extensive set of online tools. The BioCyc.org website contains such PGDBs for 14,700 organisms.The MetaFlux component of Pathway Tools enables creation and execution of quantitative steady‐state metabolic flux models. Several steps are used to refine a PGDB to a quantitative steady‐state metabolic flux model. First, we must identify a set of nutrients that support growth of the organism, and we must identify a set of biomass metabolites that form the molecular building blocks of the organism. Next we ensure that the reactome within the PGDB can synthesize each of the biomass metabolites from the provided nutrients. Often the reactome contains gaps due to incompleteness in the genome annotation. We present a new taxonomic gap‐filling algorithm with significantly improved accuracy: when evaluated on randomly degraded variants of the Escherichia coli metabolic model, it shows an average F1 score of 99.0, compared to 91.0 for the previous Pathway Tools gap filler, and compared to 80.3 for a basic gap filler. Once gaps have been filled in the reaction network, MetaFlux can execute the model, using flux‐balance analysis to compute flux rates for all reactions in the model. Those fluxes can be interpreted by visualizing them on a whole‐organism metabolic network diagram. MetaFlux can also simulate the effects of single and double‐gene knock‐out mutants on the model. And MetaFlux can execute groups of models that simulate the growth of organism communities.Support or Funding InformationThis work was supported by the National Institutes of Health under grants GM080746 and GM075742. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.