Global reconstruction of the human metabolic network based on genomic and bibliomic data

Abstract

Metabolism is a vital cellular process, and its malfunction is a major contributor to human disease. Metabolic networks are complex and highly interconnected, and thus systems-level computational approaches are required to elucidate and understand metabolic genotype-phenotype relationships. We have manually reconstructed the global human metabolic network based on Build 35 of the genome annotation and a comprehensive evaluation of >50 years of legacy data (i.e., bibliomic data). Herein we describe the reconstruction process and demonstrate how the resulting genome-scale (or global) network can be used (i) for the discovery of missing information, (ii) for the formulation of an in silico model, and (iii) as a structured context for analyzing high-throughput biological data sets. Our comprehensive evaluation of the literature revealed many gaps in the current understanding of human metabolism that require future experimental investigation. Mathematical analysis of network structure elucidated the implications of intracellular compartmentalization and the potential use of correlated reaction sets for alternative drug target identification. Integrated analysis of high-throughput data sets within the context of the reconstruction enabled a global assessment of functional metabolic states. These results highlight some of the applications enabled by the reconstructed human metabolic network. The establishment of this network represents an important step toward genome-scale human systems biology.

Fig. 1.

Human metabolic knowledge landscape. Colors represent the percentage of reactions within a pathway that have a confidence score of 3 (biochemical or genetic evidence), 2 (physiological data or evidence from a nonhuman mammalian cell), 1 (modeling evidence), or 0 (unevaluated). Metabolic pathways (primarily defined by the Kyoto Encyclopedia of Genes and Genomes LIGAND database) were classified into three categories based on their level of characterization as detailed in the text.

Fig. 2.

Normalized cumulative singular value spectra for H. sapiens, S. cerevisiae, and E. coli and dominant metabolite modes. (A) Compartmentalized networks have a greater effective dimensionality than their noncompartmentalized counterparts, requiring a larger number of singular values to completely reconstruct the network. Each spectrum shows the number of decomposed modes (x axis) required to reconstruct a given fraction (y axis) of the S matrix's content. (B) The first five modes of the human metabolite coupling matrix (38) highlight the importance of the production and exchange of energy equivalents and the potentially significant impact of osmotic regulation. c, cytoplasmic; e, extracellular; g, Golgi apparatus; m, mitochondrial.

Fig. 3.

Coupled reaction sets involving cholesterol biosynthesis and glutathione production and transport. (A) The cholesterol biosynthesis coupled set includes all reactions except those shaded in gray. Note that the groups of reactions are not all directly connected and could not be identified by visual inspection alone. (B) Reactions in the glutathione reaction set were mapped to disease associations by using Mendelian Inheritance in Man identification tags. Deficiencies in glutathione synthetase (GTHS) or glutamate-cysteine ligase (GLUCYS) both result in hemolytic anemia, supporting the notion that enzyme deficiencies in the same coupled set may have similar phenotypes. Interference with or decreases in GLUCYS activity is associated with an increased risk of myocardial infarctions (MI). A high-resolution version of this figure is available in SI Figs. 19 and 20.

Fig. 4.

Integrated analysis of gene expression data from gastric bypass patients before surgery and 1 year afterward. Expression measurements were to reactions in the global human metabolic network and then visualized on Recon 1's comprehensive collection of human metabolic maps. Reactions are color-coded based on their corresponding gene expression changes (green, down-regulated; red, up-regulated; white, no data available or reaction level conflict). Arrows next to reaction abbreviations indicate the magnitude of expression changes on a log10 scale (gray boxes indicate no data available). A high-resolution version of this figure is available in SI Fig. 21.