Skip to content

Multi-scale Erythrocyte Metabolism Platform

Summary

Mih, Brunk, Bordbar & Palsson (2016) present a multi-scale computational platform that integrates protein structural data, molecular dynamics/docking simulations, and constraint-based/kinetic genome-scale metabolic modeling to mechanistically predict how sequence variants affect drug responses in human erythrocyte metabolism. Building a "GEM-PRO" (genome-scale model with protein structures) from the existing erythrocyte metabolic model iAB-RBC-283, the authors demonstrate the workflow end-to-end on three case-study proteins — COMT, G6PD, and GAPDH — connecting a single amino-acid substitution through structural, thermodynamic, and systems-level consequences.

Method: A Four-Layer Workflow

  1. Sequence-to-structure mapping (GEM-PRO): starting from the genome-scale erythrocyte model iAB-RBC-283 (281 genes, 346 proteins), each protein-encoding gene is mapped to its 3D structure via the Protein Data Bank (PDB), or homology-modeled via I-TASSER when no experimental structure exists. The resulting model, iNM-RBC-283-GP, achieved structural coverage for 181 of 346 proteins from 1,766 unique PDB entries, plus 312 homology models.
  2. Variant classification: 6,800 exon-coding SNPs in erythrocyte-expressed genes were mapped to 247 of 281 model genes (88%); >90% were missense. Proteins were stratified into four classes by data availability — Class I (structural, pharmacogenomic, and clinical data all available; includes COMT, ALDH3A1, ADA, G6PD, GPX1, UMPS), through Class IV (structurally modeled in the network but with no external validation data), to identify where the framework adds the most value.
  3. Structural and thermodynamic simulation: molecular dynamics and ensemble docking simulations compare wild-type versus variant protein structures, quantifying changes in substrate/product/cofactor binding free energy (ΔΔG).
  4. Systems-level integration: the structurally-derived kinetic parameter changes (Km, kcat) are fed into both constraint-based modeling (iAB-RBC-283) and a kinetic rate-law model (Mass Action Stoichiometric Simulation, MASS) to predict the downstream effect on network-wide flux and metabolite concentrations.

Case Studies

  • G6PD "Andalus" variant (Arg454His): structural simulation correctly reproduced the direction of the experimentally known effect — increased binding affinity for the substrate glucose-6-phosphate (predicted ΔΔG=3.00±0.68 kcal/mol; corroborated by higher-accuracy thermodynamic integration, ΔΔG=3.59 kcal/mol) alongside a drastically decreased cofactor (NADP⁺) binding affinity. Integrating the experimentally measured kcat for this variant into the kinetic model reproduced the WHO clinical classification of this variant as causing "severe deficiency with intermittent hemolysis" — the NADPH:NADP⁺ ratio and the cell's tolerated oxidative load both dropped sharply, and constraint-based modeling of the same flux decrease produced significant downstream changes in the glutathione reductase pathway.
  • COMT variant: constraint-based modeling predicted decreased binding affinity for norepinephrine/dopamine substrates as well as for the inhibitor drugs tolcapone and entacapone, leading to decreased flux and reduced export of methylated catecholamine metabolites.
  • GAPDH variant: kinetic modeling predicted the cell could not tolerate the variant's increased Km, resulting in an infeasible model state corresponding to cell lysis — illustrating that the framework can identify variants with severe, potentially lethal systems-level consequences even for a protein without prior clinical annotation (a Class II protein in this study's data-availability scheme).

Significance

This work established the GEM-PRO integration pattern — later applied at larger scale by Recon3D across the full human metabolic network — and demonstrated that structural simulation of a single sequence variant can be mechanistically threaded through to a systems-level, disease-relevant phenotype prediction (e.g., correctly recovering a WHO clinical severity classification from first-principles molecular simulation).

See Also

  • Recon3D — a later, genome-wide extension of the same structural-integration (GEM-PRO) philosophy to the complete human metabolic network.
  • Metabolic Flux Modulation of Genetic Risk — a population-scale, genotype-driven (rather than single-variant, structure-driven) approach to linking genetic variation to organ-specific metabolic flux.

Citations

[1] Mih, N., Brunk, E., Bordbar, A., & Palsson, B.O. (2016). A Multi-scale Computational Platform to Mechanistically Assess the Effect of Genetic Variation on Drug Responses in Human Erythrocyte Metabolism. PLOS Computational Biology, 12(7), e1005039.