Human-GEM User Guide

Important

This guide applies to Human-GEM version v2.0.0 ("Human2"). If you are using a different version of Human-GEM, we cannot guarantee that it will function as described in this guide.

Overview

This guide contains instructions and examples of how to use Human-GEM, a human genome-scale metabolic model (GEM). Choose a section from the sidebar or the list below to get started!

What's new in Human2

Human2 (v2.0.0) is a major update to Human-GEM, described in Luo et al. (2026) PNAS. Highlights include:

LLM-assisted curation. Gene–reaction associations were systematically reviewed with large language models, improving gene–protein–reaction (GPR) rules and gene essentiality predictions. A public metabolic-gene analysis GPT is available.
Continuous quality control. Every proposed change is checked by automated GitHub Actions (metabolic tasks, MEMOTE, MACAW, duplicate and dead-end detection) before it can be merged.
Improved annotation. Expanded metabolite and reaction cross-references (including Rhea and MA identifiers) and structure-verified metabolite annotations at pH 7.3 (see Additional tools).
Enzyme-constrained and whole-body models. An enzyme-constrained ec-Human2 and whole-body models for different sexes and ages enable dynamic, multi-organ simulations.
Python support. In addition to MATLAB, the workflows in this guide can now also be run from Python (COBRApy, raven-toolbox, and geckopy).

Citation

If you use Human2 in your research, please cite:

Luo J, Wang H, Moyer D, Guo Z, Robinson JL, Gustafsson J, Anton M, Chen Y, Kerkhoven EJ, Nielsen J, Li F. Reconstruction of human metabolic models with large language models. PNAS 123.15:e2516511123 (2026). doi:10.1073/pnas.2516511123

If you use Human1 in your research, please cite:

Robinson JL, et al. An atlas of human metabolism. Sci. Signal. 13, eaaz1482 (2020). doi:10.1126/scisignal.aaz1482