Tier 0 · open
Data & licensing
The corpus is a versioned artifact, not a moving target. Each release carries a content hash computed over every live model, so a published result can be traced to the exact corpus that produced it.
Current release
- Versionv0.2.2
- Content hash58a8f20e2aba...
- Released2026-10-05
- Models15,664
- Redistributable9,911
- Zenodo DOI10.5281/zenodo.22019863
Licensing is stated, not assumed
Models come from many sources under different licenses. The table below is the license as understood from the source repository or publication at the time of ingestion; it is stored per-model in the corpus DB so the right answer is always derivable, not guessed. Nothing that cannot be lawfully redistributed is ever re-served. Reports will tell you which sources to obtain directly.
Redistribution is only half of license compliance. A model can be redistributable yet still restricted for commercial use. AGORA, AGORA2, APOLLO-595 and BiGG (5,753 models, ~37% of the corpus, including the hydrogenotrophic-methanogen finishers behind the methane example) are treated as NonCommercial: free to use and cite for research, but not for commercial use without a license from the source. NonCommercial restricts the commercial USE of derived results, not only their redistribution, so a paid or commercial discovery run uses a subset of 9,911 models released under CC-BY 4.0, excluding the NonCommercial sources' own model files (AGORA, AGORA2, APOLLO-595, BiGG). This subset is not commercially clean: roughly 9,900 of the 9,911 are CarveMe / embl_gems reconstructions carved from a BiGG-derived universal model, so BiGG's academic-use restriction still reaches them through that derivation unless the (contestable) non-copyrightable-facts position holds. Commercial users must complete their own IP review (and may need a UCSD BiGG license) before relying on it. Syntropa discloses license status per-model so this review is possible.
Openly redistributable (9,911 models)
These may be included in derivative datasets with attribution. Cite the original source publication, not just the Syntropa corpus.
Provenance note (stated, not assumed). Nearly all of these are CarveMe reconstructions (and embl_gems, which are also CarveMe outputs) carved from a universal model derived from the BiGG database. Syntropa releases its reconstructions as CC-BY 4.0, following the CarveMe author's own CC-BY 4.0 release of embl_gems and on the basis that reaction stoichiometries are non-copyrightable facts, not the copyrightable content of BiGG. That position is defensible but not a substitute for your own review: commercial users who need certainty about the BiGG-derived content should review BiGG's academic-use terms (bigg.ucsd.edu/license).
| Source | Models | License |
|---|---|---|
| embl_gems (CarveMe RefSeq, Machado 2018) | 5,573 | CC-BY 4.0 |
| CarveMe 1.6.6 reconstruction from NCBI RefSeq proteome | 4,326 | CC-BY 4.0 (Syntropa reconstruction; tool Apache-2.0) |
| SysBioChalmers | 10 | CC-BY 4.0 |
| biomass composition | 1 | CC-BY 4.0 (Syntropa original) |
| embl_gems (CarveMe RefSeq) | 1 | CC-BY 4.0 |
Link-only (5,753 models)
These are held locally for discovery runs but are not included in the redistributable corpus dump. Obtain them from the original source; the discovery reports always name which source each hit came from.
| Source | Models | License status |
|---|---|---|
| AGORA2 v2.01 (Heinken 2023) | 4,672 | VMH distributes AGORA2 under non-commercial terms; a specific CC license is not separately posted on the model DATA (the MIT repo license covers only the toolchain code), so it is treated conservatively as NonCommercial: restricting commercial USE, not only redistribution. Cite Heinken et al. 2023. |
| AGORA v1.03 (Magnusdottir 2017) | 601 | CC-BY-NC 4.0 (NonCommercial) per VMH terms: restricts commercial USE; the data license is not separately posted, so it is treated conservatively as NonCommercial |
| APOLLO-595 | 390 | Treated as NonCommercial: the APOLLO publication is CC-BY-NC-ND and VMH distributes the reconstructions under NonCommercial terms. |
| BiGG | 90 | BiGG (UCSD): cite the model's own source publication; academic / non-commercial use: commercial use may require a separate UCSD license (bigg.ucsd.edu). Treated conservatively as non-commercial. |
How to cite
Cite the corpus by its Zenodo DOI, or by version and content hash. In a methods section:
Syntropa microbial metabolic corpus v0.2.2 Zenodo. doi:10.5281/zenodo.22019863 Content hash (SHA-256, first 12 chars): 58a8f20e2aba... Released: 2026-10-05 https://syntropa.org/data
Each report embeds a provenance stamp (corpus version, content hash, tool version + code hash, and a full environment lock), so a result is traceable and content-auditable independently of the full corpus release. The tool tree these reports were rendered from is committed, so the exact code that produced them is recorded rather than floating. A tagged, published tool release is still pending, so independent replication today rests on the content hash.
Model source links
- EMBL-GEMs: github.com/cdanielmachado/embl_gems
- AGORA2: vmh.life (Heinken et al. 2023)
- AGORA v1.03: vmh.life (Magnúsdóttir et al. 2017)
- CarveMe reconstructions: github.com/cdanielmachado/carveme
- APOLLO-595: doi.org/10.7910/DVN/PIZCBI
- BiGG Models: bigg.ucsd.edu
- SysBioChalmers: github.com/SysBioChalmers