Open research data and tools for microbial electrochemical systems (MES) — microbial fuel cells, electrolysis cells, desalination cells, and the wider bioelectrochemical family.
Cross-study comparison in microbial electrochemistry is hard: measurements are scattered across thousands of papers, each in its own units and conventions. These repositories publish the standardized, queryable layer underneath messai.io — every extracted value carries its paper of origin, an SI-normalized value, and a link to a shared parameter ontology.
| Repository | What's in it |
|---|---|
| MESS-Parameters | The parameter ontology and the extraction dataset — 704 parameter definitions across 13 categories, 23,568 papers indexed, 196,522 extracted values with provenance. Start here. |
| MESS-Materials | DFT-computed properties for 44 MES materials — electrodes, membranes, current collectors, catalysts. Keyed by the same material slugs as MESS-Parameters. |
| MESS-Microbes | Curated database of 28 MES-relevant organisms — electrogens, electrotrophs, DIET partners, methanogens, and intentional non-MES controls, with mechanism annotations. |
| MESS-datasets | Catalog of 41 curated third-party and literature-extracted datasets relevant to MES research. |
Earlier-stage, source-available packages: MESS-Agents · MESS-Hypotheses · MESS-Learning · MESS-Methods · MESS-Simulations
Data packages (Parameters, Materials, datasets, Learning) are CC BY 4.0.
Code packages (Microbes, Agents, Hypotheses, Methods, Simulations) are MIT.
Each repository carries its own LICENSE and, where applicable, a
CITATION.cff for academic citation.
These repositories are read-only mirrors. The source of truth is a private monorepo, and mirrors are force-updated on each release — so pull requests opened here get closed automatically. That is a limitation of the setup, not a judgement on the contribution.
What genuinely helps, and what we act on:
- Open an issue. Data corrections, wrong units, a misattributed paper, a missing parameter, a bug in a schema — these are the most valuable thing you can send, and they get applied upstream and propagated on the next release.
- Start a discussion about scope, methodology, or a parameter definition you think is wrong.
- Cite and reuse. The data is openly licensed; use it.
If you want to contribute code and the mirror model is in your way, open an issue and say so — we would rather hear it than lose the contribution.
The extraction corpus is machine-assisted and imperfect. Each data repository
ships its known limitations alongside the data — see
SCIENTIFIC_INTEGRITY.md
in MESS-Parameters for extraction-quality caveats, coverage gaps, and the
metrics that should not be quoted as point estimates. Please read it before
building on these numbers.
messai.io · questions: founders@messai.io