Which reactor and process parameters decide microbial electrosynthesis, and by how much

The question the predictor answers

query    strain = C. ljungdahlii · cathode = carbon felt   (example inputs)
         E_cathode = −0.60 V vs Ag/AgCl (sat. KCl)  →  ≈ −0.40 V vs SHE (25 °C offset; ≈ −0.41 V with the reference at 35 °C) · pH 6.8 · 35 °C
         CO₂ supply = ?  (flow, partial pressure, kLa)
answer   acetate_rate  { value: pending, unit: "g/L/d",  ci_low: pending, ci_high: pending }
         titre         { value: pending, unit: "g/L",    ci_low: pending, ci_high: pending }
         FE_acetate    { value: pending, unit: "%",      ci_low: pending, ci_high: pending }
         energy        { value: pending, unit: "kWh/kg", ci_low: pending, ci_high: pending }
drivers  cathode potential vs SHE · CO₂ supply · current density · cathode area   (ranked: pending)
evidence harmonised MES literature + John-Oliver’s reactor runs (pending)
missing  CO₂ supply  →  named as missing, no default filled in, answer flagged

How this works

The path from a published paper to a number on this page. Each step feeds the next. The first three show what this site holds right now for the papers cited here.

1. 1

   Find the paper

   Partly done

   A paper enters the corpus as a record: DOI, title, authors, journal, date and abstract, taken from Crossref or the publisher.

   • 10 of 16 cited papers with a DOI are in this corpus.
   • Counted in this deployment on Oct 28, 2026.
   • On Oct 8, 2026 some of the papers cited by the protocol chart were added to the staging corpus, as records with no stored PDF and no extracted values. The counts in the first three steps are read from the deployment you are on.

   Open the corpus
2. 2

   Get the full text

   Partly done

   Where the publisher allows it, the PDF is stored and converted to text, tables and equations. Values can also be read from open-access full text without storing the PDF. A paper with neither stays at its abstract: only what the abstract states is available, without its tables or methods.

   • 5 of 10 have a stored PDF.
3. 3

   Extract values with their evidence

   Partly done

   A language model reads the text and tables and writes every value as one row: parameter, number, unit, the conditions it was measured under, the quote it came from, a confidence and the extractor version.

   • 5 of 10 have extracted values (239 distinct facts in all; a value extracted twice counts once).
   • These counts cover the papers the protocol chart cites. The readiness map below counts every paper in its scope (staging database, Oct 13, 2026).
4. 4

   Harmonise

   Partly done

   Each row is mapped to a canonical parameter where a rule matches, so “power density” means one thing across papers, and its unit is converted to SI. Conversion is not comparability: the species, the reference electrode and the area or volume basis must match too. Those checks are incomplete, so some rows that cannot be compared still reach the models.

   • Part of the distinct facts in the MES scope are already harmonised and flagged as modelable; the share is given in the section “What blocks comparison today” (staging database, Oct 13, 2026). The fixes for the problems listed there are still to do.

   Open the parameters
5. 5

   Model and predict

   Partly done

   Models are fitted on harmonised rows only: first the effect of one parameter inside each study, then pooled across studies. A prediction returns a value, a unit, an interval, a confidence and its source. The aim is that a missing input is named, never filled with a default; some MESSAI predictors still fill one today, such as an applied voltage of 0.6 V for an MEC.

   • Most of the methods already exist in MESSAI. The predictor for MES is still to be built, and nothing is fitted yet.

   Read how the predictions are made
6. 6

   Check it

   Links listed

   The papers behind the protocol chart are listed below with a link to each publisher. The chart’s steps are steps of the shared protocol graph. Signed-in readers can comment on this page. A listed paper is a pointer for checking, not a check MESSAI has made.

   Go to the papers

Correlation-readiness map

Rows are 12 groups of reactor and process parameters, columns are 7 groups of outcomes. Each cell counts the papers in the microbial-electrosynthesis scope that report both, and how many of those have an SI-converted value for both (“usable”). This is paper-level co-reporting, not an effect estimate: it shows where a correlation could be quantified today and where it cannot. Select a cell for what blocks it and how within-study variation would quantify the effect.

What blocks comparison today

The map shows how few pairs are usable. This section says why, from the same staging snapshot: the scope, the key parameters, and six harmonisation problems, each with a proposed fix.

A reporting standard for MES reactors

A minimum reporting standard is what lets a reader rebuild a run and compare it with another lab’s. John-Oliver’s work on characterising and standardising reactors is the missing input. Below: what MESSAI requires of an electrosynthesis study today, then what reactor standardisation would add.

How MESSAI turns this into predictions

MESSAI already runs this pipeline on the bioelectrochemistry corpus. For MES it needs the harmonisation fixes above, a parameter set and his reactor and run data. The steps run in this order because each one feeds the next.

1. 1

   Within-study effects

   Exists in MESSAI

   For each study, estimate how one parameter moves an outcome while the others are held fixed: cathode potential on acetate rate, CO₂ supply on faradaic efficiency. This is the cleanest evidence for a trend: differences between labs drop out, though confounders inside a study, such as run order or biofilm age, remain. MESSAI fits these as population-level Bayesian effects and shows them in the lab’s sensitivity chart (for example, hydraulic retention time → COD removal in wastewater treatment, which has not converged yet). Runs that vary one parameter at a time are the ideal input; which ones he has is an open question for him.

2. 2

   Cross-study pooling

   Exists in MESSAI

   Pool the same effect across MES papers with a hierarchical meta-analysis (MESSAI’s meta-analysis forest plots). This answers whether an effect of potential or CO₂ supply holds for other strains, reactors and labs, and how much it varies between studies. It works only on values on one basis, which is what the harmonisation fixes are for.

3. 3

   Correlation structure between parameters

   Exists in MESSAI

   Some parameters move together: a more negative cathode potential raises current density, and the area and volume used fix which basis a rate is on. MESSAI’s learned dependency graph and knowledge graph record such links, so the predictor doesn’t count one effect twice. Correlation across studies isn’t causation, so causal claims rest on step 1.

4. 4

   Compact laws by symbolic regression

   Exists in MESSAI

   MESSAI already uses symbolic regression (PySR) to search for short equations between parameters. For MES the target is a readable law such as acetate rate = f(cathode potential vs SHE, CO₂ supply, current density, cathode area) that a researcher can test, rather than a black box.

5. 5

   Predictor with intervals and drivers

   To build for MES

   A hierarchical model with study as a random effect, using the effects from steps 1–4 as priors and his reactor and run data as the core data. His own studies would be left out of those priors, so his runs are not counted twice. It would return each prediction with an interval and a ranked list of the parameters driving it. The intervals would still need calibrating on held-out studies: MESSAI’s served predictors do not return calibrated intervals yet.

   outcome ~ f(E vs SHE, j, CO₂ supply, kLa, pH, T, cathode, area, volume, strain, …) + (1 | study) + residual GP

   To be validated by

   leaving one study out at a time. The real test is predicting another lab’s results, not interpolating within one reactor.

   Planned guardrail

   outside the conditions covered by data, the answer is flagged as an extrapolation; a wider interval alone would not make it reliable

Protocol

The microbial-electrosynthesis methodology as a graph of steps, from the medium and the cathode to the electron balance and the acetate rate. Click a step for its setpoints, readouts, precautions and sources. Left to right is execution order, and every step links to the full /protocols graph. A selected step also shows which items of the reporting standard it records.

Models

The cell the predictor is about, the hydrogen route acetogens can use, and the downstream step that turns the acids into fuel. All are shared MESSAI 3D models; drag to orbit.

Papers behind this page

Every paper the protocol chart cites, with what it is cited for and where you can check it. Each row links the publisher and, if this site holds the paper, its corpus record.

Our to-dos (MESSAI)

In priority order. Estimates are rough and assume one person part-time.

1. Fix MES harmonisation (a)–(e)~1–2 weeksNot started

   Unit-spelling aliases, a separate areal-rate parameter, molar-to-mass conversion, reference-electrode conversion to SHE, and one canonical parameter per quantity. Staging first, a dry run, then prod. Every cross-study comparison on this page depends on it.

2. MES taxonomy fix~2–3 daysNot started

   Separate the system type “microbial electrosynthesis” from the field name “MES”, then re-classify, so the 397 papers that mention electrosynthesis can be selected by label.

3. Use-case spec for microbial electrosynthesis~2–3 daysNot started

   A microbial-electrosynthesis spec in @messai/use-case-specs, after the nitrogen one: scope, the parameters and outcomes of the map with canonical names and units, plausible ranges, and potentials stored against SHE. To be sent to John-Oliver to correct.

4. MES section of the minimum-reporting checklist~1 week, with himNeeds John-Oliver

   Turn the proposed additions into a checklist section, co-written with him, and add it to the reproducibility taxonomy so the scorer checks it.

5. Re-measure this map after the fixes~1 dayAfter the fixes

   Re-run scripts/collab/measure-mes-readiness.sh (read-only, staging) and compare: which pairs gained usable papers, which stay empty.

6. Within-study effects and predictor v1~3–4 weeksWaiting on data

   Encode his runs and the harmonised literature as within-study effects, pool them, and fit the hierarchical model with conformal intervals, ranked drivers and a leave-one-study-out validation report, including where it fails.

7. Reactor and run template aligned with ReactorGeometry~3–4 daysNeeds his parameter list

   A CSV template whose columns match ReactorGeometry and the run conditions (area and volume with their basis, reference electrode and offset against SHE, gas supply, run duration), so his runs land with full provenance.

8. Measurement upload~1 week (roadmap)Planned 21 Nov 2026

   The upload route (roadmap C3-M1) so lab runs land in the database with “measured” provenance.

9. In-app what-if~3–4 daysLater

   /collab/gath in English and German with parameter inputs, a prediction with interval, drivers and the live readiness map.

Deliverables to UFZ

L0 is the day John-Oliver’s parameter list arrives, T0 the day his reactor and run data arrive. Nothing is promised before the data exists; the harmonisation fixes and the reporting-standard draft do not need it.

1. This weekThis page and the proposed reporting items

   The readiness map, what blocks comparison and the proposed additions to the reporting set, for him to correct, with the open questions below.

2. L0 + 2 weeksMES parameter set v1

   Canonical names and units for the parameter and outcome groups of the map, aligned with his list, with potentials stored against SHE and the rate basis stated.

3. In ~2 weeks, no data neededHarmonisation fixes and a re-measured map

   Fixes (a)–(e) on staging, then the readiness map measured again: which pairs gained usable papers.

4. T0 + 1 weekHis runs encoded

   Within-study effects from his runs in the harmonised format, with the reactor and run template filled, or a note on what doesn’t fit.

5. T0 + 4 weeksReadiness map with numbers

   Effect size, interval and number of studies for each filled pair; empty pairs ranked by importance.

6. T0 + 6 weeksPredictor v1 and validation

   Predictions of acetate rate, titre, faradaic efficiency and energy with intervals and drivers, tested by leaving one study out at a time.

7. T0 + 8 weeksReporting-standard section, interactive page, methods note

   The MES section of the checklist co-written with him, the prediction tool in MESSAI, a co-authored methods note draft, and the dataset on Zenodo with a DOI if UFZ agrees.

What we ask of John-Oliver and UFZ

• His parameter list (L0): which reactor and process parameters he records or standardises, with the units and the basis he uses.
• Reactor and run data (T0), including runs that failed or stopped early, in our CSV template, and any runs that vary one parameter at a time.
• A check of the proposed reporting items and of the blocked cells in the map.
• 30 minutes every two weeks to review results, an agreement on data terms and authorship, and a pointer to whoever coordinates data standards in SPP 2240, if anyone does.

Open decision

Which model should answer lab-design chat?

Lab-design answers take about 50 seconds today. A smaller model answers the same questions in about 19 seconds, for about 3% of the cost, but looks up the corpus less often. We would like your view on which matters more for your work.

Method: 10 lab-design questions per model, sent in alternating order on Oct 26, 2026; no answer fell back to another provider. Times are measured on the server and exclude your network. Prices are the AI Gateway list prices; cost per answer is an estimate from median token counts with 81% of input read from cache.

Open questions for John-Oliver

1. Which rate basis is primary: volumetric (g/L/d) or areal (g/m²/d)? Which of the two lets two reactors be compared?
2. Potentials against which reference? SHE, or a named reference electrode, and with which KCl concentration? Is it stated for every run?
3. Which acetogens and which cathode mode? For example C. ljungdahlii or S. ovata; fixed potential, galvanostatic, or H₂-mediated.
4. How is CO₂ supplied and measured? Flow, partial pressure, kLa: which does he control and which does he report?
5. What makes reactors comparable across labs? Which geometry and process parameters would he fix in a standard, and which may vary?
6. Does SPP 2240 have a shared data standard or reporting template that this work should follow?
7. Would he co-own the MES section of the checklist, as co-author of the standard?