Syntropa

Tier 0 · open

Discovery reports

Each report shows the whole funnel, including what was discarded and why. A result without its drop reasons is a claim, not evidence. Open any report to read the full run on-site.

The Gap Report is the companion to these. Where a discovery report follows one funnel end to end, the Gap Report maps which two-step relay routes the corpus can already test and which it cannot reach, and separates a gap in the corpus from a gap in the science. Read the Gap Report →

Market-coverage map

The platform is run against a standing pool of real-world demands (the biomanufacturing, remediation, and critical-mineral-recovery problems that national programs prioritize), with an honest verdict published for each. Coverage here is a measured property, not a marketing claim: what the corpus can express, what is queued for a full run, and what is a genuine gap are all first-class results.

The gaps are not failures. They are a structured research roadmap. Each falls into a defined bucket with a defined fix, so “the corpus cannot do this yet” always arrives with the reason and the work that closes it. Publishing the gaps honestly, rather than dressing every screen as a win, is what makes both the capability and the roadmap credible.

Vocabulary absence 7

The target molecule or reaction is in no reconstructed model’s metabolic vocabulary, so no route can exist yet. Fix: corpus expansion by ingesting curated models that carry the missing chemistry.

e.g. Rare-earth-element bio-recovery from low-grade secondary waste · Bioleaching of Co/Ni/Li/Mn from spent Li-ion battery black mass · Bioleaching of Cu/Au from waste printed circuit boards (e-waste)

Universal re-export 5

The target is a metabolite nearly every model already exchanges, so a raw producer count is non-selective, not a discovery. Fix: selective, defined-medium modeling that isolates net production.

e.g. Struvite P-recovery from municipal wastewater (EBPR) · In-planta associative N2 fixation for cereals (corn/wheat/cotton) · Microbial phosphate solubilization of fixed soil-P / rock phosphate

Outside flux-balance scope 4

The chemistry is electrochemical, physical, or enzyme-kinetic, not stoichiometric flux balance. Fix: added modeling layers (bioelectrochemical, kinetic) in a later phase.

e.g. Methane-to-single-cell-protein (methanotrophic feed protein) · Post-consumer polyurethane foam feedstock to polyols/diamines via urethanases + laccase · Two-stage DF + microbial electrolysis (DF-MEC) biohydrogen

Substrate present, product absent 1

The feedstock is representable but the product pathway is reconstructed in no model. Fix: targeted pathway curation.

e.g. Microbial Cr(VI) -> Cr(III) reduction in tannery/plating effluent

Configured as a free nutrient 1

The target is modeled as a freely available nutrient, so it cannot register as a made product. Fix: medium reconfiguration for the specific case.

e.g. Selenium reduction from coal-mining / FGD / refinery wastewater

Targeted pathway curation 4

A specific conversion (chitin to GlcNAc, azo-dye mineralization, lignin to adipic acid) is reconstructed in no model in the current corpus. Fix: case-by-case pathway curation.

e.g. lignin -> cis,cis-muconate -> adipic · chitin/shell waste -> GlcNAc · dairy whey/permeate -> biogas (methane) via anaerobic-digestion syntrophy

These map to three application domains: environmental remediation (uranium, perchlorate, chromium(VI), 1,4-dioxane, PFAS), critical-mineral recovery (rare earths, battery metals, e-waste gold, acid-mine-drainage zinc and copper), and low-carbon biomanufacturing (1,3-propanediol, 2,3-butanediol, biohydrogen). Several already resolve to a route in the corpus; the rest define exactly where the platform earns its next increment of coverage. The 17 queued runs are the compute-heavy screens a first funded phase is built to complete and de-risk.

corpus v0.2.2 · hash 58a8f20e2aba...

Every model used in a run is named with its source and license. Reports marked CC-BY use only redistributable CC-BY models (attribution required; the CC-BY basis for BiGG-derived reconstructions is stated on the Data page); those marked some link-only models name a source Syntropa cannot re-serve, with a link to obtain it. Naming a model is not redistribution.

Worked examples & method validation

These worked examples show the discovery funnel end to end: the pool screened, the candidates found, the flux-balance verdict, the carbon-balance audit, and every drop reason. Read them for the method, not for a flagship win: most resolve to an honest single-organism result or a positive control that recovers known biology, and none is yet a novel, literature-consistent multi-species consortium discovery. That is the bar Syntropa is working toward, and it publishes the honest state of it rather than dress a single-organism or biomass-scaling result as cooperation. The thermodynamic check today is class-level and necessary-only (per-candidate TFA is Phase-2), and every candidate is a hypothesis until it is cultured.

Negatives, pending, and not-expressible use-cases

Syntropa publishes the negatives as first-class findings. NO CREDIBLE ROUTE means the run completed but the specialist catalysts the conversion needs are absent from the corpus, so every flux-feasible finisher is a gap-fill artifact (e.g. syngas and medium-chain-fatty-acid targets). HONEST-NEGATIVE means the run found no feasible producer at all. NOT-EXPRESSIBLE means no route exists through the current corpus (a research or literature gap). PENDING means a defined target with no discovery run yet. None of these is an error. Each maps a real gap (see the Data page).

How to read these reports

CONSORTIUM-ONLY (complementation) means no single organism in the corpus produces the target, but a multi-member community does. Reducible means a smaller subset of that community still works; irreducible means every member is needed. This is metabolic complementation (a community reaches what no member reaches alone), measured against the best single organism, not a tested super-additivity: Syntropa does not compare community output against the arithmetic sum of each member run alone, so it never uses the term “super-additive.” At current gap-filled corpus quality, rigorous irreducibility needs thermodynamic modeling on defined-medium-ready models (Phase 2). That corpus quality has since been measured, on the 31 benchmark consortia rather than across every case on this page, and the result is a limit on the verdicts it covers. The reactions carrying flux on a production route are gene-unsupported at 1.53 times the rate of the model as a whole, and zeroing only the gap-filled reactions on the route stops production in 31 of 31, where zeroing the same number of randomly chosen gene-supported reactions leaves production alive in 21 of 93 draws, three per consortium. So those routes run on chemistry the reconstruction never attached to a gene, and they do so specifically rather than as a side effect of how much was removed. Whether the same holds for the other cases here has not been measured.

SINGLE-ORGANISM means a single organism does the chemistry and no cross-feeding emergence was found, so no consortium is required. Syntropa deliberately does not issue a “consortium advantage” verdict from a raw flux ratio: on an open medium, adding members adds their own biomass and uptake, so a multi-member combination reaching a higher flux than the best single is expected with zero cooperation. Until a biomass-fixed, summed-member test is run (Phase 2), a higher multi-member number is biomass scaling, not a discovery, so this is an honest single-organism result, not a failure.

NO CREDIBLE ROUTE and HONEST-NEGATIVE are published negatives, listed with the pending and not-expressible use-cases above: the run completed but the corpus has no credible route (gap-fill artifacts only) or no feasible producer at all.

NOT-EXPRESSIBLE means no path through the corpus reaches the target metabolite at all. This may mean the biology does not exist in any reconstructed model, or that an intermediate step requires a model not yet in the corpus.

Each report is provenance-stamped (corpus version, content hash, tool version + code hash, and an environment lock) so an individual result is traceable and content-auditable independently of the full corpus. The tool tree is committed, so the exact code that produced a report is recorded; a tagged, published tool release is still pending, so independent replication today rests on the content hash. The corpus release is registered with a Zenodo-citable DOI (see Data).