X-Mech Suite: eight knowledge bases, one standard

The X-Mech suite is a fleet of eight curated, ontology-grounded knowledge bases that together describe a microbe at every scale: the habitat it lives in, the community it belongs to, the traits it expresses, the structures and proteins that implement them, the antibiotics that act on it, and the ingredients and media it is grown in. Each Mech follows the same curation model, one validated YAML record per entity with evidence and provenance, and the fleet is coordinated by a single orchestrator, culturebotai-claw.

8Mech knowledge bases
445,478curated records
39ontologies & databases cited
1orchestrator (claw)
Habitatwhere it lives
Communitywho it lives with
Traitswhat it does
Cell structureswhat it is built of
Proteinsthe machinery
Antibioticswhat acts on it
Ingredientswhat it is fed
Mediawhere it is grown
Layers
Vocabulary
Relationship graph of the eight Mech knowledge bases Nodes are Mechs arranged in a ring from habitat to culture medium; chords are shared ontology terms, dashed arcs are direct cross-references, and spokes connect to the central kg-microbe knowledge graph.
shared ontology terms (width = how many) direct cross-reference (arrow = who consumes) export to / input from kg-microbe fleet manifest membership

The eight Mechs

Each card carries its Mech’s own site color. Hover a card to trace its ties in the graph above; use “Show in graph” to select it.

HabitatMech

Habitat

Four habitat vocabularies harmonized into ENVO-grounded records that keep every source's attestation.

3,213habitat records · 684 reviewed
ENVONCBITaxonBTOUBERONFOODONGOLDBacDive

CommunityMech

Community

Curated knowledge base of microbial communities, their interactions, cultivation conditions and evidence.

325community records · KGX export
NCBITaxonChEBIGOENVOGTDBPMID

TraitMech

Traits

Microbial ecophysiological trait knowledge base seeded from METPO, one curated YAML per trait, with causal mechanism graphs.

477trait records · 353 with causal graphs
METPOGONCBITaxonChEBIUniProtPATO

CellStructureMech

Cell structures

Organelles, envelope layers, appendages, microcompartments and complexes: components, distribution, function and causal mechanism.

39structure records · 279 causal edges
GONCBITaxonUniProtMETPOPfamPDB

ProteinTraitsMech

Proteins

Protein sequence, structure and function trait classes seeded from InterPro, Pfam, Rhea, CATH, SCOPe, CARD and more.

429,271protein trait records
InterProUniProtRheaPfamGOChEBIARO

AntibioticMech

Antibiotics

One record per antimicrobial chemical structure, harmonizing ChEBI and CARD with targets, mode of action, resistance and evidence.

2,909antimicrobial structures · 92% ChEBI-grounded
ChEBIAROCASPubChemDrugBankNCBITaxon

MediaIngredientMech

Ingredients

LLM-assisted curation of media-ingredient ontology mappings with full audit trails; owns ingredient identity for the fleet.

2,958ingredient records · 91% mapped
ChEBICASNCITFOODONENVOMeSH

CultureMech

Media

Versioned, ontology-grounded knowledge base of culture-media recipes from MediaDive, TogoMedium, KOMODO and the major collections.

6,286merged recipes · 15,877 normalized
ChEBIKEGGFOODONUBERONCASNCBITaxon

Shared vocabulary

The Mechs are joinable because they ground records in the same public ontologies. The table counts identifier occurrences per vocabulary in each Mech’s record corpus; darker cells mean more. Click a column to filter the graph to that vocabulary.

Counts are prefix occurrences in the canonical record directories (merged recipes for CultureMech, communities for CommunityMech, habitat records for HabitatMech) as of September 2026. ChEBI binds the chemistry arm (media, ingredients, antibiotics, proteins); NCBITaxon and ENVO bind the organism arm (habitat, community, traits); GO and METPO bridge phenotype, structure and protein.

How the Mechs reference each other

Beyond shared vocabulary, Mechs name one another directly, in record fields, in schema slots, and in the curation practices they adopt from each other. Arrows point at the Mech that consumes.

CultureMechMediaIngredientMech
Unmapped ingredient strings and occurrence counts become MIM's curation backlog; MIM records link back to recipes by stable id with the CultureMechReference class.Record data · claw ingredient_curation_pipeline
MediaIngredientMechCultureMech
Curated ChEBI and FOODON mappings sync back to recipes. CultureMech vendors MIM's ingredient-role enums and consumes MIM's immutable label index.Record data · src/culturemech/schema/mim_roles.yaml
CommunityMechCultureMech
Community records name the medium they were cultivated in by CultureMech id: 21 records carry 32 culturemech_id links.Record data · docs/cross_repo_linking.md
CommunityMechMediaIngredientMech
The RelatedIngredient.mediaingredientmech_id slot is declared for MIM ingredient ids; no record populates it yet.Schema slot · docs/cross_repo_linking.md
CellStructureMechTraitMech
A structure lists the phenotypes it confers as TraitMech or METPO terms (8 trait links today), and its causal-graph node vocabulary is TraitMech's plus a STRUCTURE node type.Record data · schema slot associated_traits
CellStructureMechProteinTraitsMech
A single protein is a component of a structure, never a record of its own: it grounds to InterPro or UniProtKB and hands off to ProteinTraitsMech.Scope boundary · docs/CURATION.md
HabitatMechTraitMech
Habitat causal graphs use a superset of TraitMech's node vocabulary so the graphs stay comparable; a node may be a TraitMech or METPO trait.Schema · CausalNode enum
HabitatMechCultureMech
The single overlapping concept, BTO:0000316 culture medium, is handed to CultureMech rather than curated twice. A culture medium is a synthetic habitat.Curation decision · curation/decisions.tsv
AntibioticMechCellStructureMech
AntibioticMech wrote the licensed source-queue pattern (rank candidate sources, verify licences, record decisions); CellStructureMech adapted it.Practice · claw docs/guides/SOURCE_QUEUE.md
ProteinTraitsMechTraitMech
TraitMech adopted ProteinTraitsMech's licence-bearing download.yaml source catalogue, and a drift audit keeps the trait tokens the two share aligned in meaning.Practice · TraitMech download.yaml

Orchestration: culturebotai-claw

culturebotai-claw is the fleet’s coordinator. It does not hold science of its own; it holds the definition of the fleet, the artifacts every Mech must share byte-for-byte, and the cross-repository pipelines that move curation between Mechs safely.

Fleet manifest

fleet.yaml is the single source of truth for which repositories form the fleet. Seven Mechs are members today; HabitatMech follows the standard and is not yet a manifest member. Every member declares every capability exactly once as enabled, disabled or not applicable, with a reason, so nothing is silently off.

Vendored governance

Shared LinkML modules (mech_shared.yaml, history.yaml), validators and behavioral contracts live in claw and are vendored into each Mech byte-identically, pinned to one immutable claw commit and checked in CI. Fifteen artifacts, all seven members.

Pipelines and skills

Three cross-repository pipelines (ingredient curation, unified ingredient mapping, ENVO environment curation) and 24 agent skills. Cross-repo writes resolve exact worktree roots, take a repository lock, default to dry run and validate staged output before replacing source data.

Which fleet contracts each member has adopted, from the manifest:

MechCuration historyStrict validationVendored syncDeep researchKnowledge-gap scanEnvironment coverageSSSOM exportKGX exportSource queueSource catalogueSite contract
CultureMech
MediaIngredientMech
CommunityMech
TraitMech
ProteinTraitsMech
AntibioticMech
CellStructureMech
HabitatMechnot yet a manifest member; vendors mech_shared.yaml without a pinned claw revision
enableddisabled, with a recorded reasonnot applicable to this corpus

What makes a Mech

The suite is cohesive because every Mech is built the same way. The fleet standard, documented in claw’s MECH_STANDARD.md, asks for:

Bibliography

  1. Santangelo BE, Hegde H, Caufield JH, Reese J, Kliegr T, Hunter LE, Lozupone CA, Mungall CJ, Joachimiak MP. KG-Microbe — Building Modular and Scalable Knowledge Graphs for Microbiome and Microbial Sciences. GigaScience. 2026;giag077. doi:10.1093/gigascience/giag077
  2. Naseem S, Miller MA, Martinez-Gomez NC, Sun N, Joachimiak MP. MicroGrowAgents: An Agentic AI System for Microbial Cultivation Engineering. bioRxiv. 2026. doi:10.64898/2026.06.04.729985

See the full bibliography on the Publications page.