Ontology alignment

ChemDCAT-AP maps every class and slot to established ontology terms via class_uri and slot_uri. This page documents the ontology choices, the slot_uri replacement strategy, and the rationale for BFO alignment.

Ontologies used

Prefix	Ontology	Used for
CHEBI	Chemical Entities of Biological Interest	Chemical entity types (`ChemicalEntity`, `Atom`)
CHEMINF	Chemical Information Ontology	Structure descriptors (`InChi`, `SMILES`, `MolecularFormula`, `IUPACName`, `InChIKey`). Also used in PubChemRDF alongside SIO.
CHMO	Chemical Methods Ontology	Analytical methods, concentration (`Concentration`), yield (`Yield`)
SIO	Semanticscience Integrated Ontology	Chemical processes and substances (`ChemicalReaction`, `Reagent`, `Catalyst`, `DissolvingSubstance`, `SubstanceSample`), attribute predicate (`SIO:000008`), pH, density. Also used in PubChemRDF alongside CHEMINF.
OBI	Ontology for Biomedical Investigations	`MaterialSample`
BFO	Basic Formal Ontology	`MaterialEntity`, parthood (`BFO:0000051`)
IAO	Information Artifact Ontology	Information content entity concept; identifier class (`IAO:0020000`) used for reaction identifiers (e.g. RInChI) without a dedicated CHEMINF class
PATO	Phenotype And Trait Ontology	Physical states, close mappings for physical quantities
QUDT	Quantities, Units, Dimensions, Types	Quantity classes (`Temperature`, `Mass`, `Volume`, `Pressure`, etc.), quantity kinds, units
RO	OBO Relations Ontology	Reaction input/output (`RO:0004009`, `RO:0004008`)
RXNO	Named Reaction Ontology	Reaction types, catalyst role (`RXNO:0000425`)
MOP	Molecular Process Ontology	Molecular process types (exact mapping)
PROCO	Process Chemistry Ontology	Starting material (`PROCO:0000029`)
NCIT	NCI Thesaurus OBO Edition	Product, catalyst, reagent, pH (exact/close mappings)
VOC4CAT	Voc4Cat - A SKOS vocabulary for Catalysis	Catalysis-specific terms (exact mappings)
AFE/AFR	Allotrope Foundation Ontology	Equipment (`Reactor`), molar mass, concentration (exact mappings)
ENVO	The Environment Ontology	`Laboratory` (`ENVO:01001405`)

The `slot_uri` replacement strategy

DCAT-AP+ assigns intentionally generic predicates to its slots. The Dublin Core Terms predicates (dcterms:relation, dcterms:hasPart) are semantically weak by design, chosen as the lowest common denominator that avoids conflicting with domain-specific vocabularies. The PROV-O predicates (prov:used, prov:generated, prov:wasAssociatedWith) are well-defined within the provenance model but carry no domain-specific semantics: prov:used does not distinguish a starting material from a reactant, and prov:wasAssociatedWith does not distinguish a catalyst from a solvent.

DCAT-AP+ slot	Default `slot_uri`
`has_quantitative_attribute`	`dcterms:relation`
`has_qualitative_attribute`	`dcterms:relation`
`had_input_entity`	`prov:used`
`had_output_entity`	`prov:generated`
`carried_out_by`	`prov:wasAssociatedWith`
`has_part`	`dcterms:hasPart`

When ChemDCAT-AP creates a sub-slot (via is_a), it may assign a different, semantically richer slot_uri that replaces the parent's predicate in the generated RDF. This is a valid specialization per the DCAT-AP+ extension rules, but it has interoperability implications. The following table summarizes all replacements; the subsections below document the rationale for each.

ChemDCAT-AP sub-slot(s)	Parent slot	Parent `slot_uri`	Replacement `slot_uri`
`has_temperature`, `has_mass`, `inchi`, `smiles`, `has_concentration`, `has_yield`, etc.	`has_quantitative/qualitative_attribute`	`dcterms:relation`	`SIO:000008` (has attribute)
`used_starting_material`, `used_reactant`	`had_input_entity`	`prov:used`	`RO:0004009` (has input)
`generated_product`	`had_output_entity`	`prov:generated`	`RO:0004008` (has output)
`used_catalyst`	`carried_out_by`	`prov:wasAssociatedWith`	`RXNO:0000425` (has catalyst)
`has_part` (on MaterialEntity), `composed_of`, `has_reaction_step`	`has_part`	`dcterms:hasPart`	`BFO:0000051` (has part)
`used_solvent`, `used_reactor`	`carried_out_by`	`prov:wasAssociatedWith`	(not replaced)

SIO:000008 as the chemistry attribute predicate

All ChemDCAT-AP attribute sub-slots (has_temperature, has_mass, inchi, smiles, has_concentration, has_yield, etc.) use slot_uri: SIO:000008 (has attribute) instead of the DCAT-AP+ default dcterms:relation. The rationale:

dcterms:relation is maximally generic and says nothing about the nature of the relationship.
SIO:000008 is a well-established predicate in the Semanticscience Integrated Ontology that means "has attribute." It is semantically more appropriate for linking an entity to its characteristics.
SIO is already used elsewhere in ChemDCAT-AP for class mappings (ChemicalReaction, Catalyst, etc.), so adding it as a predicate does not introduce a new dependency.

RO predicates for reaction inputs/outputs

The reaction participant slots use OBO Relations Ontology predicates instead of the PROV-O defaults:

ChemDCAT-AP slot	Replaces	New `slot_uri`	Meaning
`used_starting_material`	`had_input_entity` (`prov:used`)	`RO:0004009`	has input
`used_reactant`	`had_input_entity` (`prov:used`)	`RO:0004009`	has input
`generated_product`	`had_output_entity` (`prov:generated`)	`RO:0004008`	has output

RO:0004009 (has input) and RO:0004008 (has output) are process-to-entity relations from the OBO Relations Ontology. They are more specific than prov:used and prov:generated and align with OBO Foundry practices.

RXNO for catalyst association

The used_catalyst slot uses slot_uri: RXNO:0000425 (has catalyst) from the RXNO Ontology, replacing the parent carried_out_by's prov:wasAssociatedWith. This predicate precisely captures the catalytic role. used_solvent and used_reactor keep the parent's prov:wasAssociatedWith because there currently is no established predicate that provides equivalent precision for these roles.

BFO:0000051 for parthood

has_part on MaterialEntity, ChemicalEntity, composed_of, and has_reaction_step uses slot_uri: BFO:0000051 (has part), replacing the DCAT-AP+ default dcterms:hasPart. BFO's parthood relation carries formal mereological axioms (transitivity, reflexivity) that dcterms:hasPart does not.

Interoperability implications

When a ChemDCAT-AP instance is serialized to RDF, the triples use the domain-specific predicates (SIO:000008, RO:0004009, etc.), not the DCAT-AP+ defaults (dcterms:relation, prov:used, etc.). This means:

A SPARQL query written against DCAT-AP+ predicates will not find ChemDCAT-AP data unless the query accounts for the replaced predicates.
Conversely, a query written against ChemDCAT-AP predicates will find exactly the right data with higher precision.

To bridge both query patterns, use SPARQL CONSTRUCT rules that add DCAT-AP+ predicates alongside the domain-specific ones:

CONSTRUCT {
  ?entity dcterms:relation ?attr .
}
WHERE {
  ?entity SIO:000008 ?attr .
}

This is the same approach recommended in the DCAT-AP+ design patterns documentation for avoiding blank node duplication when projecting between vocabularies. Generate RDF from the domain-specific schema and use CONSTRUCT rules to add the generic predicates to the existing nodes.

Planned: formal projection queries

We plan to provide a complete set of SPARQL CONSTRUCT queries covering all predicate replacements listed above, with worked examples from the ChemDCAT-AP test data. These will allow any triplestore hosting ChemDCAT-AP data to materialize the DCAT-AP+ base predicates alongside the domain-specific ones, ensuring backward-compatible querying without losing semantic precision.

Why BFO alignment matters

Several of ChemDCAT-AP's ontology choices (CHEBI, OBI, RO, PATO) are part of the OBO Foundry, which is built on the Basic Formal Ontology (BFO). This alignment has practical value:

A formal mapping between PROV-O and BFO has been published. This means ChemDCAT-AP data (which uses PROV-O via DCAT-AP+ and BFO-aligned terms via OBO ontologies) can be reasoned over coherently without manual bridging.
OBO Foundry ontologies share a common upper-level structure, so terms from CHEBI, OBI, CHMO, and RO interoperate without custom alignment work.
BFO's material entity (BFO:0000040) is the natural class_uri for MaterialEntity, and BFO:0000051 (has part) is the established parthood relation in this ecosystem.

Why SIO

The Semanticscience Integrated Ontology (SIO) is used in ChemDCAT-AP for both its has attribute predicate (SIO:000008) and several class mappings. Three reasons drove this choice.

BFO-compatible upper level

SIO has its own upper-level structure, but it aligns naturally with BFO. SIO's three top-level branches under entity correspond to BFO's continuant subtypes, though not perfectly:

attribute (with subtypes quality and realizable entity) corresponds to BFO's specifically dependent continuant (SDC) branch
object groups together material entity, information content entity, and spatial region. In BFO, these are split: material entity and spatial region (via immaterial entity) fall under independent continuant, while information content entity is a generically dependent continuant (GDC), a sibling branch to both independent continuant and SDC
process corresponds to BFO's occurrent/process branch

A formal OWL bridge module has not been published yet, but producing one would be a straightforward alignment task given this structural correspondence.

Established prior art in PubChemRDF

SIO is used alongside CHEMINF in the PubChemRDF graph (Fu et al. 2015). The concrete overlap is the compound descriptor pattern: PubChemRDF links compounds to CHEMINF-typed descriptors via SIO:000008 (has attribute), which is the same predicate and descriptor vocabulary ChemDCAT-AP uses for its chemical entity attribute slots (inchi, smiles, molecular_formula, etc.). This shared pattern means SPARQL queries over chemical descriptors can target both ChemDCAT-AP and PubChemRDF data using the same predicate and type assertions.

Gap-filling for missing OBO classes

The SIO classes reused in ChemDCAT-AP (e.g., SIO:010345 for chemical reaction, SIO:010344 for catalyst) were chosen because no equivalent existed in OBO ontologies. Given SIO's BFO-similar upper level, their use was considered appropriate. If OBO equivalents are minted in the future, the class_uri mappings may change, but the SIO correspondence would be preserved via exact_mappings or close_mappings in the schema.

Why most quantity classes map to `qudt:Quantity`

A glance at the schema shows that Temperature, Mass, Volume, Pressure, AmountOfSubstance, and others all carry class_uri: qudt:Quantity. This looks like a missed opportunity for more precise ontology mappings. It is, but the gap is not in ChemDCAT-AP; it is in the OBO/BFO ontology ecosystem.

The issue comes down to a BFO distinction between three layers:

Qualities (specifically dependent continuants in BFO): the physical property inhering in a material entity. PATO provides these: PATO:0000146 (temperature), PATO:0000125 (mass), PATO:0000918 (volume), etc. In QUDT terms, these correspond to QuantityKinds.
Information content entities (ICEs): a class from the OBO/BFO ecosystem (defined in the Information Artifact Ontology, IAO) that represents an information artifact about a quality, i.e. the recorded measurement value as it appears in a dataset or lab notebook, not the physical property itself. In QUDT terms, this corresponds to a Quantity (a value + unit + quantity kind). This is what ChemDCAT-AP's attribute classes actually represent: a recorded value, not the quality itself.
The QUDT model collapses layers 1 and 2 into a single qudt:Quantity class that carries both the quantity kind reference and the numeric value.

The problem: OBO/BFO ontologies provide good coverage of layer 1 (PATO qualities) but almost no coverage of layer 2 (ICEs about those qualities). There is no established OBO class for "a recorded temperature measurement" as distinct from "the quality of temperature." Chris Mungall has created a PATO-to-QUDT QuantityKinds SSSOM mapping, which bridges layer 1 across the two ecosystems. But what ChemDCAT-AP needs for its class_uri values are layer 2 classes, and those mostly do not yet exist in the OBO world.

Four exceptions exist where a more precise class is available:

Density is mapped to SIO:001406, an information content entity about density in SIO.
Yield is mapped to CHMO:0002855 (yield percentage), a CHMO class representing the recorded yield value.
Concentration is mapped to CHMO:0002820, a CHMO class for concentration measurements.
Molar mass is mapped to AFR:0002409, an Allotrope Foundation class for molar mass values.

These four demonstrate that precise mappings are possible when the right class exists in an established ontology. For the remaining quantities, ChemDCAT-AP maps to qudt:Quantity and uses has_quantity_type (pointing to QUDT QuantityKinds) to differentiate them. The PATO terms appear as close_mappings rather than class_uri values, because they represent the quality, not the information entity about the quality.

Future work

The long-term goal is to mint proper ICE classes for recorded quantity values in a suitable OBO ontology, allowing class_uri mappings that are both BFO-aligned and semantically precise. Until then, the QUDT mapping combined with has_quantity_type typing is the pragmatic solution, and the close_mappings to PATO terms preserve the link to the OBO ecosystem for future alignment.

Known workarounds

Duration on sub-steps

The DCAT-AP+ has_duration slot is only available on EvaluatedActivity and its subclasses (like ChemicalReaction). For sub-steps of a reaction that are modeled as generic Activity instances (dissolving, stirring, drop-wise addition), has_duration is not in scope. As a temporary workaround, these sub-steps express their duration using the generic has_qualitative_attribute slot with type: schema:Duration and an ISO 8601 value (e.g. PT10M). This is tracked in DCAT-AP+ issue #69. Once has_duration is made available on Activity, the workaround will be replaced.

Mapping annotations

ChemDCAT-AP uses LinkML's exact_mappings, close_mappings, broad_mappings, and narrow_mappings to document cross-ontology correspondences. These are documentation annotations, not formal OWL axioms. More mappings will be added over time, and it will be explored in how far they can be used to produce SSSOM mapping sets, which can then be transformed into OWL bridge modules using the Ontology Development Kit (ODK).

Annotation	Meaning	Use in ChemDCAT-AP
`exact_mappings`	Equivalent semantics	`ChemicalReaction` maps exactly to `MOP:0000543`, `REX:0000002`, `AFP:0003711`
`close_mappings`	Similar but not identical	`Concentration` close to `PATO:0000033` (PATO captures the quality, not the measurement)
`broad_mappings`	Target is more general	`SubstanceSampleCharacterization` broadly maps to `OBI:0000070` (assay), which covers more than just substance characterization
`narrow_mappings`	Target is more specific	`ChemicalReaction` has narrow mapping to `RXNO:0000329` (a planned synthesis)