Ontologies: OBO and OWL

Formats

kgdiff reads OBO 1.4, and OWL ontologies serialized as RDF: RDF/XML (the .owl files Protégé and ROBOT write by default), Turtle, N-Triples and JSON-LD. OWL/XML, Functional Syntax and Manchester Syntax are recognised and rejected with a hint. Convert them to RDF/XML or Turtle first, for example with ROBOT:

robot convert --input onto.owx --output onto.owl

Recommended options

kgdiff release-1.obo release-2.obo --match id

• --match id: ontology ids (CURIEs, IRIs) are stable by design, so pair terms by id only. Obsolete terms still pair with their replacements through replaced_by (OBO) or IAO:0100001 (OWL, "term replaced by") and are reported as merges. The default, auto, may also pair a removed term with a new one by label similarity.
• --hierarchy <predicate>: the predicates that place a term under a parent, so that a changed parent reads "Moved … under …". The defaults are rdfs:subClassOf, rdfs:subPropertyOf, skos:broader, is_a, subClassOf and parent. Giving --hierarchy replaces them; repeat it for several, e.g. --hierarchy is_a --hierarchy part_of to read partonomy as hierarchy too.
• --lang <tags> (RDF): which language's label to show when a term has several, e.g. --lang en or --lang zh,en.
• --ignore-attr <key>: leave out annotations that change in every release (repeatable). Keys are OBO tag names such as creation_date, or compact predicates such as dcterms:date in RDF; the detail panel of the report shows them.

What kgdiff reports

You edit	kgdiff reports
Obsolete a term with replaced_by or IAO:0100001	Merged “old term” into “new term” (identity)
Obsolete a term without a replacement	Deprecated “term” (content)
Change a term's parent (is_a, rdfs:subClassOf)	Moved “term” under “new parent”, from “old parent”
Change a label	Renamed “old label” → “new label” (content)
Edit synonyms, definitions, cross-references, comments	Edited “term” or Updated attributes of “term” (content)
Replace a term with several new ones	Split “term” into “a” and “b” (identity), with a confidence
Add or remove an existential restriction (part_of some X)	a relation added, removed or retargeted (structure)

The ontology-release example shows most of them:

  Restructured 7
  ⊕ Merged “juvenile diabetes” and “insulin-dependent d…   identity · via replacement h…
  ⊕ Merged “heart attack” into “myocardial infarction”     identity · via replacement h…
  ⊕ Merged “chronic obstructive airway disease” into “c…   identity · via replacement h…
  ⇄ Moved “Crohn disease” under “inflammatory bowel dis…   structure · from “intestinal…
  ⇄ Moved “Hashimoto thyroiditis” under “autoimmune thy…   structure · from “hypothyroi…
  ⇄ Moved “tuberculosis” under “bacterial infectious di…   structure · from “lung disea…
  ⑂ Split “diabetes insipidus” into “arginine vasopress…   identity · inferred · 80% co…

The merges come from replaced_by: the obsolete terms are kept as deprecated and their relations carried over. The split has no replaced_by (the obsolete term only says consider): kgdiff infers it because the new terms are named after the old term's synonyms and took over its relations, and it reports how confident it is. --no-merges and --no-splits turn the detection off.

Limits

kgdiff diffs the graph an ontology describes: terms, labels, synonyms, annotations, the hierarchy and existential restrictions (SubClassOf R some C becomes an edge from the term to C). It does not run a reasoner, and other class expressions (unions, universal restrictions, equivalent-class definitions) appear as anonymous nodes rather than as readable axioms. To review logical axioms, use an axiom-level tool such as robot diff next to kgdiff.