Skip to content

Ontology Conversion

BOOMER can import OBO and OWL ontology files directly, converting structural axioms into hard facts and cross-references/SKOS mappings into probabilistic facts suitable for reasoning.

Quick Start

# Convert an OBO ontology to a boomer KB
pyboomer convert my_ontology.obo -o kb.yaml

# Convert an OWL ontology
pyboomer convert my_ontology.owl -o kb.yaml

# Solve directly from an ontology file
pyboomer solve my_ontology.obo -O markdown

How It Works

The ontology converter reads an ontology file and produces a KB with two kinds of facts:

Hard facts (probability 1.0) from structural axioms:

  • is_a / SubClassOf → ProperSubClassOf
  • equivalent_to / EquivalentClasses → EquivalentTo
  • disjoint_from / DisjointClasses → DisjointWith

Probabilistic facts from cross-references and mappings:

  • xref / oboInOwl:hasDbXref → EquivalentTo (default prob 0.7)
  • skos:exactMatch → EquivalentTo (default prob 0.9)
  • skos:closeMatch → EquivalentTo (default prob 0.7)
  • skos:broadMatch → ProperSubClassOf reversed (default prob 0.7)
  • skos:narrowMatch → ProperSubClassOf (default prob 0.7)

Disjoint groups are auto-generated per ID prefix, so entities from different namespaces (e.g. MONDO: vs ORDO:) are placed in disjoint groups.

Example: OBO Ontology

Given an OBO file with cross-references and SKOS mappings:

format-version: 1.4
ontology: disease-mappings

[Term]
id: MONDO:0001234
name: Alpha disease
is_a: MONDO:0000001 ! disease
xref: ORDO:123
property_value: skos:exactMatch OMIM:456789

[Term]
id: MONDO:0005678
name: Beta disease
is_a: MONDO:0000001 ! disease
xref: ORDO:456
property_value: skos:broadMatch ICD10:K72

Convert and solve:

# Convert to KB
pyboomer convert disease-mappings.obo -o disease.yaml

# Solve
pyboomer solve disease.yaml -O markdown

The converter produces:

  • Hard facts: MONDO:0001234 ProperSubClassOf MONDO:0000001, etc.
  • Pfacts: MONDO:0001234 EquivalentTo ORDO:123 at 0.7, MONDO:0001234 EquivalentTo OMIM:456789 at 0.9, MONDO:0005678 ProperSubClassOf ICD10:K72 at 0.7 (broadMatch is reversed)
  • Disjoint groups: one group per prefix (MONDO, ORDO, OMIM, ICD10)

Configuration

Python API

from boomer.ontology_converter import OntologyConverterConfig, obo_to_kb

config = OntologyConverterConfig(
    # Xref probabilities
    xref_default_probability=0.5,
    xref_prefix_probabilities={"OMIM": 0.9, "ICD10": 0.6},

    # SKOS mapping probabilities
    skos_exact_match_prob=0.95,
    skos_close_match_prob=0.7,
    skos_broad_match_prob=0.7,
    skos_narrow_match_prob=0.7,

    # Filtering
    skip_obsolete=True,       # skip obsolete terms (default)
    include_xrefs=True,       # include xrefs as pfacts (default)
    include_skos=True,        # include SKOS mappings as pfacts (default)
    auto_disjoint_groups=True, # generate MemberOfDisjointGroup per prefix (default)
    min_probability=0.01,     # filter out very low probability pfacts
)
kb = obo_to_kb("my_ontology.obo", config=config)

YAML Config File

# ontology_config.yaml
xref_default_probability: 0.5
xref_prefix_probabilities:
  OMIM: 0.9
  ICD10: 0.6
  ORDO: 0.8
skos_exact_match_prob: 0.95
skip_obsolete: true
auto_disjoint_groups: true

from boomer.ontology_converter import load_ontology_config, obo_to_kb

config = load_ontology_config("ontology_config.yaml")
kb = obo_to_kb("my_ontology.obo", config=config)

OWL Support

The OWL backend uses py-horned-owl and supports multiple serializations:

  • OWL Functional Syntax (.ofn)
  • OWL/XML (.owx)
  • RDF/OWL (.owl)
# Any OWL serialization works
pyboomer convert ontology.ofn -o kb.yaml
pyboomer convert ontology.owx -o kb.yaml
pyboomer convert ontology.owl -o kb.yaml

The same axiom types are extracted as for OBO. IRI-to-CURIE conversion uses the ontology's prefix declarations and handles OBO-style IRIs (e.g. http://purl.obolibrary.org/obo/GO_0008150 → GO:0008150) automatically.

Combining with Other Sources

You can merge ontology-derived KBs with SSSOM mappings or other KBs:

# Merge ontology structure with SSSOM mappings
pyboomer merge ontology.obo mappings.sssom.tsv -o combined.yaml

# Solve the merged KB
pyboomer solve combined.yaml -O markdown

Seed-Based Extraction

For large ontologies, extract a neighborhood around entities of interest:

# Extract cluster around a specific entity
pyboomer extract ontology.obo --id MONDO:0001234 -o cluster.yaml

# Multiple seeds
pyboomer extract ontology.obo --id MONDO:0001234 --id ORDO:123 -o cluster.yaml

# Limit hop distance
pyboomer extract ontology.obo --id MONDO:0001234 --max-hops 2 -o cluster.yaml