How to Validate an OBO ontology using Obo Metadata Ontology schema

Steps

Step 1: Obtain the sqlite version of the ontology

Currently the sqlite version of ontologies are not distributed alongside them in OBO

You can either

• (a) make the sqlite file yourself, see INCATools/semantic-sql

• (b) get a ready-made download from the s3 bucket

The second option is likely easiest.

For example:

wget https://s3.amazonaws.com/bbop-sqlite/uberon.db

Step 2: Install oaklib

pip install oaklib

Check your install works:

runoak --help

Step 3: Validate an individual ontology

runoak -i sqlite:uberon.db validate

This will stream yaml output. The output is linkml objects using the SHACL Validation vocabulary

type: https://w3id.org/shacl/MinCountConstraintComponent
severity: ERROR
subject: CARO:0000003
predicate: rdfs:label
info: Missing slot (label) for CARO:0000003

type: https://w3id.org/shacl/MinCountConstraintComponent
severity: ERROR
subject: CARO:0000006
predicate: rdfs:label
info: Missing slot (label) for CARO:0000006

Step 3 (alternative): Validate multiple ontologies

runoak validate-multiple db/*.db -o obo-validation.tsv

Caveats

Currently only the following are implemented:

• MinCountConstraintComponent checks (required or recommended)

• MaxCountConstraintComponent checks

• DeprecatedPropertyComponent

• DatatypeConstraintComponent: basic type checks (literal vs object) DOES NOT YET CHECK SPECIFIC LITERAL TYPE

• ClosedConstraintComponent

Using your own schema

TODO: add an option to pass in your own yaml file

How this works

The Python API is described here:

• ValidatorInterface

Currently there is only one implementation, the SqlDatabase implementation

The validation is driven entirely by a LinkML schema

Currently this schema lives within this repo, but the goal is to have it live outside and be imported

• linkml generated docs

• source: src/oaklib/datamodels/ontology_metadata.yaml

Different implementations are free to use this in different ways

The SqlDatabase implementation attempts to do this in a performant way doing whole-database predicate-based queries

Validation results use the Validator Datamodel, which reuses many URIs from SHACL

Analysis

See notebooks folder in https://github.com/cmungall/obo-metadata