This post describes a recent improvement to the Semantic Farm alignment workflow with the Open Biological and Biomedical Ontology (OBO) Foundry enabled by contributing upstream data and code improvements to the OBO Foundry itself.

While the OBO Foundry Principle 3 (URI/Identifier Space) states

Each ontology MUST have a unique IRI in the form of an OBO Foundry persistent URL (PURL) that includes the ontology’s short namespace.

there are four ontologies for which this isn’t the case:

Prefix Name URI Prefix
CHEMINF Chemical Information Ontology http://semanticscience.org/resource/CHEMINF_
FMA Foundational Model of Anatomy Ontology http://purl.org/sig/ont/fma/
MAMO Mathematical Modeling Ontology http://identifiers.org/mamo/MAMO_
SWO Software Ontology http://www.ebi.ac.uk/swo/SWO_

Updating the OBO Foundry

Because the OBO Foundry previously assumed that all OBO Foundry ontologies followed principle four, it had a hard-coded rule for generating JSON-LD and SHACL contexts that incorrectly created incorrect URI prefixes like http://purl.obolibrary.org/obo/CHEMINF_ for the Chemical Information Ontology following the rule for well-behaved ontologies like the Gene Ontology (GO), whose URI prefix is http://purl.obolibrary.org/obo/GO_.

Therefore, I made a pull request to the OBO Foundry in https://github.com/OBOFoundry/OBOFoundry.github.io/pull/2951 that extended its data model to include a uri_prefix field and add the explicit override for each of the four ontologies above. It also updated the context-generating scripts to take into account these annotations.

Further, I added into the JSON-LD context a mapping from uri_prefix to the predicate from the vocabulary for annotating vocabulary descriptions (VANN) for describing the URI prefix for a given resource (vann:preferredNamespaceUri). This gets explicitly reflected in the Turtle version of the JSON-LD export of the registry. The specific usage of the new predicate can be seen here.

Updating VANN

During the process, I actually realized that there was a misconfiguration on the purl.org redirects for the VANN PURLs. Specifically, PURLs like http://purl.org/vocab/vann/usageNote were redirecting to https://vocab.org/vann/usageNote when they should have been redirecting to https://vocab.org/vann/#usageNote.

Luckily, I was able to reach out to VANN’s maintainer via its GitHub repository and suggest a fix on https://github.com/iand/vocabdotorg/issues/4. Ian was able to fix the configuration the same day, once again demonstrating that open science is awesome.

Updating the Semantic Farm / Bioregistry

The gap in the OBO Foundry’s curated data about its ontologies wasn’t new. I had actually already written custom overrides in the Semantic Farm / Bioregistry’s alignment script to override one (but not all four) of these during ingestion of the OBO Foundry.

After the OBO Foundry pull request was accepted, I was able to remove the custom overrides in a pull request to the Semantic Farm / Bioregistry (https://github.com/biopragmatics/bioregistry/pull/2130).

Like my previous post describing contributions I made to the World Avatar project, this post is yet another demonstration of the goodness of openness, shared technical infrastructure, and learning to love the drive-by curation.