– DRAFT –
FHIR RDF

Introductions

Joe: At Johns Hopkins, working w Chris Mungall

Chris: Working at U Colorado

OWL usage

joe: Have some questions on how to represent things.
… Working on PR for ont access kit. We have a FHIR server and want to load content from a number of sources. Need to convert to FHIR, some data already in OWL/RDF. Not sure how to structure it.
… There's a github w existing OWL converter, but it's leaving out a lot of content.

https://github.com/aehrc/fhir-owl The Australian e-Health Research Centre

joe: Working on a PR: https://github.com/INCATools/ontology-access-kit/pull/320
… How to represent CURIEs? How to include classes from external code systems? Synonym structuring?
… Is-a hierarchy -- want to check it.

jim: Two issues in converting OWL. In our IRI stem proposal, we assume lots of code with one prefix. But a lot of OBO ont include terms from a number of namespaces.
… Also, our proposal assumes that the code is a local code and not a CURIE.
… RDF group has expectations about how that should work?

eric: Converting OWL to FHIR? joe: Yes.

joe: Converting OWL ont to FHIR code system resource. Suppose some content maps could be involved.

dbooth: Your intent is to use those with FHIR data later?

chris: Yes.

gaurav: Not sure how pyrophen is related, but it looks like it could be useful: https://pyrophen.readthedocs.io/en/latest/example-json.html It highlights the problem we have.

chris: It was written by the phenopackets guy.

https://pyrophen.readthedocs.io/en/latest/representation.html

jim: For a code like "HP:000300", if we prepend an IRI stem directly, then it won't match the OBO IRIs, that OBO assigns.

joe: i like those CURIEs in there but Idk if that's conformant?

joe: code: Indicates that the value is taken from a set of controlled strings defined elsewhere (see Using codes for further discussion). Technically, a code is restricted to a string which has at least one character and no leading or trailing whitespace, and where there is no whitespace other than single spaces in the contents xs:token JSON string

Regex: [^\s]+( [^\s]+)*

This datatype can be bound to a ValueSet

joe: i don't mind using a fhir extension element for the prefix/namespace, but whenever extensions come into play that is certainly an area where we'd want to have a consensus or explicit standard

joe: I'm thinking like: CodeSystem.extension.curie_map

joe: CodeSystem.concept.extension.uri maybe

eric: The code should just be the number part. You can still have valuesets that draw from multiple code systems.

gaurav: We could ask everybody to use either:
… { system: OBO, code: “HP:101” } OR { system: HPO, code: “101” }

houcemeddine: An excellent practice will be the use of OBO Ontologies instead of SNOMED-CT.

joe: I was thinking one OWL file should generate one FHIR resource.

jim: If you're making a FHIR resource for Mondo, maybe it should only have the Mondo terms.

gaurav: So then { system: MONDO, code: “UBERON:101” }?

eric: If there's no need to unify codes w other ontologies, then ... if you have hp: on the front of each, and take those off, but leave others off, you'd effectively be making aliases for them in your code system.
… But the cleanest way to do it is to emit one code system for each prefix.

rob: Agreed.

joe: Suppose I have one with only one term. Would that confuse?

eric: I don't think so. It would be more confusing if you kept the prefixes.

chris: Want to check with Peter.

jim: Our goal is that it comes out looking how people expect it to look.

joe: Tooling available for FHIR RDF?

jim: The playground.

eric: Also Harold's demo of using SNOMED and reasoning with FHIR RDF.

https://yosemiteproject.org/tutorial-fhir-rdf-as-a-bridge-to-the-semantic-web-in-healthcare/

gaurav: Also concept IRI https://github.com/fhircat/fhir-rdf-playground/pull/12

dbooth: Also HAPI server does FHIRRDF R4. Not yet R5.
… and the playground: https://fhircat.github.io/fhir-rdf-playground/

gaurav: Given that FHIR thinks of codes as system-code pair, and RDF uses IRIs. We're planning to use the HL7 terminology website to host the IRI stems.

https://jira.hl7.org/browse/UP-364

https://bit.ly/fhir-rdf-concept-iris-july-2022

eric: hapi-fhir RDF tests: https://github.com/hapifhir/hapi-fhir/blob/master/hapi-fhir-structures-r4/src/test/java/ca/uhn/fhir/parser/RDFParserTest.java

gaurav: We've made the proposal to the terminology group, and want to start adding actual IRI stems.

dbooth: And the underscore should be a part of the IRI stem, because it isn't a part of the code.

joe: For sysnonyms, if it's exact, shoudl I use "designation"?

rob: Yes, that sounds right.
… For broader/narrower, it could be done with a property.

Other updates

gaurav: No progress on concept IRIs

jim: No word from grahame.

dbooth: Eric, HAPI progress?

eric: Need time to work on it.

dbooth: And simplifying the playground URI?

eric: James is doing that.
… Might want an OWL-friendly option. So it might be useful to keep those optoins somewhere.

ACTION: DBooth to check with James about trimming the playground UI

ADJOURNED