19 Mar 2019


David_Booth, EricP, Emily_Pfaff, Harold_Solbrig, Samson_Tu, Paweł_Szmeja, Gerard_Wed
David Booth



Pawel: Polish Academy of Sciences. Was doing an interop IoT project. Interested in semantic interop. Worked on ITSM for interop. INterested in comparing that with ShEx and ShExMap.

Gerard Wed: Working in Berlin, orig from Australia. Project w German gov. Part is video conferencing.

scribe: Not FHIR experts. Challenge working with the schemas. Currently our stack is GRPC protocol buffers, exposed via REST. Heavily looking now at pub/sub.
... German gov is very supportive of FHIR. Aware of waste caused by data silos.f
... Currently not using FHIR/RDF, but using FHIR/JSON, using the biowiki JSON library.
... Bypassed FHIR/RDF.

Emily Pfaff: dir informatics group at Univ of North Carolina. Bioinformatics. Needed to represent clinical data in a more semantically interoperable manner.

scribe: Very focused on relational DBs. No concept of interop yet, but moving there. But needing to share with collaborators, interop is getting more attention now.
... We pass FHIR messages, but if we could put it into RDF that might be even better, cross query data sources. Working now on a project of post-surgical patients on opiods. Building a risk model for chronic opioid use.
... Opioid crisis is more than just about health record. Lifestyle and other factors influence it also. Want to link to other data sources.

harold: We're also involved in the translator project. At Johns-Hopkins. We're turning stuff into knowledge graphs: reified triples with attribution and all that good stuff.
... Looking at FHIR. Modeling language is not easiest to use, but it really does model-driven architecture.
... Working with Evidence Based Medicine on FHIR group. Goal is to map EBM on FHIR to translator knowledge graph.
... I see the FHIR/RDF value is that you can get more or less than a resource if you need it. It also gives URIs to things.

Eric: W3C, longtime Sem Web geek. Devleoped a lot of W3C standards for healthcare and life science.

Samson: Following W3C for decades. Developed systems interested in updating to FHIR. Also interested in modeling, system to access knowledge in a more integrated way.
... Stanford University, with Mark Musen's group.

Draft ShEx - IPSM comparison article (Pawel)

pawel: Did new diagrams last week. Pub target deadline April 3 submission deadline for abstract for ISWC.
... Want contributions from others.
... Still drafting conclusions, but not yet in the repo.


scribe: The paper shows a case study of two examples, both in ShExMap and IPSM.
... The project we did was called Inter-IoT. One pilot was a medical pilot. What Emily said caught my attention, because that was one of the problems that we had to tackle. Can put you in touch with our Spanish partner, eHealth pilots.

FHIR/RDF projects (Gerard Wed, Cory Slep; Emily Pfaff, Robert Bradford)

gerard: Trying to build an enterprise stack with FHIR. Looking at pub/sub.

harold: Someone at the last HL7 meeting took the ball at getting the HAPI server to implement RDF.

gerard: Discussed with Cory the similarity between what GraphQL and JSON-LD are trying to achieve.

harold: Need to think about: tried to use JSON-LD using @context [a couple years ago], and it was possible with embedded @context scattered through the RDF -- flattened.
... But it made extremely ugly JSON. But JSON-LD 1.1 could now accomplish it. But the issue is that it would not come out exactly the same as the RDF we have now.

Scoped @context in JSON-LD 1.1 (David)



<ericP> JSON-LD playground example of FHIR JSON with JSON-LD 1.1 @context

<ericP> ericP: JSON-LD 1.0 is context-free; 1.1 is context-sensitive

david: Could use JSON-LD 1.1 to map regular FHIR/JSON to RDF.

harold: There are some places where we add triples, so JSON-LD 1.1 would not be able to add those.
... There may also be a few other spots that may not be consistent. But worth an experiment.
... Could use fhir.ttl to figure out where we would need the scoped @context.
... Would be worth an experiment. If the only differences are the extra triples that we add, it would simplify things. Would eliminate the need to implement RDF in FHIR servers.

eric: If we were super lucky, we could even do JSON-LD framing to recognize FHIR/RDF back into JSON.
... JSON-LD is (normally) one-way from JSON to RDF. But some stuff can go back in the other direction, from RDF to JSON.
... But maybe this would get us most of the way back to JSON using JSON-LD framing.

harold: Would be nice to get a grad student or someone to try JSON-LD 1.1 for this, and see if there are any other FHIR issues needing to be resolved. They're going for a lock-in in a few weeks.

gerard: Cory and I might be able to help.

<inserted> Previous investigation: http://dbooth.org/2015/fhir/json-ld/fhir-in-json-ld.pdf

harold: We can also provide fhir.ttl . We also have shex to describe it. And we have a python program to convert the FHIR to RDF.

<scribe> ACTION: David to follow up with Gerard on FHIR/JSON-LD 1.1 possibility

FHIR/RDF projects (Gerard Wed, Cory Slep; Emily Pfaff, Robert Bradford)

emily: What aspects of what we are doing are of interest to this community?
... Cannot share data. (protected information).

harold: Where we're bridging between FHIR and the knowledge graph.
... Our community is bridging between GraphQL and bio RDF.
... Interested in transformations from FHIR to biolink knowledge graph format.
... The biolink knowledge graph format is a custom format -- reified triples, to include provenance -- and classification of the items in the triples. Model of gene to chem association.
... Using that in robocop reasoner to connect RDF identifiers and traverses disease to system to chemical to env factors, etc.

emily: Interested in enabling robocop, but big challenge is gap between native EHR data, where most diagnoses are not labeled with SNOMED codes, but have ICD9 or 10 codes. What is the best way to get from ICD10 to a better ont that can be used?
... Chris Mungall's answer was that there is no great way to do that now.
... So we're a little stuck, but linking to other things is a challenge.

harold: We face that also. LOINC is a little easier. This could be a long discussion.
... Chris Chute has taken ont in bio info space.

<ericP> https://tinyurl.com/y3fm733s



<scribe> ACTION: David to follow up with Emily and Harold about Yosemite Project webinar on FHIR/RDF projects

<scribe> ACTION: David to connect Emily with Pawel for followup with Spanish collaborators

Summary of Action Items

[NEW] ACTION: David to connect Emily with Pawel for followup with Spanish collaborators
[NEW] ACTION: David to follow up with Emily and Harold about Yosemite Project webinar on FHIR/RDF projects
[NEW] ACTION: David to follow up with Gerard on FHIR/JSON-LD 1.1 possibility

Summary of Resolutions

[End of minutes]

Minutes manually created (not a transcript), formatted by David Booth's scribe.perl version 1.154 (CVS log)
$Date: 2019/03/19 21:47:02 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.154  of Date: 2018/09/25 16:35:56  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00)

Succeeded: i/Pawel: Polish Academ/Topic: Introductions
Succeeded: i| We can also provide fhir.ttl|Previous investigation: http://dbooth.org/2015/fhir/json-ld/fhir-in-json-ld.pdf
Succeeded: s|on FHIR/JSON-LD 1.1 possibility|with Gerard on FHIR/JSON-LD 1.1 possibility|
Present: David_Booth EricP Emily_Pfaff Harold_Solbrig Samson_Tu Paweł_Szmeja Gerard_Wed
No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth

WARNING: No date found!  Assuming today.  (Hint: Specify
the W3C IRC log URL, and the date will be determined from that.)
Or specify the date like this:
<dbooth> Date: 12 Sep 2002

People with action items: david

WARNING: IRC log location not specified!  (You can ignore this 
warning if you do not want the generated minutes to contain 
a link to the original IRC log.)

[End of scribe.perl diagnostic output]