FHIR Chat · Gene Panel membership · genomics / eMerge Pilot

Stream: genomics / eMerge Pilot

Topic: Gene Panel membership

Larry Babb (Mar 06 2019 at 12:58):

We will be making extensive use of the ObsGenePanel profile in our pilot as it seems to be the only mechanism to structure (group and aggregate) the complex array of results analyzed in the genetic assay associated with the eMERGE effort (molecular diagnosis based on indication, PGx panel with nested individual PGx results, and polygenic risk scores - coming soon).

The "hasMember" list of references seems quite complete except for ObsTrasnporter (the Medication Transporter Impact profile). I'm guessing this was an oversight. But it may be worth reviewing the list and making sure other allowable member types have not been overlooked as well.

Question 1: Can ObsTransporter be added to the hasMember list of ObsGenePanel explicitly?

Hypothetical Question 2: If we (the emerge pilot) needed to create an Observation Profile that did not currently exist would it be permissible to make it a member without breaking the validation or definition of the ObsGenePanel? (we would notify the CG IG team of any new Observation profiles of course, but I'm thinking about how this might be handled in as "new" observation profiles might be defined in a post-IG release world)

Kevin Power (Mar 06 2019 at 15:19):

Question 1: Can ObsTransporter be added to the hasMember list of ObsGenePanel explicitly?
That certainly sounds like an oversight, you can log a gForge tracker for that one. The intention of the Gene Panel was that it would act as a grouping mechanism for any of the findings and impact profiles.

Other entities are always welcome to create their own FHIR profiles. I think in this case, you would also have to profile GenePanel to indicate it could include your other finding or impact related profiles.

Bret H (Mar 06 2019 at 15:22):

@Larry Babb But why do you feel you need to be explicit that the ObsTransporter can be linked to? Not why you need to link it, but why do you think it needs to be explicitly stated?

Kevin Power (Mar 06 2019 at 15:25):

@Bret H My answer to that is that we show everything else, why would we not show this one profile?

Bret H (Mar 06 2019 at 15:25):

? what do you mean

Kevin Power (Mar 06 2019 at 15:29):

Genomics Panel says that the Panel profile can have a 'hasMember' referenced to all our finding and impact related profiles, except this one. I am quite sure it was just an oversight in the definition of the Panel profile.

Bret H (Mar 06 2019 at 15:33):

@Kevin Power Are there any restrictions stated? if the the term 'can' is used, then it is not restrictive. We might have other members in the future, and there are items which can be added. For example, if I did an enzymatic activity test, while it is not mentioned, I could still have hasMemeber to it. Not opposed to adding mention of the specific profile, but let's change the language to be clearer that 'while this list is extensive it is not restrictive or complete. Any profile meeting the base requirements can be added.' or something like 'the profiles mentioned here are exemplary and not restrictive'

Larry Babb (Mar 06 2019 at 15:35):

@Bret H that is the clarification I am looking for. I wasn't sure if what you are proposing here is true or not. It might be good to verify with the other in the CG group unless you are certain. Otherwise, why list any of the profiles, why not just cite the IG ObsXXX profiles list and say that these are examples.

Bret H (Mar 06 2019 at 15:44):

We sliced hasMember but did not remove the base hasMemeber. http://build.fhir.org/ig/HL7/genomics-reporting/obs-gen-panel-definitions.html#Observation.hasMember
So, you can still add a Reference to Observation or QuestionnaireResponse or MolecularSequence using hasMemeber.

Kevin Power (Mar 06 2019 at 15:45):

My impression of the Genomics Panel profile is to group the findings and impacts found in this IG - not to group any observation you want. So I would have said it was a mistake to not remove the base hasMember options. Perhaps that is too restrictive and unnecessary?

Bret H (Mar 06 2019 at 15:50):

As I recall, the specific profiles were called out to provide guidance...yeah, as Kevin said an oversight. Looking at the list the only two have a cardinality that provides guidance (i.e. card 0..* is pretty loose but 0..1 is strongly informative especially from a computational perspective). @Kevin Power suppose pulling in a enzymatic activity test would be outside the expected scope of Gentetic results. However, I recently asked a vendor to shoot me a list of all their genetic tests and they included enzymatic assays that imply a mutation in a specific gene. @Larry Babb would you want more than is listed with the addition of the one you mentioned? Partly the IG is informative for an implementer and partly the profiles in the IG are meant to be used to validate/create conformant profiles automatically without less additional effort.

Larry Babb (Mar 06 2019 at 15:59):

Interesting. I did not know that the IG was meant "partly" as informative and that it could be used to create conformant profiles not included in the IG itself. I like the idea of knowing that I can be conformant with the IG even if I extend it to create my own profiles (assuming they meet the IGs definition of conformant.)

How do we determine if the profiles I may or may not create are conformant? Is there a validation process or section in the IG spec that stipulates how to meet this criteria?

Bret H (Mar 06 2019 at 16:01):

Validation. Yes there are tools, our current IG needs work. We've an ongoing effort to fix it @Patrick Werner comment?

Kevin Power (Mar 06 2019 at 16:02):

We need to create some better guidance around conformance testing / validation, but the basics can be found here: http://hl7.org/fhir/validation.html

Kevin Power (Mar 06 2019 at 16:03):

And I wouldn't say our IG breaks the tools, but our IG still has a few things that are not right and/or the validators don't handle.

Grahame Grieve (Mar 06 2019 at 22:00):

do we have a list somewhere? tasks?

Kevin Power (Mar 06 2019 at 22:07):

Most of the work is on us. We have our own tasks to clean up our examples, and make things easier to validate. When we chatted at San Antonio, we mentioned seeing this error quite a bit: Error null validating Coding http://terminology.hl7.org/CodeSystem/v2-0074#laboratory [null}, and you weren't sure why it was showing "null" - but honestly, not sure if we logged a tracker for it. @Patrick Werner might have, and maybe logged others?

Our whole qa file is here: http://build.fhir.org/ig/HL7/genomics-reporting/qa.html

Patrick Werner (Mar 27 2019 at 14:44):