Stream: genomics
Topic: more examples
Bob Milius (Sep 27 2019 at 18:48):
GF#16510 is about adding more examples. We created a sign-up for these, found here: https://docs.google.com/spreadsheets/d/1kYzfQMbTXwb4yt-tOKh9FfNcYjiK0W-PXwOJyoPH8PU/edit#gid=0
Looking at the spreadsheet it looks like we wanted
- clinical oncology tests - @Patrick Werner
- PGx and/or cancer risk assessment - @Bret H
- PGx examples - @Bob Dolin
- Somatic variant reporting - @Sameer Malhotra, @Alexander Mankovich
- more HLA - @ Bob Milius
I have more HLA examples. Have any of the others been done?
Bret H (Sep 27 2019 at 19:00):
I'm remiss. I will work on it this weekend. I just played with queries so I am pretty familiar with the IG. thanks for the reminder Bob!
Bob Dolin (Sep 27 2019 at 19:03):
@ Bob Milius I have examples that represent output from a VCF-to-FHIR translator, containing the region-studied, variant, and sequence phase relationships. This file ('HG00403.MEN1_variants.fhir.json.txt') contains all variants identified in the MEN1 gene from 1000 Genomes patient HG00403. HG00403.MEN1_variants.fhir.json.txt
Jamie Jones (Sep 27 2019 at 19:10):
@Bob Dolin I had uploaded one of the CYP2C19 files from before with some updates, will QA and upload this one too. Thanks!
Bob Dolin (Oct 03 2019 at 18:30):
@James Jones Jamie, do you know if we have any examples that show the inclusion of the ClinVar variation ID and allele ID?
Jamie Jones (Oct 03 2019 at 18:37):
No, that would be excellent if you could provide one
Bob Dolin (Oct 03 2019 at 18:55):
ha ha - I would love to, once I figure out how... It looks like we have a terminology system defined for the ClinVar variation ID, but not for the allele ID. to keep it simple, I'll just use variation ID for now, and put together an example and let you know when it's done.
Kevin Power (Oct 03 2019 at 18:58):
If the part you don't know is how to commit it to GitHub, just post the example file here and someone can take care of that for you.
Bob Dolin (Oct 03 2019 at 19:03):
@Kevin Power ClinVar has both a variant ID, and an allele ID. On the FHIR terminology systems page (https://www.hl7.org/fhir/terminologies-systems.html), we have http://www.ncbi.nlm.nih.gov/clinvar/, which scopes the ClinVar Variant ID. I don't see how we'd send the ClinVar Allele ID.
Bob Milius (Oct 03 2019 at 19:10):
@Bob Dolin , you raise a good point. From clinvar's documentation,
Allele ID
A unique integer identifier, the Allele ID, is assigned to each individual variant in ClinVar. The numbering systems for the Allele ID and the Variation ID described above overlap, so it is important to note the context of any integer identifier.
Our documentation says to use variant ID, but we don't provide anything for allele ID. I think I remember that @Larry Babb suggested that we only use variant ID and discouraged the use of allele ID, but I may be remembering wrong.
Kevin Power (Oct 03 2019 at 19:10):
We will have the table that @Patrick Werner and @Liz Amos put together (https://docs.google.com/spreadsheets/d/1XeXVPZ2DLf69Uohhzl0b-XQOTF3-91CQ0l2Hvn_Uvi4/edit#gid=1667123875 ) in the IG, but it doesn't list Allele ID either. However, I am not sure we are going to use Allele ID? I was thinking I heard that Allele ID would be going away longer term? But might be wrong on that. Perhaps @Larry Babb can answer.
I think the example with Variant ID is fine.
Kevin Power (Oct 03 2019 at 19:10):
Ha - Looks like @ Bob Milius beat me to the punch.
Bob Milius (Oct 03 2019 at 19:11):
jinx!
Bob Dolin (Oct 03 2019 at 19:11):
Ok, thanks. I'll use variant ID, and put together an example.
Larry Babb (Oct 03 2019 at 23:43):
Please don't use the clinvar allele ID. Every allele id in clinvar has a corresponding unique variant ID. And for the variants that are complex... they also have a variant ID but no allele ID. The allele id was an early approach and the variant id is the superset of all variation, whether it is a simple allele or a complex variant. The allele id identifier set will only complicate the sharing of variation within the ClinVar data set.
Please let me know if there are any questions or concerns.
Bob Dolin (Oct 04 2019 at 15:17):
@James Jones Hi Jamie, here is an example that includes the ClinVar variant ID. It's a fake patient that carries a diagnosis of hypercholesterolemia. We tested the patient for variants in LDLR, to see if their hypercholesterolemia might be familial, and a variant was found. We use the inherited disease pathogenicity profile to annotate the condition, and we use the variant profile to include the identified variant. The variant profile includes the ClinVar variant ID. AnnotationExample.fhir.json.txt
Jamie Jones (Oct 04 2019 at 15:50):
Looks good! I think we have a LOINC for genomic coordinate system now, I can update that and include it.
Thanks!
Bret H (Oct 04 2019 at 16:52):
@Bob Dolin are you missing a derived from Link in that connects the obs-inh-dis-path with a variant annotation?
Bret H (Oct 04 2019 at 17:02):
it is hard to see that it is required unless you look at the snapshot table (because we require on the base profile). We need guidance on every profile to look at both the snapshot and differential.
Jamie Jones (Oct 04 2019 at 17:03):
also, should we be using "associated phenotype" (81259-4) instead of "condition" 75323-6?
Bret H (Oct 04 2019 at 17:04):
yep
Bret H (Oct 04 2019 at 17:14):
I am not sure if this example got into the mix. It is the same compound heterozygote example that we had but using a sequence reference, position and change, instead of HGVS strings. complexVariant_nonHGVS.xml
Jamie Jones (Oct 04 2019 at 17:15):
I can validate and include, meant to get that one uploaded during the connectathon
Jamie Jones (Oct 04 2019 at 17:59):
@Bob Dolin Should the value for the inherited disease pathogenicity Observation be "Pathogenic", LA6668-3? we have it bound to https://r.details.loinc.org/AnswerList/LL4034-6.html
Bob Dolin (Oct 04 2019 at 22:29):
Hi,
My scenario is that the patient has 238076009 | Primary hypercholesterolemia on his problem list, prompting a reanalysis of prior genetic testing for various condition-gene interactions. I'm experimenting with using the FHIR Genomics DiagnosticReport to convey annotations on the problem list item.
The required derivedFrom link is tricky in that for some problem list items, we just want to say that no disease-gene interactions were found. I'm the last one in the world to recommend any changes to the IG before we publish ;-) but... can we consider relaxing derivedFrom to 0..*?
The rationale for using "condition" 75323-6 instead of "associated phenotype" (81259-4) is that I'm primarily trying to annotate an existing condition with possible disease-gene interactions. The patient's condition isn't necessarily the same as the phenotype associated with the variant.
Rationale for using an observation value of "present" rather than "pathogenic", is that I'm simply noting that a disease-gene interaction is present - the variant my be pathogenic or likely pathogenic, may affect the metabolism of a drug commonly used to treat the condition, may alter risk factors associated with condition, may suggest an alternative diagnosis, etc. Nuances of each identified interaction are conveyed elsewhere.
Bottom line for me is that we can change the example to use associated phenotype and pathogenic, but I'm wondering if we can soften cardinalizy of derivedFrom?
Jamie Jones (Oct 04 2019 at 22:31):
I personally think that whole implication space needs major tidying
Kevin Power (Oct 04 2019 at 22:54):
Can’t argue with the need for tidying.
I do feel uneasy with using Observation for “existing conditions” however. I will have to review what options we have at some point to provide a different suggestion though, but was thinking one of our extensions (SupportingInfo maybe?) could be used to reference details in record. But I see that this is a little different as you want to relate things together and say somethings out that relationship.
Jamie Jones (Oct 04 2019 at 23:41):
We're looking at making an example for high-risk allele and yeah, linking to the disease seems important...
Bret H (Oct 05 2019 at 01:21):
@Bob Dolin the inherited disease profile has codeable value where you can state Pathogenic | Likely pathogenic | Uncertain significance | Likely benign | Benign . But if this is a condition from the patient's problem list then you should use a reference to an instance of Condition Resource in my opinion
Bret H (Oct 05 2019 at 01:24):
Sounds like your stating an association. The patient has X variation. The patient has Y condition. Without the necessity of a connection between them. In your bundle or diagnostic report you can have both an instance of Condition Resource and instances of the genetic variation. The conclusion can carry some statement about association. What do you think? Will that let you provide the co-occurrence you're seeking
Jamie Jones (Oct 08 2019 at 17:02):
question on http://build.fhir.org/ig/HL7/genomics-reporting/SNVexample.html, should we be advocating sending implied p.HGVS without the g. or c. information?
Bret H (Oct 08 2019 at 17:04):
Where? I see component
code: Amino acid change (pHGVS) (Details : {LOINC code '48005-3' = 'Amino acid change (pHGVS)', given as 'Amino acid change (pHGVS)'})
value: p.(Ala412Val) (Details : {http://varnomen.hgvs.org code 'p.(Ala412Val)' = 'p.(Ala412Val)', given as 'p.(Ala412Val)'})
Jamie Jones (Oct 08 2019 at 17:04):
If it was specifically measured p.HGVS, my understanding is it would be 'p.Ala412Val' rather than the given 'p.(Ala412Val)'
Jamie Jones (Oct 08 2019 at 17:04):
i thought the parenthesis meant it was implied from a measured genomic change (from my brief time spent on
https://varnomen.hgvs.org/recommendations/general/)
Bret H (Oct 08 2019 at 17:08):
You are asking about properly formatted HGVS. The guidance from HGVS changed in 2012 ("to indicate that the description at protein level is without any experimental evidence it is recommended that, when RNA nor protein has been analysed, the description is given between brackets, like p.(Arg22Ser)") per https://www.hgvs.org/mutnomen/recs-prot.html. The example predated the change in the nomenclature. So, we're at a an interesting spot. Use the more current convention.
Bret H (Oct 08 2019 at 17:16):
note for any would-be implementer: HGVS syntax is not governed by HL7. The fields exist in the HL7 CG IG as per current practice. One should consult HGVS to understand how to build an HGVS parser. (but... in general, there are older conventions and systems which will not adhere to the parenthesis convention. Unless the test explicitly measures the correct target molecule - do not make assumptions about directly observed protein change.).
FYI: Additionally, there can be implications about how the change occurred (insertion and deletion versus substitution) which HGVS can be used to convey. In terms of patient care, the history of how the change occurred is not typically relevant.
Jamie Jones (Oct 08 2019 at 18:15):
up to 18 examples, will try to upload a couple somatic predictive/prognostic
Last updated: Apr 12 2022 at 19:14 UTC
