Stream: genomics
Topic: FISH result related data components
SangHo Kim (Feb 21 2022 at 10:02):
Hello,
Below are list of "data" that the lab would like to store in FHIR.
I am well aware that there isn't an official IG for Cytogenomics but would still like to ask for learning purposes!
- Gain/Loss -> I believe LOINC 62378-5 (Chrom - copy-number-change-type) can be used
- Translocation -> https://loinc.org/LL4033-8/ can be used as custom component if non existent?
- Rearrangement -> Couldn't find any code for answer list in LOINC any suggestions?
- Amplification -> i think this is has a component/extension (copy number event)
- 3
or 5GAIN/LOSS - Partial Del / Partial Rearrangement
Thanks!
Kevin Power (Feb 21 2022 at 16:41):
@SangHo Kim - Do you happen to have an example report (or several) you could share? I ask because the list you provided above seems to be mostly 'answers' rather than 'questions' -- as you have no doubt discovered, there is a great deal of nuance in this space where the exact questions/answers tend to blend together and, in many cases, overlap. Often times, if we can see some instances of a report, it can help clarify.
SangHo Kim (Feb 22 2022 at 01:12):
Thanks for the reply @Kevin Power
Below are images of mentioned 'data". As you have mentioned, I think answers overlap but I'm not completely sure. For example, Hemato-Oncology related results have "translocation" but I was told that "rearrangement" is more correct since it could mean both translocation and "something (I fogot... Inversion? or Fusion)". My goal is to find components that I can just add to let say "FISH observation profile (custom)". This would be, at least for now, be specific for the lab I am working with.
-
Hemato-Oncology (images in order : Dual Color & Dual Fusion, Single Color, Dual Color, Break Apart)
image.png
image.png
image.png
image.png
image.png -
Tissue (Images in order : Break Apart, Dual Color (2 images) )
image.png
image.png < - DC & Example of Amplification (but with ASCO, HER2)
image.png <- Another Example of HER2 -
Microdeletion
image.png
I tried to write in some FHIR related info in the images as well (used the mouse to write... sorry :p )
Thanks!
SangHo Kim (Feb 22 2022 at 01:13):
Another Microdeletion Lab Result as an example (Williams Syndrome)
image.png Dual color probe Williams syndrome ish 7q11.23(ELNx2)
ish del(7)(q11.23q11.23)(ELN-)
SangHo Kim (Feb 22 2022 at 01:15):
Other samples for Hemato-Oncology lab Result (KMT2A (MLL) )
image.png KMT2A(MLL) nuc ish(KMT2A(MLL)x2)[200]
nuc ish(KMT2A(MLL)x2)(5'MLL sep 3'MLLx1)[10/200]
nuc ish(KMT2A(MLL)x3)[10/200]
nuc ish(5'MLLx2,3'MLLx1)(5'MLL con 3'MLLx1)[10/200]
SangHo Kim (Feb 22 2022 at 01:17):
More "TISSUE" Sample (BAP, BCL2)
image.png BCL2 Negative for BCL2 (18q21.3) translocation by FISH
Positive for BCL2 (18q21.3) translocation by FISH
"Negative for BCL2 (18q21.3) translocation by FISH
Positive for BCL2 copy number gain/amplification"
"Positive for BCL2 (18q21.3) translocation by FISH
Positive for BCL2 copy number gain/amplification"
you have copy number gain/amplification
SangHo Kim (Feb 22 2022 at 01:25):
This is a little extra to the type of change AnswerList in LOINC but is there another LOINC Code that would hold below answers?
[ marker chromosome, ring chromosome, dicentric chromosome, isodicentric chromosome, derivative chromosome ]
This is the answerlist I was able to find in LOINC which I thought would be enough but later on found out we needed more...
EDIT
I read the DNA Change Type Stream and found the SO link.
I can't say I understand the site but I think this holds more answers?
So you would use SO as system, SO:000### as code with the value[x] as one of the answers?
SangHo Kim (Feb 22 2022 at 01:53):
Or would http://www.sequenceontology.org/browser/current_release/term/SO:0000240 this be more appropriate for Cytogenomics as answerlist for change type?
Kevin Power (Feb 22 2022 at 14:08):
This is a lot of great information @SangHo Kim , thanks for sharing! This will take some time to review, and I ask that @Bret H review, and will also tag @Bob Freimuth who has been thinking about how to model structural variants.
I am afraid any quick answers we can provide, except for the obvious of using the cytogenomic-nomenclature component to deliver the ISCN string, will be very short term and likely not the best answer long term. Which leads me to the question - what is your timeline? Do you have a commitment to deliver something by a certain date, or are you simply exploring options?
SangHo Kim (Feb 22 2022 at 16:03):
No problem. Great to hear that the information is of any help.
As per your question on the timeline, I am committed to this Friday 2/25/21 (Korea time, GMT+9).
As you mentioned @Kevin Power , short term answers will be great to due to the deadline.
The lab is well informed regarding the absence of cytogenetics part in the official IG and they know that things will change.
The goal is to come up with components that are able to express "selected data by the lab which they think is important" in FHIR for both FISH and Chromosome (results like image.png 46,XY,t(9;22)(q34;q11.2)[10]/47,idem,+8[5]/47,idem,+8,i(17)(q10)[5] <- BM related test)
The list that was asked inhttps://chat.fhir.org/#narrow/stream/179197-genomics/topic/FISH.20result.20related.20data.20components are for FISH though.
Regarding the Chromosome (including metaphase terms like idem, der, etc) are being discussed with @Bret H .
Anyway any quick implementation method would be of great help though!
Thanks!
Bob Freimuth (Feb 22 2022 at 20:03):
Thank you, @SangHo Kim ! I appreciate the input and examples. I'll reference them when we dive into cytogenetics... hopefully you'll be able to join some calls and help us figure out how to handle all of that complex data. :-)
Bret H (Feb 24 2022 at 15:14):
chromosomal band (18q21.3) goes into component:cytogenetic-location http://build.fhir.org/ig/HL7/genomics-reporting/StructureDefinition-variant-definitions.html#Observation.component:cytogenetic-location @SangHo Kim
this is distinct from the component:cytogenetic-nomenclature (the place for ISCN strings)
Bret H (Feb 24 2022 at 18:17):
@SangHo Kim Yes, for DNA change type, the parent term is SO:0002072, so you would use terms that are children of it as valueCodeableconcept values with SO as the system. We've examples in the IG Build using so terms such as
http://build.fhir.org/ig/HL7/genomics-reporting/Observation-ExampleGermlineCNV.json.html
SO hierarchy partial screen shot
image.png
Bret H (Feb 24 2022 at 18:22):
with the probes, unless specific LOINC codes exists for the test with that probe, it'll be hard to put the probe name in.
You can describe what was observed/inferred from the probe detection assay - that there was a deletion at a specific region of the chromosome. And describe the method somewhat, but the probe name and description is harder to place.
Bret H (Feb 24 2022 at 18:28):
you can describe the location the probe interrogates by defining the region.
SangHo Kim (Feb 25 2022 at 01:01):
Bret H said:
with the probes, unless specific LOINC codes exists for the test with that probe, it'll be hard to put the probe name in.
You can describe what was observed/inferred from the probe detection assay - that there was a deletion at a specific region of the chromosome. And describe the method somewhat, but the probe name and description is harder to place.
Thanks for the response @Bret H !
If that's the case, I think we would just have to store the probe info (ex. FISH Trisomy 21 -> probe LSI 21 , FISH with probes for Chromosome 21 (band region 21q22.13) ) as one full string in component.valueString part for every obx created for the test.
you would have 2 observation.variant resource created. One for D13S319/D13S25 and another for D13S1825.
I know the D... is not for component.gene-studied but end user and I have come to a conclusion to store them under gene-studied as there is no real element for locus-name.
For every observation created, under component.value-string, the FISH probe info (FISH with probes for Chromosome 13q14.3 (D13S319, D13S25) and D13S1825) will stored as full string (maybe concatenate Probe name as well -> Del(13)(q14.3) )
Also, observation variant would express Dna-change-type as Del for both observations with copy number of 1.
The actual agreement that we (end user and I) have made is to just hold the "common" data in Observation Interpretation Profile.
As shown in below image, have cell phase and cells analyzed. Maybe include the probe information -> FISH with probes for Chromosome 13q14.3 (D13S319, D13S25) and D13S1825) will stored as full string (maybe concatenate Probe name as well -> Del(13)(q14.3) ) under conclusion.string?
스크린샷-2022-02-25-오전-9.58.07.png
Bret H (Feb 25 2022 at 03:21):
Locus name - for that could you consider a delivering it as a genotype or haplotype profile (like in this HLA example: http://build.fhir.org/ig/HL7/genomics-reporting/Observation-genotype-hla-a-glstring-r4.html)? Or do you think the cytoband component would work? Also, do you mean component.value-string for gene-studied? Another option to consider is using component:variation-code instead of gene-studied. With a codeable concept there is a display field, this is textual display to accompany the code. So if the locus has a coded representation you could use that as the code and a display string. But to fully describe the locus I would suggest considering using some combination of references, chromosome, start and end components. OR codeableconcept.text by itself without a code
Nice work by the way! Are you suggesting two new components cell-phase and number-cells-analyzed? You could slice component (the CG IG leaves it open) and use the LOINC codes for component.code
Bret H (Feb 25 2022 at 03:22):
@Patrick Werner is our binding fixed to codeable concept on gene-studied and variation-id?
Bret H (Feb 25 2022 at 03:24):
@SangHo Kim (your first image wouldn't render on my machine : ( but the second one did and is awesome! I'd recommend everyone here check it out.
SangHo Kim (Feb 25 2022 at 04:34):
Bret H said:
SangHo Kim (your first image wouldn't render on my machine : ( but the second one did and is awesome! I'd recommend everyone here check it out.
The first image was this!
image.png
SangHo Kim (Feb 25 2022 at 04:50):
Bret H said:
Locus name - for that could you consider a delivering it as a genotype or haplotype profile (like in this HLA example: http://build.fhir.org/ig/HL7/genomics-reporting/Observation-genotype-hla-a-glstring-r4.html)? Or do you think the cytoband component would work? Also, do you mean component.value-string for gene-studied? Another option to consider is using component:variation-code instead of gene-studied. With a codeable concept there is a display field, this is textual display to accompany the code. So if the locus has a coded representation you could use that as the code and a display string. But to fully describe the locus I would suggest considering using some combination of references, chromosome, start and end components. OR codeableconcept.text by itself without a code
Nice work by the way! Are you suggesting two new components cell-phase and number-cells-analyzed? You could slice component (the CG IG leaves it open) and use the LOINC codes for component.code
For clarification, I am uploading an image of Variant Profile as named as CNV (just cuz it has copy number event on genes or locus as part of the result of FISH tests)
스크린샷-2022-02-25-오후-1.47.55.png
SangHo Kim (Feb 25 2022 at 04:59):
Aslo, answers to the questions asked.
- genotyp or haplotype profile can be considered but the end user wants to simplify on the number of information being separated. The more resources introduced means more "input texts" to create if we were to implement it in LIS.
- Or do you think the cytoband component would work? -> ANS: is this an existing component in Variant Profile? I don't think I've seen one from the site. Please if you could share the component link that would be helpful!
- Also, do you mean component.value-string for gene-studied? -> ANS: no, from what I've seen from the Variant profile page, there is conclusion-string AND gene-studied Component Elements. Gene-Studied would be used for gene names (TNF, MTHFR, and the likes) while conclusion-string would be used to share Probe Information (since you've mentioned that Probe Information can't be stored currently because there are no elements or LOINC codes to represent those information)
- Are you suggesting two new components cell-phase and number-cells-analyzed? -> From what I've experienced (as this is my first time to work on this type of projects) I think at least in Korea, Probe information or test information is a MUST (regulation). Thus, those two information would need to be present in the profile? for FISH AND for hemato-oncology related profiles? (again, I am not expert in this field but as far as I have been told by the lab, they said it is required and is a in the regulation)
Bret H (Feb 25 2022 at 14:04):
- Yes, it would mean multiple profiles. Likely you'll also want to use diagnostic report, even if you do not use haplotype and genotype, so you will be likely to send more than one profile anyway. Multiple profiles allow for compartmentalization. But for the locus name, you might try the variation id component, but it depends on what you mean by locus name. If you mean something like 3p22.1 then that would go into cytogenetic-location but I think you mean something else.
- I meant component:cytogenetic-location it's not the cytoband label that you need. It's for position use the cytogenomic banding nomenclature (e.g. 3p22.1). From your other example, I think you want a place for 'Banding technique' like C-banding and G-banding, when you say chromosome band. This would be part of the method.
- Can you give me an explicit example of what you mean by locus name?
- Cell-phase and number-cells-analyzed are an important subset for sure. Seems reasonable to me to have them in their for FISH. But some methodology information could go into the methodology sections, take a look at the EMERGE based Pharmacogenomics report on how they handled methodology sections. http://build.fhir.org/ig/HL7/genomics-reporting/pharmacogenomics.html Take a look at how they use Plan Definition https://www.hl7.org/fhir/plandefinition.html to see how they used it.
Banding type (e.g. G-banding), banded level (e.g. 450) could all be observations linked to a Diganostic report. But you showed me an example with them as components. I think that's a good move, you have the specific LOINC codes, but more importantly banded-level and banding-type are meaningless by themselves. The profiles in the CG IG are very much focused on results not attributes of the study. We have an active effort to allow for a better way to describe attributes of the study (e.g. how many cells observed).
Have you had a look at diagnostic report yet? http://build.fhir.org/ig/HL7/genomics-reporting/StructureDefinition-genomics-report.html There are fields there that you might find useful. The pattern is to use Diagnostic Report at the top level and then connect to results underneath it. Each OBX can be one observation result linked as a result to the Diagnostic Report. There is also a coded-annotation field you might find useful for some statements. Here's an example of a bundle one might send to show how things are tied back to Diagnostic Report (not cytogenetic specific): http://build.fhir.org/ig/HL7/genomics-reporting/Bundle-bundle-cgexample.html
Also, for the conclusion that the patient has a specific disease, the diagnostic implication profile would be suggested as a structured means of reproting: http://build.fhir.org/ig/HL7/genomics-reporting/StructureDefinition-diagnostic-implication.html
Bret H (Feb 25 2022 at 14:13):
In this report PDF example: https://diagnosticcytogenetics.com/wp-content/uploads/2017/01/Sample-Abnormal-PB-Cyto-Report.pdf
you see the Karyotype image. The karyotype might be attached with document reference, or take a look at Media in Genomic Diagnostic Report or genomic-report-note
Bret H (Feb 25 2022 at 17:38):
for relationships between resources see http://build.fhir.org/ig/HL7/genomics-reporting/general.html#relationships-between-resources
AND
http://build.fhir.org/ig/HL7/genomics-reporting/general.html#genomic-implications
Bret H (Mar 02 2022 at 23:19):
@SangHo Kim in case you need it. Take a look at this example to see how one could use genotype profile to provide a human friendly display name in valueCodeableConcept without having a code. http://build.fhir.org/ig/HL7/genomics-reporting/Observation-Genotype-Clinical-Trial-Example-using-haplotypes.json.html
Last updated: Apr 12 2022 at 19:14 UTC
