Stream: genomics
Topic: functional class
Patrick Werner (Apr 25 2019 at 11:27):
In the context of Terminology work on the IG and the creation of GC oncology reports we identified the following issue:
Our lab always reports a functional class per Variant found (splice_site_variant, inframe_varient, etc.). This is done using the sequence ontology http://www.sequenceontology.org.
Right now i didn't find a place to put this information. So i was thinking about another component to capture the functional_class of the variant. Or maybe i missed another spot to capture this information?
Bret H (Apr 29 2019 at 15:53):
am I in the wrong profile: http://build.fhir.org/ig/HL7/genomics-reporting/obs-variant.html
I see "component SΣI 0..1 BackboneElement DNA change type" and " component SΣI 0..1 BackboneElement Amino acid change type" @Patrick Werner value sets are example. so you can you use SO as long as you provide code system with the code can you map into the value set?
Bret H (Apr 30 2019 at 15:42):
got it. so in off-zulip chatter, it is the absence of a place for SO terms for functional class. Previously in the Sequence resource and profiles of observation we had fields for SO. Maybe we can change the value set binding to be preferred. This would allow one to use SO, but also other value sets. @Patrick Werner thoughts?
Patrick Werner (May 03 2019 at 11:36):
My Question is: is a functional class the same as a DNA Change Type?
Patrick Werner (May 03 2019 at 11:37):
DNA Change Types to me are: substitution, deletion, ...
Functional Classes: FrameShift, missense, esseantial splice site, ....
Patrick Werner (May 03 2019 at 11:39):
oh, i'm stupid. Of course, this is captured as Amino Acid Change Type. I was really blind :-)
Patrick Werner (May 03 2019 at 11:44):
but the answer list is to short. Intron Variant / Splice Site Variant are missing
Kevin Power (May 03 2019 at 15:45):
So you are talking about additions to this answer list, right? https://r.details.loinc.org/AnswerList/LL380-7.html
Jamie Jones (May 03 2019 at 15:56):
I'm realizing with this that my text description of the 'Amino Acid Change Components' Group at the bottom of http://build.fhir.org/ig/HL7/genomics-reporting/obs-variant.html is naive/wrong and should be updated...
Bret H (May 21 2019 at 15:08):
we used to have SO. should we extend the values allowed by using the LOINC as preferred? can't recall why we restricted the value set to the few LOINC terms. I think it was due to concerns about labs currently using LOINC rather than SO terms.
Bret H (May 21 2019 at 16:10):
discussed on the call today was a new component. With DNA change type and AA change type as required an app developer (or recipient system) can rely on DNA change type and AA change type to have a specific set of values. But where will the information be found? in the new component or the DNA change or AA change fields. Maybe we need to consider DNA change type and AA change type to have cardinality 1..X. This would mean that where overlap exists one need only look in DNA change type or AA change type. Or we leave it open and acknowledge that implementers will need to be prepared to find/store data with either of the three fields. Think about the query required to return all annotations, and the query to try to discover all variants where there is a DNA change. For the DNA change query, one would need to look both in the DNA change element and the new,proposed component. Am I missing something? (it happnes) : ^ ) @Patrick Werner
Bob Dolin (May 21 2019 at 16:46):
@Bret H I digress, but I wonder why it is, that in cancer genomics, it seems prevalent to report at the AA change level, whereas for germline it's at the DNA level? Anyhow, I think DNA and AA change types are orthogonal to functional annotation. The former tells you just what the variant is. The latter helps you decide if the variant may be clinically relevant (e.g. because the DNA change occurs in an exon rather than in an intron).
Patrick Werner (Jun 13 2019 at 15:02):
(deleted)
Patrick Werner (Jun 13 2019 at 15:03):
(deleted)
Patrick Werner (Jun 13 2019 at 15:04):
@Bret H 0..1 cardinality for change type is ok, we can have multiple Condings in the CodeableConcept.
Patrick Werner (Jun 13 2019 at 15:06):
i agree that we need a new component for this: "functional annotation"
Patrick Werner (Jun 13 2019 at 15:26):
for the new component:
- Ask for a LOINC concept code (in the meantime define our own code)
- ValueSet binding to SequenceOntology (preferred)
Jamie Jones (Jun 14 2019 at 17:26):
@Patrick Werner Can't you not use the same coding multiple times though?
Patrick Werner (Jun 14 2019 at 17:40):
You can have as many codings per CodeableConcept as you like. As long they represent the same concept. As functional annotations aren‘t the same the AA change type we have to have a different component
Patrick Werner (Jul 02 2019 at 14:52):
after giving this more thought i think it would be good to have a component (0..*) named "functional annotation" with a preferred binding to sequence ontology. What do you think?
Patrick Werner (Jul 15 2019 at 16:02):
While working on functional annotation GF#22815 and variant type GF#22814 I browsed a lot through Sequence Ontology and now think we can merge these two into one component and call it Variant Type.
Functional annotation would use SO:0001060 - sequence_variant children concepts which also includes SNV, Indel and Structural Variant.
Kevin Power (Jul 15 2019 at 22:00):
I am fine with that, but we did vote on creating "Functional Annotation" last week, so we would probably need to re-vote on it tomorrow if we want to go with Variant Type?
Kevin Power (Jul 15 2019 at 22:37):
@ Bob Milius ^^^ topic for the Tuesday call ...
Patrick Werner (Jul 16 2019 at 11:54):
I am fine with that, but we did vote on creating "Functional Annotation" last week, so we would probably need to re-vote on it tomorrow if we want to go with Variant Type?
Yes unfortunately we have to revote.
Patrick Werner (Jul 16 2019 at 12:00):
i added the Topic 2: Variant Type to todays agenda @ Bob Milius
Bob Milius (Jul 16 2019 at 12:50):
thanks, @Patrick Werner , I added it to Topic 3: Functional Class as well. You're proposing to merge these two, correct?
Patrick Werner (Jul 16 2019 at 13:24):
yes.
Patrick Werner (Jul 17 2019 at 10:18):
after yesterdays discussion on the topic we agreed on first defining the meaning of each of these components:
- DNA change type
- Variant type
- functional annotation
Patrick Werner (Jul 17 2019 at 12:57):
Ncbi defines variant type as: Variant type is the type of any sequence change reported relative to a reference sequence. https://www.ncbi.nlm.nih.gov/variation/docs/glossary/
Patrick Werner (Jul 17 2019 at 12:58):
which i think is synonymous to DNA change type
Patrick Werner (Jul 17 2019 at 13:17):
functional annotation is not about a structural change (like the variant type) but about the (potential) functional change caused by the structural change.
Examples:
- synonymous_variant
- splice_site_variant
- stop_lost
Kevin Power (Jul 17 2019 at 14:32):
So, do we keep DNA change type for now, and still add something called Functional Annotation and use SO items in the value set?
Patrick Werner (Jul 17 2019 at 15:49):
So, do we keep DNA change type for now, and still add something called Functional Annotation and use SO items in the value set?
Yes. My merge intent was motivated through the possibility of using SO for both elements. This motivation was not correct, like @Bob Dolin said: "first we have to be clear about the meanings and then care about answer lists"
Patrick Werner (Jul 17 2019 at 15:50):
For DNA Change Type i still would prefer SO instead of the LOINC list as SO has relations and hierarchies.
DNA Change could be everything under sequence_alteration : http://www.sequenceontology.org/browser/current_release/term/SO:0001059
Patrick Werner (Jul 17 2019 at 15:53):
through using an ontology you can transport the alteration on different levels: http://www.sequenceontology.org/browser/current_release/term/SO:1000002
Substition or SNV/MNV
Patrick Werner (Jul 17 2019 at 15:54):
SO also contains definitions for their terms (which is missing in the LOINC answer list) and DB Xrefs for some terms
Patrick Werner (Jul 17 2019 at 15:56):
For DNA Change Type i still would prefer SO instead of the LOINC list as SO has relations and hierarchies.
DNA Change could be everything under sequence_alteration : http://www.sequenceontology.org/browser/current_release/term/SO:0001059
this is practically what ncbi and ebi defined here (their list is missing the abstract concepts like structural variation which i would include) : https://www.ncbi.nlm.nih.gov/variation/docs/glossary/
Kevin Power (Jul 17 2019 at 17:18):
I realized we didn't define a binding to for our "DNA Change Type" component to the LOINC answer list (https://r.details.loinc.org/AnswerList/LL379-9.html). That is probably an oversight that we missed since we include a "short description" which shows some of the answers in that list, but not all of them. So as part of this fix, perhaps we should include an example binding to the LOINC answer list, but realize that would still allow someone to use SO codes if they choose? We could instead setup our own specific ValueSet with the SO codes, which I think was mentioned on the call. I guess I am thinking the example binding to the LOINC answer list is easier for now and still allows others to send SO codes if they choose.
Patrick Werner (Jul 17 2019 at 19:16):
(deleted)
Patrick Werner (Jul 17 2019 at 19:17):
(deleted)
Patrick Werner (Jul 18 2019 at 15:46):
(deleted)
Patrick Werner (Jul 18 2019 at 15:47):
(deleted)
Last updated: Apr 12 2022 at 19:14 UTC
