Closed karafecho closed 1 year ago
Hello, Does anyone have the curies for the above phenotypes handy? Just trying to avoid doing extra legwork that someone else might have done. We are looking at doing the following process:
treat
those genesComments welcome
Here are some of the phenotype CURIES:
Great suggestion @marcdubybroad, and thanks for getting the process started @khanspers. I changed the phenotype lists in Kara's original post above to be tables with a column with IDs, and added in the Kristina's mappings. I think most of us have the ability to edit each other's comments so feel free to add to the tables above directly. Or message me and I'll add them...
Additional terms (curies / IDs) I started from:
Want to review with SMEs to see if we want to add them to the cases...
EDIT: added whether SME approved or didn't approve them
Gathering information. I'm using ARS-CI.
only 4 / 5 had links out to other Things. The "results"/Things were ATP1A3 (of course), and a bunch of diseases / phenotypes.
Simpler query gets the full list of diseases linked to the gene: (Gene ATP1A3) → Disease
Query Topology:
Diseases to pick to look for treatments:
Name | ID | Cases |
---|---|---|
dystonia 12, rapid-onset dystonia-parkinsonism (RDP) | MONDO:0007496 | 1, 2, 3, 4, 5 |
alternating hemiplegia of childhood 2 (AHC) | MONDO:0013900 | 1, 2, 3, 4, 5 |
CAPOS syndrome | MONDO:0011038 | 1, 2, 3, 5 |
Developmental and epileptic encephalopathy 99 (DEE99) | OMIM:619606 | 2 |
However, DEE99 may be incorrect for Case 2 (doesn't mention seizures).
Aside: an interesting article (2020) on cardiac phenotype in ATP1A3-related syndromes
(Gene ATP1A3) → ChemicalEntity
S-Adenosylmethionine
Cystamine
UMLS:C4709333
(actually Adeno-associated Virus Vector
) Translator / UI-related feedback:
Ammonium chloride
and ALUMINUM CHLORIDE
from Automat-CTD, without more helpful info. But I can go to CTD and look up ATP1A3's chemical interactions. And then I can click on the references for the ammonium chloride (link, link) and aluminium chloride (link). And see that the "evidence" articles for these relationships aren't particularly related to this set of cases...Pasted from my post in Slack:
Some responses to things in the chat for the QotM first meeting:
EDIT: Also I did queries and found that the variants are directly connected to various "loss-of-function" diseases and lots of diseases / phenotypes in general (when looking at BTE's results). See the "What are the variants connected to" section of this post.
EDIT: it doesn't sound like the SMEs are super interested in "similar genes where mutations cause similar phenotypes" and treatments for those diseases. However, if we want to use this as a strategy or these come up in our analysis of results (as they did for me in this post), that's probably fine.
General issues (for Translator)
Variant IDs (and any cross-mapping)
- CLINVAR:1333833 or ClinVarVariant:1333833?
It looks like both are listed as valid prefixes in biolink, but only CLINVAR is in the context.jsonld, so we probably need to remove ClinVarVariant? Probably a good biolink issue.
Yes, I think that's right. (Note the comment in biolink that rs ids are kind of a poor match for variants - they are really more of a locus id (though everybody uses them like this))
- CAID:CA358802 or remove the "CA" part?
Keep "CA" b/c then the url resolves correctly. FWIW, the CAID is really meant to function like a node normalizer for variants, providing all the other names for that entity (in this case including variations across different versioning of the underlying sequences)
- Thoughts on curies for genomic deletions / CNVs?
For small deletions CAIDs should work; for bigger and for CNV IMO there is not a perfect approach.
Disease IDs:
- Is the biolink-model prefix ORPHANET or ORPHA:2131?
Looks like it should be ORPHANET, but I am confused as well about why we have both in different places. Probably worth a biolink issue.
- SRI Node Normalizer didn't recognize OMIM:619606 (DEVELOPMENTAL AND EPILEPTIC ENCEPHALOPATHY 99)
Looks like dev NN ( https://nodenormalization-sri.renci.org/ ) has fixed this.
Replying to https://github.com/NCATSTranslator/testing/issues/233#issuecomment-1271835560:
Working with the 4 identified in the finding a "disease" section of this post:
Responses: (4 Disease IDs) → ChemicalEntity Note: re-ran on ARS-CI, may need re-running due to errors for multiple tools?
BTE had only 2 results:
Injection of botulinum toxin
for focal dystonia and considerations for anesthetic management for a patient with rapid-onset dystonia-parkinsonism (RDP)
UMLS:C0002932
(label should be Anesthetics
)Acetazolamide
(aka Diamox
) for patients with CAPOS syndrome / episodic ataxia
Based on feedback from Dr. Friedman, I edited Kara's original post to: 1) add additional phenotypes and HPO terms for each of the 5 cases, and 2) clarify the presumed (non-)significance of the 17p12 deletion
Slightly different strategy for find chemical linked to diseases, starting with the 4 diseases identified by Colleen: tried separate queries for each disease, ChemicalEntity -> Disease. Ran in ARAX UI:
Anesthetics
. The associated paper describes anesthetic management of a dystonia patient during a procedure to treat the dystonia with botox injection, so thats how it was picked up. However, it seems not relevant as it is not the actual treatment. The actual treatment described in the paper, Injection of botulinum toxin
(UMLS:C1321035), could have maybe been found if Procedure or another node category was used instead of ChemicalEntity.Flunarazine
and topiramate
.Topiramate
.Notes:
Brief notes from the 10/14 call:
More brief notes:
@gaurav : To Colleen's second point, I think CAM-AOP KP may be able to contribute here, sort of similar to the queries we've been testing for the TCDC get_creative() workflow.
For example:
144 results for CHEBI:63840-(?)-UniProtKB:P08684
This has some interesting results, e.g. CHEBI:63840("5'-hydroxyomeprazole") biolink:participates_in GO:0006739 ("NADP metabolic process") biolink:caused_by NCBIGene:100861540
We have a new tool, called OARD (Open Annotation for Rare Diseases), which is similar to COHD in providing co-occurrence counts between pairs of concepts, except OARD focsues on phenotypes and rare diseases mined from notes. OARD's association calculation endpoints can take multiple phenotypes and calculate joint association scores against diseases. This hasn't been validated or TRAPI-fied yet, but we just wanted to give it a quick test on this QoTM since we have a list of phenotypes, which suits this function.
Here's a notebook with results against the Case 1 phenotypes: https://colab.research.google.com/drive/1HSRLqXOuPdnUaWkae9UwOZl8MVkCpvCS
From there, we did a couple of TRAPI queries taking the diseases returned by OARD, finding the ones related to ATP1A3, and the chemical entities that treat those diseases.
During the call today, @suihuang-ISB brought up again the issue of whether a genetic disease results in irreversible damage by the time of diagnosis (in which case clinicians would want to focus on ameliorative treatments that address phenotypes and symptoms) versus cases where the genetic variation results in ongoing and reversible changes (in which case mechanistic / curative treatments focusing on the mutated gene are still worth pursuing -- cystic fibrosis / CFTR would be an example here). (Sui, please correct me if I've incorrectly or incompletely summarized this issue.)
As far as we know, there is no curated resource that classifies genetic diseases into these two classes (or somewhere on a spectrum between these two classes). As one idea, we could look at levels of expression (or variation in levels of expression) through development as a proxy for how likely mutation in a gene is likely to have reversible/irreversible effects. In an effort to look at relevant developmental gene expression datasets, I see some data in the EBI's Expression Atlas. For example this link shows the expression of ATP1A3 through developmental time (columns) in several tissues (rows):
I like @andrewsu 's idea above. A complementary approach might be to look at Go annotations. For instance ATP1A3 is annotated as being involved in GO:0021987 (cerebral cortex development). Which is a subclass of GO:0032502 (developmental process)
Notes on approaches:
This is a BTE-centric view
A few years back, the PMI team did a literature review for ATP1A3 and found this clinical report where a child with a de novo ATP1A3 variant was treated with oral ATP and showed significantly decreased frequency and shorter duration of hemiplegic episodes. Has anyone seen ATP come up as a potential treatment option?
This suggestion (@amfoksinska) , and comparing it to the above approaches (e.g. @colleenXu), really once more exposes the need to distinguish between two fundamental, natural classes of uses cases from the perspective of the (clinical) researcher:
(A) finding an existing (empirical) report, a cohort study or case report, deposited somewhere that explicitly proposes/evaluates a treatment for a defect in ATP1A3.
A few years back, the PMI team did a literature review for ATP1A3 and found this clinical report (...) Has anyone seen ATP come up as a potential treatment option?
(B) Try to come up with a new hypothesis for a treatment based on mechanistic reasoning
Not sure how to get from ATP1A3 -> related genes -> ChemicalEntity, Procedure, Treatment
Not sure if this natural fundamental distinction maps unto our current distinction of CREATIVE MODE (CM) versus non-CM (however we draw the line). Can someone answer tis questions??
If so, then we have essentially "rediscovered" a natural mode of approach that researchers use anyway. And if so, we should not use an artificial (and obscure) neologism to refer to it, and treat a fundamentally different class of an operation as merely a "variant" and give it an obscure name. It would be very useful if we are all aware of these two distinct processes, which also have distinct consequences in optimizing our product - it also affects the organization of EPC. A researcher in the "search/lookup mode" (A) is in a totally different mindset than when she is in a "explorative/reasoning mode" (B). Although we do it all the time naturally, assembling a set of known facts of relationships to a new relationship is a tricky, deep epistemological problem that is not fully understood in medicine, and we cannot expect to solve it.
I know, I know , there are many grey zones, but it does not help to muddle natural distinctions. To appreciate 'gray', on needs to first internalize the existence of 'black' and 'white'.
Per @amfoksinska's comment, I have not seen ATP in any of the results I've reviewed. Has anyone else?
@karafecho : The problem with ATP is that it is a mega-hub in our network connected to 1000s of pathways. So we have to mask it when doing graph traversal based queries. I think this is now a case for old fashioned literature/human reasoning based investigation. ATP is released as a neurotransmitter and has a plethora of (often INHIBITORY) post-synaptic effects. This may have been a rationale for suggesting ATP as therapy for the epilepsies. One could also ask whether the receptor in this neurotransmission, PY2 ... is somehow involved in ATP1A3 pathogenetic pathways. Or more boldly, one could ask if drinking Gatorade (blue) exacerbates the symptoms because the blue dye "Blue #1" in it antagonizes the neuronal function of ATP ...
@suihuang-ISB : Yeah, I completely understand the complexities related to ATP. I was just hoping that Translator, via get_creative() mode perhaps, would be able to identify (oral) ATP (or P2Y, as you suggest) as a potential treatment (or target), or at least surface the 2016 case report that @amfoksinska alerted us to (admittedly a challenging task).
I was pleased to see flunarazine in several of the query responses. It might be interesting to query for non-drug therapies (e.g., ketogenic diet, Gatorade?), too. Just a thought ...
@karafecho do we need to keep this ticket open? i ran it in the Test UI , for What drugs may treat:Atp1a3-associated Neurological Disorder
Thanks for checking, @sstemann. I think we can close all of the QotM tickets. The JCTS paper has been published (I returned v3 of the proofs this morning), and any ongoing work has been moved elsewhere, so I'll go ahead and close these tickets.
Submitting team
Exploring Agent, Service Provider (Andrew Su)
SME
Colleagues at the Rady Children's Institute for Genomic Medicine
Jennifer Friedman, MD Clinical Professor Neurosciences and Pediatrics UCSD/Rady Children's Hospital San Diego Clinical Investigator Rady Children's Institute for Genomic Medicine
Laura Forero, MD Genetics Fellow. UCSD/Rady Children’s Hospital
Charlotte Hobbs, MD Vice President for Research & Clinical Management Rady Children's Institute for Genomic Medicine
Challenge Question
Given a mutation in gene ATP1A3 and a case description of associated phenotypes, can Translator propose new therapies?
Background
For this challenge, Translator team members will consider five real-world use cases, all involving mutations in the gene ATP1A3. (Please note that the case summaries provided below were provided by the clinical SMEs and have not been edited for grammar or style.) Please post progress on this challenge to this GitHub ticket.
Case 1 The focus case is a 13 month female with alternating hemiplegia phenotype. She presented initially shortly after birth with abnormal eye movements/nystagmus. She currently displays global developmental delay, and hypotonia with paroxysmal episodes of dystonia, tremors and/or weakness as well as probable seizures.
Gene: ATP1A3 (NCBIGene:478) Variant: c.2591A>G; p.(Gln864Arg)
HPO/Phenotype Terms:
Case 2 A 17-year-old male presenting at age 16 with progressive ataxia, neuro-behavioral impairment and cerebellar atrophy with regression. There is history of expressive language delay and longstanding mild difficulties with coordination. Current symptoms/examination notable for scanning speech, dysmetria, ataxic gait, hand tremor, dystonia, parkinsonism and dysphagia.
Gene: ATP1A3 Variant: c.985G>A; p.Gly329Ser HPO/Phenotype terms:
Case 3
7-year-old male who presented at five months of age with episodic spells described as head turning and head tremor. He has been diagnosed with alternating hemiplegia of childhood, global developmental delay and myoclonic epilepsy.
Gene: ATP1A3 Variant: c.2440G>A p.Asp814Asn HPO/Phenotype terms:
Case 4 17-year-old female who presented with global developmental delay and spells described by caretakers as leg weakness and inability to move limbs. These episodes vary from side to side. She also has multiple behavioral concerns, including depression and self-aggression. She has as well abnormal muscle biopsy with electron chain abnormalities and an array showing deletion 17p12 involving PMP gene. (EDIT 2022-10-12: The 17p12 deletion likely indicates a second genetic disease, though typically the second disorder does not onset till adulthood so the phenotype in child not likely attributable to the deletion.)
Gene: ATP1A3 Variant: c.2303A>G (p.Try768Cys) HPO/Phenotype terms:
Case 5 A 15-year-old male with global developmental delay and paroxysmal episodes of primarily dystonic posturing associated with possible myoclonus and chorea. Brain MRI within normal limits.
Gene: ATP1A3 Variant: c.2767G>A, p.Asp923Asn HPO/Phenotype terms:
Timeline
October 7, 2022 - First Friday of the Month Standup
Check-in/update from teams on progress
October 14, 2022 - Second Friday of the Month Standup Check-in/update from teams on progress
October 20, 2022 - Third Thursday of the Month Translator QotM mini-hackathon Review answers and refine queries with SME; wrap-up challenge
October 28, 2022 - Fourth Friday of the Month Translator QotM Challenge summary published in Gazette; next month's QotM Challenge announced