Closed paolaroncaglia closed 4 years ago
Note for self: I'll document my brain dump here.
Current EFO TA name: Cardiovascular disease
EFO has: EFO:0000319 cardiovascular disease (disease location: UBERON:0004535 cardiovascular system) Among its children are: EFO:0003777 heart disease (disease location: UBERON:0000948 heart) EFO:0004264 vascular disease (disease location: UBERON:0001981 blood vessel)
UBERON 'cardiovascular system' in EFO has 2 subclasses: EFO heart component + UBERON heart conduction system UBERON:0000948 heart and UBERON:0001981 blood vessel in EFO are both part_of UBERON 'cardiovascular system'
[x] Check the class usage of EFO:0000319 cardiovascular disease and EFO UBERON cardiovascular system. EFO:0000319 cardiovascular disease has 322 uses. Though many are annotations to the term itself. All others seem to be a) diseases that are children of CV disease b) measurement terms that are is_about CV disease. EFO UBERON cardiovascular system has 136 uses.
[x] Remove them from EFO - how? Obsolete the EFO term? Can't, because it's gwas trait = true. Mark the UBERON term not for use? (Rather than obsoleting it in EFO, bcs David OS advised against that practice) But would it create trouble if it stayed, since we import UBERON dynamically? Zoe thinks it should be ok if we tag it "not for use" and implements checks to ensure it stays that way when we build EFO and when we import UBERON.
[x] Remove EFO:0000319 cardiovascular disease from MONDO mapping files
[x] Check class usage of EFO heart component + UBERON heart conduction system
[x] Do they need to stay?
MONDO has 'cardiovascular disease' with children 'heart disease' and 'vascular disease'. Their disease locations are, respectively, UBERON cardiovascular system, UBERON:0000948 heart and UBERON:0002049 vasculature (NOT blood vessel)
UBERON blood vessel is a descendant of UBERON vasculature UBERON is multi-species, and vasculature works for non-blood systems too
TBC
Summary for requested OT TAs cardiac disorders + vascular disorders:
cardiac disorder already exists in EFO as EFO:0003777 heart disease (isn't a TA but we can make it so) Subclass of cardiovascular disease and thoracic disease
vascular disorder already exists in EFO as EFO:0004264 vascular disease (isn't a TA but we can make it so) Subclass of cardiovascular disease
In order for heart disease + vascular disease to be directly under disease, we’d need to
But:
Check children of cardiovascular disease to see if they can be classified under heart disease, vascular disease or the other therapeutic area for blood disease.
As agreed above, here's a review of all children of 'cardiovascular disease', shortened in CVD, starting from the top in the OLS view. All parents will be looked at. Any parent that is not specifically mentioned is good as is and should not be modified. Note, if links to be deleted are inferred, we need to look at what's causing the inference. Presumably, many inferred links are due to blood being part of cardiovascular system in EFO; it isn't in UBERON, so we'll need to address that first.
[x] EFO > Orphanet:168612 Congenital deficiency in alpha-fetoprotein => delete link to CVD
[x] EFO:1000289 Hemangiopericytic Neoplasm => delete link to CVD and make it subclass of vascular disease instead
[x] EFO:0004264 vascular disease => Consider broadening its equivalence axiom Equivalent to: disease and has_disease_location some (blood vessel or part_of some blood vessel) so it points to vasculature instead of blood vessel - that would allow better autoclassification as some terms point to vasculature and some to blood vessel. MONDO has vasculature. EFO has both UBERON blood vessel and UBERON vasculature, but they are not connected. In UBERON, blood vessel is_a vessel and part_of blood vasculature, and blood vasculature is_a vasculature. In EFO, blood vasculature is missing. So I think I need to
[x] import UBERON blood vasculature
[x] broaden axiom vascular disease has disease location vasculature
[x] run the reasoner and check that more terms now classify correctly under vascular disease (currently 45 direct children, though looking at class usage of vasculature, we may only gain 1 child at the moment)
[x] EFO > Orphanet:168615 Hereditary persistence of alpha-fetoprotein => delete link to CVD
[x] EFO:1000297 Histiocytic and Dendritic Cell Neoplasm => delete link to CVD
[x] EFO:1000343 lympoid tissue (stemming from the term above) Subclass of 'cell type'; no definition; only used in axiom Histiocytic and Dendritic Cell Neoplasm has disease location lympoid tissue; Presumably a typo and mis-placement for UBERON:0001744 lymphoid tissue => fix it (i.e. change axiom Histiocytic and Dendritic Cell Neoplasm has disease location lympoid tissue to lymphoid tissue; obsolete EFO:1000343 lympoid tissue bcs it's incorrect; replace with UBERON:0001744 lymphoid tissue)
[x] EFO:1000339 Lymphangiosarcoma => delete link to CVD and make it subclass of lymphatic system disease instead (but see https://github.com/EBISPOT/efo/issues/485#issuecomment-510016443)
[x] EFO:1001405 Postthrombotic Syndrome is the same as EFO:0007452 post-thrombotic syndrome. Merge the two and keep EFO:0007452 post-thrombotic syndrome, which is better defined/placed/mapped; make sure all dbxrefs are retained (EFO:1001405 wasn't added for UK Biobank work, so no problem there).
[x] EFO:1000014 acidosis => delete link to CVD
[x] All the leukemia/lymphoma terms should not be children of CVD. Presumably this is due to the disease location axiom pointing to blood. If so, that should be fixed when we remove blood from being a part of cardiovascular system, see above. May also want to check corresponding MONDO terms and their axioms using Ontobee. Also, note from Sandra: "All the leukemia/lymphoma terms – Could go to Neoplasm TA and TA Blood & lymphatic tissue disorder".
To be continued.
Continued:
(CVD = cardiovascular disease)
[x] EFO:1000782 altitude sickness We debated its location in https://github.com/EBISPOT/efo/issues/383. In the end we didn't make it a direct child of 'disease' because we couldn't really reach a decision. But perhaps we should. If so, re-open #383. Also, note from Sandra: "EFO:1000782 Altitude sickness (also linked to ticket: #383.) - Could go to a Respiratory TA. Definition is here: “Altitude sickness occurs because the partial pressure of oxygen decreases with altitude. (…) there is a substantially lower amount of oxygen available for the individual to inhale “ It shouldn’t go to some blood related TA as it’s the oxygen intake that’s the problem, thus resulting in too little oxygen in the blood. But it is not a blood disorder.".
[x] EFO:1001263 amniotic fluid embolism => delete link to CVD and make it a subclass of vein disease instead.
[x] EFO:1001493 cardiac embolism => delete link to CVD and make it a subclass of vascular disease instead (not heart disease, because it's about heart vessels; we may need to add to/improve the UBERON representation in EFO for this area, and clarify with OT if it's ok to keep heart and heart vessels separate)
[x] EFO > MONDO:0024757 cardiovascular neoplasm This needs to stay under CVD, but it will also be in the neoplasm TA, and all its children have ancestors in other TAs (haven't checked children's children though), except for
[x] EFO > MONDO:0024296 vascular neoplasm => delete link to CVD and make it a subclass of vascular disease instead
[x] EFO > MONDO:0024239 congenital anomaly of cardiovascular system This needs to stay under CVD, but its children have parents in other TAs too. Also, note from Sandra: "Congenital anomaly of cardiovascular system – Maybe consider moving this to the Congenital/Genetic TA. Note from @d0choa: under both (congential + CVD) would make sense for us".
To be continued with terms from epistaxis down.
Notes from S. on some of these actions:
Continued from above (OLS view under CVD, from epistaxis down):
[x] EFO:0003895 epistaxis Change parent CVD into more specific vascular disease, and add parent hemorrhage.
[x] EFO:1000938 fibromuscular dysplasia Change parent CVD into more specific arterial disorder.
[x] EFO:0007319 hyperprolactinemia Should not be child of CVD. Presumably this is due to the disease location axiom pointing to blood. If so, that should be fixed when we remove blood from being a part of cardiovascular system, see above. May also want to check corresponding MONDO terms and their axioms using Ontobee.
[x] EFO:0005251 hypotension "Blood pressure that is abnormally low." (from NCIT). This is tricky. It's a direct child of CVD in both EFO and MONDO. But compare with EFO hypertension/MONDO hypertensive disorder which is child of arterial disease ("Persistently high systemic arterial blood pressure"). The Wikipedia entry says "Hypotension is low blood pressure, especially in the arteries of the left sided systemic circulation.", so not conclusive as to whether it'd be ok to move under arterial disease. No other parent. => move under arterial disease, but also discuss with MONDO?
[x] EFO:1001032 malignant lymphatic vessel tumor Change parent CVD into a more specific one depending on where lymphatic vessels end up, so that the term sits, in OT's view, under the TA for blood and/or lymphatic system disorder (as well as under the neoplasm TA, which is already the case).
[x] EFO:0009087 non-typhoidal Salmonella bacteremia Should not be child of CVD. Presumably this is due to the disease location axiom pointing to blood. If so, that should be fixed when we remove blood from being a part of cardiovascular system, see above. Also, add parent salmonellosis.
TBC (still missing: pernicious anemia, septicemic plague)
@d0choa Could you and Sandra please provide feedback on the placement of EFO > MONDO:0000474 'pericardium disease' with regards to the therapeutic areas of cardiac disorders and/or vascular disorders. It is currently a direct child of cardiovascular disease (CVD) (as well as thoracic disease). Its textual and logical definitions point to pericardium which contains the heart and the roots of the great vessels; for the OLS view of pericardium, see https://www.ebi.ac.uk/ols/ontologies/uberon/terms?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FUBERON_0002407 We would keep this term under CVD in EFO, but for the OT TAs, would it be enough to have it under the thoracic disease TA? Perhaps we could also place it under connective tissue disease based on the composition of the pericardium, but I'm really not sure because of its multi-tissue structure.
Thanks,
Paola
Reply from Sandra: Pericardium disease would best go to Cardiac disease (cardiac disorders are defining the heart and associated blood vessels, whereby vascular disorders are pertaining to the blood vessels not associated with the heart). It should not go to connective tissue disease.
Regarding the thoracic disease, we are aiming to make this TA obsolete as it refers to a purely anatomical region with no clinical relevance. Meddra has no equivalent System organ class
, so we expect to accommodate all diseases currently contained there in at least 1 other TA.
@d0choa (and Sandra) Thanks. Could you please let me know what exactly is your source for "cardiac disorders are defining the heart and associated blood vessels", so I can quote that in EFO. I can see the logic, but I'd like to avoid the reasoner throwing errors if possible. I might need to create cardiac disease as a parent of heart disease - or I might simply delete the location axiom from pericardium disease - I'd like to find an easily maintainable solution. The only place where I could easily find "cardiac" as referring to heart as well as its surroundings is the Merriam-Webster: "of, relating to, situated near, or acting on the heart" (https://www.merriam-webster.com/dictionary/cardiac). Thanks, Paola P.S. I might have found the solution: UBERON has a grouping class for 'heart plus pericardium'; that would sort the logical definitions. I just need sources/references to be able to create a cardiac disease class that refers to heart plus pericardium and that would not be in disagreement with other resources. I think that should be acceptable. A possible issue would be to fix cases where cardiac disease is used as an exact synonym of heart disease, while it should be narrow if we go for this fix.
Continued from above (OLS view under CVD, last two children):
[x] EFO:0005576 pernicious anemia Should not be child of CVD. Presumably this is due to the disease location axiom pointing to blood. If so, that should be fixed when we remove blood from being a part of cardiovascular system, see above. It's already in the hematologic disease TA.
[x] EFO:0007481 septicemic plague Should not be child of CVD. Presumably this is due to the disease location axiom pointing to blood. If so, that should be fixed when we remove blood from being a part of cardiovascular system, see above. It's already in the hematologic disease TA.
[x] EFO:0004264 vascular disease By definition, it comprises the lymphatic vessels, but OT would like a separate TA for lymphatic system disorders. That might still be ok, but need to check.
@d0choa please note, I extended my comment/question to you above, thanks.
The UBERON class heart plus pericardium
is what we were aiming for. If you look at Mesh the heart
class includes the pericardium
. However, their definition is not very verbose and makes no distinction.
We would advise against "cardiac disease as a parent of heart disease" as cardiac and heart should be synonyms.
Sandra found some accurate definitions in this page:
See #485 for discussion about lymphatic vessels
@d0choa and Sandra Thanks for your prompt feedback, I do appreciate it. All things considered then, I think it would be quicker and simpler to make pericardium disease a child of heart disease (exact synonym cardiac disease) in the slim for OT TAs only, and keep it as it is in 'regular' EFO. Should be an acceptable compromise for the time being. That would avoid having to make cardiac disease different from heart disease. The descriptions from the Wake Forest health provider are helpful, but I'm afraid not ideal as ontology references :-) Best, Paola
Note for self: summing up, in short,
This solution will work for us. However, we think the proposed changes are not an OT specific problem and will benefit EFO as a whole.
Since the UBERON pericardium is part of UBERON heart, we understand pericardium disease should be a child of heart disease and not of cardiovascular disease.
We thought it will be simple as UBERON provides the desired anatomical description. If the problem is with the definition we can try to find a better source, but the one we trust the most we can not share ;)
Thanks @d0choa . That's the problem - UBERON pericardium is not part of UBERON heart; it is part of UBERON 'heart plus pericardium', see the section on "Superclasses & Asserted Axioms" in the Ontobee link you provided above. UBERON heart refers to the muscle alone, and UBERON pericardium refers to the sac around the heart. To my knowledge, those UBERON anatomical descriptions are correct and shared by the community. I understand the reasons why the OT view would like pericardium diseases associated with heart diseases, and I understand that cardiac doctors deal with pericardium as well as heart, and we'll have that in the slim we'll provide to you. But to have the same in EFO, we would need the disease class to refer to 'heart plus pericardium', and then we can't call it 'heart disease' because that refers to heart alone. And we can't call it cardiac disease either because that's an exact synonym of heart disease. It doesn't simply depend on the definition. As I said, don't worry, it will be ok in the slim we'll provide. :-)
Sorry it feels like I'm misreading the ontology and waisting your time.
Can I have an extra question for me to understand? Why does the OLS pericardium entry show part_of some heart
as a Term relationship? This and the OLS tree made me think that according to UBERON pericardium
is a part of the heart
.
@d0choa No problem at all. Yes the OLS view of the EFO pericardium term is confusing and doesn't reflect the placement that the term should have based on the UBERON file that EFO imports (and that shows the same as the Ontobee view we discussed). @zoependlington and @S2Ola are looking into this issue in case something went wrong during the UBERON import. It's good that we spotted this! Thanks David!
@d0choa FYI, my colleagues have identified the issue and will address it over the next couple of days. The fix may not be immediately visible in the EFO version in OLS, so for the time being, if you wish/need to double-check the placement of anatomy terms, please refer to the UBERON version currently in OLS. To make things even more interesting, please be aware that the UBERON version loaded in Ontobee is 6 months behind (OLS has the most recent one, but the default URI for UBERON terms redirects to Ontobee!). This said, while I wait for the issue to be fixed, I'll continue to work on the TAs. Best, Paola
Note for self: Most issues in this ticket have been addressed via creation of an OT TA slim/profile. Other issues in this ticket will be addressed as part of https://github.com/EBISPOT/efo/issues/571. I'll check boxes as I go to reflect the above, so I can focus on what really still needs to be done in EFO3. UPDATE: I went through all comments in this ticket;
Note for self: left to do for this ticket: looking at the post-reasoner inferred view in Protege (also see OLS),
Notes for self (no action item for EFO):
opened a Mondo ticket for 'fibromuscular dysplasia' (https://github.com/monarch-initiative/mondo/issues/1029);
opened a Mondo ticket for 'hypotension' (https://github.com/monarch-initiative/mondo/issues/1030).
@smnorthen @d0choa I've finished the revision of the cardiac + vascular disease areas as outlined in this ticket. There may still be issues with Orphanet terms and their parentage, but those will be addressed elsewhere. If you spot any problem with EFO-specific (or Mondo) descendants of cardiac + vascular diseases at a lower/more granular level, please let us know by opening new tickets. Thanks.
Proposed TA name:
car·di·ac = 1. Pertaining to the heart., 2. Pertaining to the esophageal opening of the stomach. , 3. (Obsolete). A remedy for heart disease. https://medical-dictionary.thefreedictionary.com/cardiac Cardiovascular = pertaining to the heart and blood vessels. https://medical-dictionary.thefreedictionary.com/cardiovascular vascular = pertaining to blood vessels or indicative of a copious blood supply. https://medical-dictionary.thefreedictionary.com/vascular
Currently in EFO: [OLS view of 'heart disease' and 'vascular disease']
ACTION: Update the therapeutic area list (ontology will keep the same): remove cardiovascular disease add heart disease add vascular disease We can do a pull-request for this action