Repository navigation
Add Journal Article -> main subject -> MeSH Terms #32
Description
Activity
Hi @Daniel-Mietchen . Yes, this is something that fits in to our workflow, and is something that could be added to WikidataIntegrator. This would be easy to implement provided the Wikidata item corresponding to the MeSH term already existed. Ideally, at some point we would import MeSH into Wikidata (or at least some parts of it). I imagine this would require a lot of work matching up items to prevent duplication though...
Looking at the API response from EuropePMC (and from Pubmed), the mesh terms don't actually include the mesh ID (of course, why would they?), but only the name. So either way, we would have to parse the MeSH dump file to get the MeSH ID from the name, unless you know of a better API?
We could also look into adding the items from the "chemicalList" field.
Reacted by Daniel Mietchen- changed the title
[-]MeSH-to-"main subject" pipeline[/-][+]Add Journal Article -> main subject -> MeSH Terms[/+]on Jun 12, 2017 Yes, the ~7K MeSH terms that are in Wikidata at the moment (which don't seem to be differentiated in terms of being Descriptors or Concepts) are just about 1/4 of the total number of MeSH Descriptors, but perhaps a good starting point for testing such workflows. For the rest, we would probably need to think about using Mix'n Match.
In terms of getting the ID, I just left the following message with the NLM Helpdesk (ticket #28045-219557):
I am interested in getting the MeSH Descriptor IDs associated with a PubMed ID via an API call.
I know how to get the MeSH Descriptor name from a PMID via API, e.g. "Coenzymes" for https://www.ncbi.nlm.nih.gov/pubmed/16046484 via
https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=16046484&tool=my_tool&email=my_email@example.com .I am also aware that I can use regexes in SPARQL to try to convert that string into the Descriptor ID, e.g. https://id.nlm.nih.gov/mesh/query?query=PREFIX+rdf%3A+%3Chttp%3A%2F%2Fwww.w3.org%2F1999%2F02%2F22-rdf-syntax-ns%23%3E%0D%0APREFIX+rdfs%3A+%3Chttp%3A%2F%2Fwww.w3.org%2F2000%2F01%2Frdf-schema%23%3E%0D%0APREFIX+xsd%3A+%3Chttp%3A%2F%2Fwww.w3.org%2F2001%2FXMLSchema%23%3E%0D%0APREFIX+owl%3A+%3Chttp%3A%2F%2Fwww.w3.org%2F2002%2F07%2Fowl%23%3E%0D%0APREFIX+meshv%3A+%3Chttp%3A%2F%2Fid.nlm.nih.gov%2Fmesh%2Fvocab%23%3E%0D%0APREFIX+mesh%3A+%3Chttp%3A%2F%2Fid.nlm.nih.gov%2Fmesh%2F%3E%0D%0APREFIX+mesh2015%3A+%3Chttp%3A%2F%2Fid.nlm.nih.gov%2Fmesh%2F2015%2F%3E%0D%0APREFIX+mesh2016%3A+%3Chttp%3A%2F%2Fid.nlm.nih.gov%2Fmesh%2F2016%2F%3E%0D%0APREFIX+mesh2017%3A+%3Chttp%3A%2F%2Fid.nlm.nih.gov%2Fmesh%2F2017%2F%3E%0D%0A%0D%0A+SELECT+%3Fd+%3FdName+%3Fc+%3FcName%0D%0A+FROM+%3Chttp%3A%2F%2Fid.nlm.nih.gov%2Fmesh%3E%0D%0A+%0D%0A+WHERE+%7B%0D%0A+++%0D%0A+%3Fd+a+meshv%3ADescriptor+.%0D%0A+%3Fd+meshv%3Aactive+1+.%0D%0A+%3Fd+meshv%3Aconcept+%3Fc+.%0D%0A+%3Fd+rdfs%3Alabel+%3FdName+.%0D%0A+%3Fc+rdfs%3Alabel+%3FcName%0D%0A+FILTER%28REGEX%28%3FdName%2C%22Coenzymes%22%2C%22i%22%29+%7C%7C+REGEX%28%3FcName%2C%22Coenzymes%22%2C%22i%22%29%29%0D%0A+%0D%0A+%7D%0D%0A+ORDER+BY+%3Fd%0D%0A%0D%0A&format=HTML&inference=true&year=current&limit=50&offset=0#lodestart-sparql-results for "Coenzymes", which gives https://id.nlm.nih.gov/mesh/D003067.html .
I do not see a direct way to convert from a PMID to the associated Descriptor IDs, though, and would appreciate pointers.
A few notes from me.
-
It looks like the XML file will go directly from a PMID to a MeSH ID (eg https://www.ncbi.nlm.nih.gov/pubmed/16046484?report=xml&format=text).
-
The XML also contains an attribute for "MajorTopicYN". For P921 "main subject" you might want to consider restricting to records with a value of "Y".
-
I think a bot to systematically load MeSH (or parts of it) will be part of the renewal proposal. Of course, some manual review will definitely be necessary...
-
You may find this service useful here. SPARQL over MeSH graph. https://id.nlm.nih.gov/mesh/
@andrewsu Thanks for the pointer to the XML.
I'm not sure yet what the "MajorTopicNY" switch does — in the example, it seems to only be used for qualifiers, while we would likely want to go mainly for descriptors.
@Daniel-Mietchen definitely used for descriptors too, eg https://www.ncbi.nlm.nih.gov/pubmed/26611529?report=xml&format=text
Yes, for other PMIDs, it is used for descriptors as well, but my point from above was that if the goal is to turn MeSH IDs into P921 statements, we probably cannot solely rely on those descriptors for which "MajorTopicNY" is "Y", since that would give nothing for the PMID=16046484 example.
If we were to include those descriptors for which at least one qualifier's "MajorTopicNY" is set to "Y", then we would get
- Antioxidants
- Heart Failure
- Ubiquinone,
which is probably close to what we would want for P921.
For PMID=26611529, we would get
- Endoplasmic Reticulum-Associated Degradation
- Peptide-N4-(N-acetyl-beta-glucosaminyl) Asparagine Amidase,
which seems reasonable as well.
I have made those sample edits to
https://www.wikidata.org/w/index.php?title=Q30247623&oldid=500642437#P921
and
https://www.wikidata.org/w/index.php?title=Q30274598&oldid=500644169#P921 .Is there a way to know whether there are PMIDs for which no "MajorTopicNY" value would be "Y"?
Yes, in principle, we'd like to make use of the information provided by the qualifiers as well, but not sure P366 is the way to go.
To go back to https://www.ncbi.nlm.nih.gov/pubmed/26611529?report=xml&format=text , we have, for instance
<MeshHeading> <DescriptorName UI="D043524" MajorTopicYN="Y">Peptide-N4-(N-acetyl-beta-glucosaminyl) Asparagine AmidaseDescriptorName> <QualifierName UI="Q000737" MajorTopicYN="N">chemistryQualifierName> <QualifierName UI="Q000235" MajorTopicYN="N">geneticsQualifierName> <QualifierName UI="Q000378" MajorTopicYN="N">metabolismQualifierName> MeshHeading>
I understand this to mean that the paper is about
- the chemistry of Peptide-N4-(N-acetyl-beta-glucosaminyl) Asparagine Amidase
- the genetics of Peptide-N4-(N-acetyl-beta-glucosaminyl) Asparagine Amidase
- the metabolism of Peptide-N4-(N-acetyl-beta-glucosaminyl) Asparagine Amidase
Unless we have dedicated items for such things, I could imagine translating this into statements like
- P921 (main subject): chemistry (Q2329), with a qualifier
- of (P642): Peptide-N4-(N-acetyl-beta-glucosaminyl) Asparagine Amidase (Q7166515)
I have made such a test edit to a sandbox item: https://www.wikidata.org/w/index.php?title=Q13406268&oldid=500822215#P921
Other issues:
- the number of statements derived from MeSH terms might be a concern — 26611529 has 14 Descriptors, which is close to the upper end of what I think is reasonable in terms of P921, and adding in the qualifiers means going well beyond 20, which I think is adding more noise than signal
- the hierarchy of the Descriptors, as seen through their Wikidata graph. For instance, "human" is on a subtree of "mammal", so perhaps we don't need to add "mammal" here unless it's MajorTopicYN="Y"?
Yeah, have to admit I'm not a fan of that sandbox model and the use of "of". Don't have an alternate suggestion, but just sharing my gut reaction.
I'd argue that you'd only want to use P921 "main subject" if MajorTopicYN="Y". That would also cut down on the number of MeSH term statements and eliminate less informative/specific ones eg "mammal"...
Meanwhile, all MeSH IDs are in Mix'n Match, and about 14k have been matched.
I'm looking for ways to get the following workflow going:
Would this be something that could be fit into GeneWiki workflows?