Test grouping terms in mapping tool via NCImt hierarchy

[balhoff]

See https://github.com/cancerDHC/tools/issues/7#issuecomment-693516046

[gaurav]

Also covered by https://github.com/gaurav/umls-rrf-scala/issues/3

[decorons]

Are your referring to NCI Metathesaurus hierarchy or NCI Thesaurus hierarchy for this activity title? Thanks. Sherri

[gaurav]

I don't think we'd ever need to use the NCI Thesaurus hierarchy -- if we have an exact match to an NCIt term, then we're done, and if we have a broad match, I don't think there's any way of using the hierarchy to improve the match. So, in practice we'd only need to use the NCI Metathesaurus hierarchy here.

My idea is to use two pieces of information from the NCI Metathesaurus to try to group terms that we don't have a match for (and, possibly, to validate other matches, see #24):

• In some cases, such as SNOMED 397522002 (NCImt C0181448) (Keratoscope), the NCI Metathesaurus entry says that the "Semantic Type" of this term is "Medical device".
• Some terms, such as SNOMED 129722001 (NCImt C1268650) (Finding partially removed since previous mammogram), don't have this information. However, we can look up the parent of this term in NCImt (in this case, from the SNOMED hierarchy), and keep going until we get to an NCI Thesaurus term, in this case NCIt C77140 (NCImt C0456984) (Clinical Test Result, semantic type: Finding), which has a parent term of NCIt C3367 (NCImt C1285578) (Finding). That might be useful information in grouping terms that remain to be matched.

That's my plan at the moment -- let me know if there's some way to use either NCIt/NCImt hierarchies that I don't know about!

[decorons]

Thanks, Gaurav. Clever. I think I get it. It may be a more general purpose approach for non NCIt terms. Forwarded to Gilberto and Lyuba who also attend the terminology meetings I think. Sherri

From: Gaurav Vaidya notifications@github.com Date: Monday, November 9, 2020 at 12:08 PM To: cancerDHC/tools tools@noreply.github.com Cc: De Coronado, Sherri (NIH/NCI) [E] decorons@mail.nih.gov, Comment comment@noreply.github.com Subject: Re: [cancerDHC/tools] Test grouping terms in mapping tool via NCImt hierarchy (#20)

I don't think we'd ever need to use the NCI Thesaurus hierarchy -- if we have an exact match to an NCIt term, then we're done, and if we have a broad match, I don't think there's any way of using the hierarchy to improve the match. So, in practice we'd only need to use the NCI Metathesaurus hierarchy here.

My idea is to use two pieces of information from the NCI Metathesaurus to try to group terms that we don't have a match for (and, possibly, to validate other matches, see #24https://github.com/cancerDHC/tools/issues/24):

• In some cases, such as SNOMED 397522002 (NCImt C0181448)https://ncim.nci.nih.gov/ncimbrowser/ConceptReport.jsp?dictionary=NCI+Metathesaurus&code=C0181448 (Keratoscope), the NCI Metathesaurus entry says that the "Semantic Type" of this term is "Medical device".
• Some terms, such as SNOMED 129722001 (NCImt C1268650)https://ncim.nci.nih.gov/ncimbrowser/ConceptReport.jsp?dictionary=NCI+Metathesaurus&code=C1268650 (Finding partially removed since previous mammogram), don't have this information. However, we can look up the parent of this term in NCImt (in this case, from the SNOMED hierarchy), and keep going until we get to an NCI Thesaurus term, in this case NCIt C77140 (NCImt C0456984)https://ncim.nci.nih.gov/ncimbrowser/ConceptReport.jsp?dictionary=NCI+Metathesaurus&code=C0456984 (Clinical Test Result, semantic type: Finding), which has a parent term of NCIt C3367 (NCImt C1285578)https://ncim.nci.nih.gov/ncimbrowser/ConceptReport.jsp?dictionary=NCI%20Metathesaurus&code=C1285578 (Finding). That might be useful information in grouping terms that remain to be matched.

That's my plan at the moment -- let me know if there's some way to use either NCIt/NCImt hierarchies that I don't know about!

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/cancerDHC/tools/issues/20#issuecomment-724147081, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABVZR4EGXEHYP6WDU5RNSWLSPAOXHANCNFSM4SPQOPPA.

[gaurav]

Sounds good, Sherri! I'd love to know what they make of this plan.