Search Mailing List Archives
[bioontology-support] [BioPortal] Feedback from Michael Baudis
Jennifer Leigh Vendetti
vendetti at stanford.edu
Wed Feb 3 16:10:37 PST 2016
On Feb 2, 2016, at 5:16 AM, Michael Baudis <mbaudis at gmail.com<mailto:mbaudis at gmail.com>> wrote:
(Maybe the message was truncated? It was ..)
Medulloblastomas <- Gliomas <- Neoplasm, benign
… which both are wrong (not glioma, not benign, most gliomas not benign).
Maybe there is a general error in sub-/up-classing ICD-O 3? The last of the codes’ 5 digits is special and designates “malignancy” (well, histologically). So 8500/0 is not a superset of 8500/3, but something different (true adenomas may not be carcinoma precursors etc.).
Anyway, a search leading from medulloblastomas (a set of PNET tumors) to benign gliomas is extremely misleading (biologically & medically).
How was this schema derived?
We use NLM's MetamorphyoSys installation tool  to load the UMLS sources into a MySQL database hosted here at Stanford. We then use a Python script to convert all of the UMLS ontologies in the MySQL database to RDF files that are useable by the BioPortal application. The “umls2rdf” Python code is open source and available in our GitHub repository:
I’m not the original author of the script, but my understanding is that the UMLS hierarchy is transformed into a subClassOf tree and the other attributes into RDF predicate-value sets. I don’t know offhand if the irregularities you pointed out above are encoded in the UMLS source data, or if it’s an issue with our conversion script. I’ve created a ticket in our issue tracker so that we can investigate (NCBO-1640).
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the bioontology-support