Medgen updates #7
Conversation
joeflack4
force-pushed
the
update1
branch
5 times, most recently
from
df9d980
to
f208076
Compare
| @@ -0,0 +1,115 @@ | |||
| """Mapping status between Medgen and Mondo""" | |||
There was a problem hiding this comment.
Repurposed some of my GARD code to create these attached atefacts.
Feels a bit duplicative, but this is probably fine for now. However I'm likely to do more of this for other ingests as well.
I did simplify some of it. Most of the code is the same between Medgen and GARD, but some things are special to each.
medgen_terms_mapping_status.tsv.zip
obsoleted_medgen_terms_in_mondo.txt
Some counts:
# tot_medgen_only = len(existing_overlap_df[existing_overlap_df['status'] == 'medgen']) # n=66,224
# tot_mondo_only = len(existing_overlap_df[existing_overlap_df['status'] == 'mondo']) # n=2,362
# tot_both_only = len(existing_overlap_df[existing_overlap_df['status'] == 'both']) # n=14,263
| mondo_df['prefix'] = mondo_df['object_id'].apply(lambda x: x.split(':')[0]) | ||
| mondo_df = mondo_df[mondo_df['prefix'].isin(MEDGEN_PREFIXES)] # n=16,627 | ||
| del mondo_df['prefix'] | ||
| # preds = list(mondo_df['predicate_id'].unique()) # only skos:exactMatch |
There was a problem hiding this comment.
exactMatch thoughts
I imagined we want to do the same thing here as with GARD where we only care about exact matches.
However a couple things I discovered so far are:
- Our prior Mondo->Medgen mappings were only of
skos:exactMatch - All of the mappings coming out of Chris' pipeline are only of:
oboInOwl:hasDbXreforowl:equivalentClass
There was a problem hiding this comment.
Replace all owl:equivalentClass to skos:exactMatch in the Medgen ingest. owl:equivalentClass is no longer relevant anywhere across our pipelines.
There was a problem hiding this comment.
I believe I've figured out how to do this. In OBO, this:
equivalent_to: <CURIE>
Should be changed to this:
property_value: exactMatch <CURIE>
I just need to change the Perl that creates the OBO like this. Then the pipeline will generate the correct OWL.
…hey started with 'C' and a number. - Update: prefixes: In addition to new classes above, renamed UMLS prefix with Medgen for all other classes (which happen to all start with 'CN:' - Update: prefixes: Renamed prior MEDGEN: xref prefixes to Medgen_UID: These IDs don't start with C (CUI; Concept Unique Identifier) or CN (Common Name?). These are internal Medgen UIDs that are duplicative and not for clinical or analytical use. - Rename: bin/ -> src/ - Add: output/: For both release outputs and non-release. - Rename: release/ -> output/release/ - Add: mondo_mapping_status.py: For generating artefacts related to the reporting and management of mappings between Mondo and Medgen. - Add: Python dependency requirements files. - Add: run.sh: For running commands in ODK - Add: config/medgen.sssom-metadata.yml
joeflack4
force-pushed
the
update1
branch
4 times, most recently
from
a89f466
to
05e8ecd
Compare
…riples as a function, (ii) updated namespacing of classes based on what type of MedGen/UMLS identifier they are. - Update: Namespaces MedGen, MedGen_UI (removed), MedGenCUI - Bugfix: SSSOM metadata yaml had a typo preventing conversion - Bugfix: Makefile: (i) needed to rename a dependency, (ii) needed to run 'analyze' step after 'stage' - Update: Makefile: Simplified some goals - Bugfix: For UMLS CUIs (e.g. starts with C then #s), we chose to do duplicate classes with namespaces UMLS and MedGen. However, I just now made it so that also all references (e.g. xrefs) are also duplicated, e.g. MedGen:1 maps to MedGen:2 and UMLS:2.
|
@matentzn I'm going to merge this one for now. A handful of misc work was done on this ingest, but we're going to be pausing it for now, holding off to see if the MedGen team is able to handle a lot of the work that this ingest would otherwise be doing. For future work on this ingest, when/if ever needed, I'll open up new PRs. I did wrap up a couple things from our least meeting though. Had to change a conditional block, and updated the namespaces, e.g. MedGen -> MEDGEN. |
Updates
a4eff72a96a4a018fb46a1222f25c968312cecb9f461d52abf3c7eb4981c43ab0a63aca653a333db