These notes are for the EDITORS of mondo
PLEASE READ THE DOCUMENTATION HERE
https://docs.google.com/document/d/19bp9MpCHCxbjMmbntB2e5gZNzzNlu06DnDB8xcoSXK8/edit#
The rest of this document describes generic ontology editing guidelines
Make sure you have an ID range in the idranges file
If you do not have one, get one from the head curator.
The editors version is mondo-edit.obo
** DO NOT EDIT mondo.obo OR mondo.owl in the top level directory **
../../mondo.owl is the release version
To edit, open the file in Protege. First make sure you have the repository cloned, see the GitHub project for details.
These are stored in the file
** ONLY USE IDs WITHIN YOUR RANGE!! **
If you have only just set up this repository, modify the idranges file and add yourself or other editors. Note Protege does not read the file
- it is up to you to ensure correct Protege configuration.
We aim to put this up on the technical docs for OBO on http://obofoundry.org/
For now, consult the GO Tutorial on configuring Protege
You should only attempt to make a release AFTER the edit version is committed and pushed, and the travis build passes.
to release:
cd src/ontology
make
If this looks good type:
make prepare_release
This generates derived files such as mondo.owl and mondo.obo and places them in the top level (../..). The versionIRI will be added.
Commit and push these files.
git commit -a
And type a brief description of the release in the editor window
Finally type
git push origin master
IMMEDIATELY AFTERWARDS (do not make further modifications) go here:
- https://github.com/monarch-initiative/mondo/releases
- https://github.com/monarch-initiative/mondo/releases/new
The value of the "Tag version" field MUST be
vYYYY-MM-DD
The initial lowercase "v" is REQUIRED. The YYYY-MM-DD must match what is in the versionIRI of the derived mondo.owl (data-version in mondo.obo).
Release title should be YYYY-MM-DD, optionally followed by a title (e.g. "january release")
Then click "publish release"
IMPORTANT: NO MORE THAN ONE RELEASE PER DAY.
The PURLs are already configured to pull from github. This means that BOTH ontology purls and versioned ontology purls will resolve to the correct ontologies. Try it!
- http://purl.obolibrary.org/obo/mondo.owl <-- current ontology PURL
- http://purl.obolibrary.org/obo/mondo/releases/YYYY-MM-DD.owl <-- change to the release you just made
For questions on this contact Chris Mungall or email obo-admin AT obofoundry.org
Check the build status here:
Note: if you have only just created this project you will need to authorize travis for this repo. Go to https://travis-ci.org/profile/monarch-initiative for details
Mondo QC is split into two conceptual parts:
- Violations
- Warnings
Violations have to be fixed and are executed every time a pull request is created. They are implemented as sparql queries with a postfix "violation", like no-label-violation.sparql. To create a new violation, we create a new SPARQL SELECT query that queries for the violation, and name it including the -violation postfix. The violation query is executed using ROBOT verify, which will make the CI (Travis/Jenkins) pipeline fail if the query has more than 0 results.
Warnings do not have to be fixed, but are indications of something going wrong. Warnings are implemented in the same way as violation queries with a SPARQL Select query, named with a postfix (example: single-child-warning.sparql).
Warnings generate a report in the src/ontology/reports/ directory, with that name scheme:
src/ontology/reports/ontologyname.owl-sparql-check-name.tsv, for example: src/ontology/reports/mondo.owl-single-child-warning.tsv. The same logic is used for ROBOT reports.
There now is a jupyter notebook (src/ontology/reports/mondo_analysis.ipynb) that automatically reads all files following the above naming schemes (and some more) and generating overviews. The notebook can be edited using:
sh run.sh make run_notebook
But usually, the only thing you would want to do is generating a markdown overview like so:
sh run.sh make reports/mondo_analysis.md
All tags are embedded into components/mondo-tags.owl. The way it works:
- All DOSDP matches are tagged with their corresponding pattern(s), using
dc:conformsTo. - All SPARQL queries with
-tags.sparqlin the name are expected to be CONSTRUCT queries (example:src/sparql/single-child-tags.sparql). They will match a pattern, and set a tag usingdc:conformsTo.