Creating an ontology project
January 8, 2014 Leave a comment
UPDATE: see the latest post on this subject.
This article describes how to manage all the various files created as part of an ontology project. This assumes some familiarity with the unix command line. The article does not describe how to create the actual content for your ontology. For this I recommend the material from the NESCENT course on building anatomy ontologies, put together by Melissa Haendel and Matt Yoder, as well as the OBO Foundry tutorial from ICBO 2013.
Some of the material here overlaps with the page on making a google code project for an ontology.
Let’s make a jelly project
For our purposes here, let’s say you’re a Cnidaria biologist and you’re all ready to start building an ontology that describes the anatomy and traits of this phylum. How do you go about doing this?
Well, you could just fire up Protege and start building the thing, keeping the owl file on your desktop, periodically copying the file to a website somewhere. But this isn’t ideal. How will you track your edits? How will you manage releases? What about imports from other ontologies (you do intend to import parts of other ontologies, don’t you? If the answer is “no” go back and read the course material above!).
It’s much better to start off on the right foot, keeping all your files organized according to a directory structure common layout, and making use of simple and sensible practices from software engineering.
As part of your release process you’ll make use of OWLTools and OORT, which can be obtained from the OWLTools google code repository.
Make sure you have OWLTools-Oort/bin/ in your PATH
The first thing to do is to create a directory on your machine for managing all your files – as an ontology developer you will be managing a collection of files, not just the core ontology file itself. We’ll call this directory the “ontology project”.
To make things easy, owltools comes with a handy script called create-ontology-project to create a stub ontology project. This script is distributed with OWLTools but is available for download here:
The first thing to do is select your ontology ID (namespace). This *must* be the same as the ID space you intend to use. So if your URIs/IDs are to be CNIDO_0000001 and so on, the ontology ID *must* be “cnido“. Note that whilst your IDs will be in SHOUTY CAPITALS, the actual ontology itself is all ~~gentle lowercase~~~, even the first letter. This is actually part of OBO Foundry ID policy.
Running the script
Now, type this on the command line:
(you will need to add this to your path – it’s in the OWLTools-Oort/bin directory).
You will see the following output:
* Create a git or svn project. E.g.:
git add -A
git commit -m 'initial commit'
* Now visit github.com and create project cnido-ontology
* Edit ontology and create first release
* Create a jenkins job
You may not need all the stub files that are created from the outset, but it’s a good idea to have them there from the outset, as you may need them in future.
I recommend your first step is to follow the instructions above to (1) initiate a local git repository by typing the 4 commands above (2) publish this on github. You will need to go to github.com, create an account, create a project, and select “create a project from an existing repository“. (more on this later).
Once this is done, you can start modifying the stub files.
The top level README provides a high level overview of your project. You should describe the content and use cases here. You can edit this in a normal text editor — alternately, if you intend to use github (recommended) then you can wait until you commit and push this file and then edit it via the github web interface.
You will probably spend most of your time in the src/ontology directory, editing cnido-edit.owl
If you intend to have multiple people editing, then the cnido-idranges.owl file will be essential. You can edit this directly in Protege (but it may actually be easier to edit the file in a text editor). Assign each editor an ID range here (just follow the existing file as an example). Note that currently Protege does not read this file, so this just serves as formal documentation.
In future, id-ranges may be superseded by urigen servers, but for now they provide a useful way of avoiding collisions.
If you use github or another hosting solution like google code, you can use their wiki system. You should keep any files associated with the documentation (word docs, presentations, etc) in the doc/ folder. You can link to them directly from the wiki.
You can ask the OBO admins to help you set up a purl redirect such that URLs of the form
Will redirect to your images/ directory, which is where you will place any pictures of jelly fish or anenome body parts that you want to be associated with classes in the ontology (assuming you have the rights to do this). I recommend Jim Balhoff’s depictions plugin.
Managing your imports can be a difficult task and deserves its own article.
For now you can browse the ctenophore-ontology project to see an example of a setup, in particular:
This setup uses the OWLAPI for imports, but others prefer to make use of OntoFox.
You can use OORT to create your release files. The auto-generated Makefile stub should be sufficient to manage a basic release pipeline.In the src/ontology directory, type this:
This should make both cnido.obo and cnido.owl – these will be the files the rest of the world sees. cnido-edit is primarily seen by you and your fellow cnidarian-obsessed editors.
Depending on the specific needs of your project, some of the defaults and stubs provided by the create-ontology-project script may not be ideal for you. Or you may simply prefer to create the directory structure manually, it’s not very hard – this is of course fine. The script is provided primarily to help you get started, hopefully it will prove useful.
Finally, if you know any cnidarian biologists interested in contributing to an ontology, let me know as we are lacking detailed coverage in existing ontologies!