This is a complicated ticket for various reasons. Currently NIF has cell types
in two resources:
* An OWL ontology in the NIFSTD suite called "NIF_Cell"
* A semantic mediawiki (http://neurolex.org)
NIF_Cell is the easiest to work with as it follows relatively standard
conventions for OWL ontologies. However, this is extremely out of date, and has
not been updated for years, and there appear to be no plans to update it.
Neurolex is more current. However, it is extremely difficult to work with, the
representation is highly non-standard, it is full of broken links.
Additionally, it incorporates broken mass imports of other resources, such as
the neuron subset of the Drosophila anatomy (FBbt). We do not need this, as we
already have FBbt in the source OWL format which is well-integrated with CL
already.
In order to work with neurolex, I have developed a programmatic library for
interrogating it and for extracting what is the best I can do for some kind of
sane OWL: https://github.com/cmungall/nlx-pl
Please see the docs on this site for the various caveats.
Whether we are dealing with nlx or NIF_GA, one important thing to do is to map
brain regions; this is because so many of the neurons are brain type-specific.
Most of this work has been undertaken as part of the uberon project. See:
* https://github.com/obophenotype/uberon/wiki/Integration-with-neuroanatomical-atlases
* https://github.com/obophenotype/uberon/issues/300
For the uberon integration, we decided to lump nlx and nif_GrossAnatomy
together. We use the prefix http://uri.neuinfo.org/nif/nifstd/ for all
entities, whether from nlx or the OWL.
Another issue to be aware of is the fact that in CL we use part_of where we
should perhaps be using the shortcut soma_part_of
So far, we have done some preliminary alignment with NIF_Cell (see src/obol).
There is a suprisingly low degree of overlap. CL tends to have more grouping
classes, with nif/nlx having lots of very highly specific classes. I will add
more on the reporting later.
This can be visualized by opening
http://purl.obolibrary.org/obo/cl/cl-plus-nif.owl in Protege. This brings in
both CL and NIF_Cell with equivalence axioms.
TODO:
* Analyze further distinctions in content between NIF_Cell and nlx
* Devise strategy for specific cell types - leave as 'leaf' nodes or integrate fully?
Original issue reported on code.google.com by cmung...@gmail.com on 10 Apr 2015 at 1:19
Original issue reported on code.google.com by
cmung...@gmail.com
on 10 Apr 2015 at 1:19