Standardized Term Names

[Chris Mungall]

Hi Alex - I agree completely. I will create a specific cell ontology
alignment report next obol run

On Mon, 21 Feb 2005, Alexander Diehl wrote:

> Chris,
>
> While I would like to see the cell ontology continued to be improved,
> I think that its use in constructing GO terms cannot be forced at
> this time for the reasons I mentioned previously.  In particular, I
> don't feel the cell ontology can be considered authoritative at this
> time and its naming conventions differ significantly from the GO.
> Nevertheless, it would be useful for interested parties to compare
> the cell ontology with the implicit GO cell ontology either
> computationally or by hand to look for discrepancies between the two
> and add appropriate synonyms where needed.  I would also suggest the
> GO editors at EBI could make a point of comparing the two ontologies
> whenever a GO term containing a cell type name is being coined or
> modified in order to ensure the two ontologies are more in harmony by
> suggesting changes to one or the other ontology.
>
> -- Alex
>
>
> At 11:46 AM -0800 2/21/05, Chris Mungall wrote:
> >Alexander,
> >
> >The cell ontology is still young and not yet seeing much active use, so
> >it's not surprising it's incomplete and has problems. I had envisioned one
> >of the main users of the cell ontology as being the GO curators. It seems
> >that instead GO curators are focusing their energies on getting the
> >implicit cell ontology within GO correct, and ignoring the cell ontology
> >(I may be mischaracterizing here, I don't know to what extent annotators
> >are doing curation).
> >
> >I guess this is perfectly understandable, given that this is the GO
> >consortium and not the cell ontology consortium. However, I think it is
> >extremely counterproductive in the long run.
> >
> >If we can somehow divert more energy to getting the cell ontology correct,
> >then we can use this for automatic correction in GO.
> >
> >I realise this may seem quite daunting. There are indeed glaring
> >inconsistencies. I'm not suggesting anyone sits down and manually sorts
> >these out. The technology is there to automate this task. Obol currently
> >reports inconsistencies, and obol is also capable of exporting the
> >implicit cell (or anatomy or chemical or protein) ontology from GO.
> >
> >Soon we will have the ability to make genuine cross products between GO
> >and CellOnt, with the cross-product recorded in the obo file. This will
> >allow OBO-Edit to automatically flag inconsistencies (rather than having
> >the curator examine a seperate text report).
> >
> >Of course, if this approach is to succeed, then it would require a slight
> >shift in working practice. There would have to be more cross-talk between
> >the cell ontology and GO. GO curators would still of course be free to
> >define their 'own' cell ontology within GO that need not be consistent
> >with the cell ontology (in either DAG structure, synonyms or spelling),
> >but this would be deeply counterproductive. The correct approach would be
> >to correct the cell ontology and have this automatically propagate into
> >GO.
> >
> >Does this sound reasonable?
> >
> >Cheers
> >Chris
> >
> >On Mon, 21 Feb 2005, Alexander Diehl wrote:
> >
> >>  The correct spelling is "eosinophil," which gets about 225000 hits in
> >>  Google.  "Oesinophil" seems to be a mispelling, and gets about 30
> >>  hits in Google.
> >>
> >>  As currently constituted, I would have a lot of problems with using
> >>  the cell ontology as a source for correct nomenclature for cell types
> >>  in the GO.  The cell ontology differs from standard usage for many
> >>  terms and is far from complete.  It also lacks definitions for its
> >>  terms as well as many commonly used synonyms.  There are glaring
> >>  discrepancies between cell types in the GO and cell ontology terms,
> >>  such as the use of T-cell in the GO and T_lymphocyte in the cell
> >>  ontology, with no synonyms in either place to provide a mapping
> >>  between the two terms.  All in all, one could probably construct an
> >  > equally valid cell type ontology by deconstructing GO terms.
> >>  Nevertheless, I do use the cell ontology in annotation notes to
> >>  provide clarifying information when the GO term is inadequate, and I
> >>  welcome its continued development.
> >>
> >>  GO term names should reflect actual usage and proper English, which
> >>  was the source of my initial complaint.
> >>
> >>  -- Alex
> >>
> >>  At 10:05 AM -0800 2/21/05, Chris Mungall wrote:
> >>  >[cc-ing Sue, even though I suspect she is already on the development
> >>  >mailing list]
> >>  >
> >>  >Let me see if I understand
> >>  >
> >>  >You want to just have name='oesinophil differentiation', and not bother
> >>  >having exact_synonym='oesinophil cell differentiation'?
> >>  >
> >>  >That seems very sensible, since it's not a good use of curator time to
> >  > >insert these automatable synonyms. You certainly shouldn't be spending any
> >>  >time on this to help an automatable parser like Obol!
> >>  >
> >>  >Are you also saying you'd like Obol to automatically add these synonyms?
> >>  >We're not really set up for this kind of dataflow at the moment. There
> >>  >would be some lag, during which time users search on Amigo for "oesinophil
> >>  >cell differentiation" would come back empty handed (I think this should
> >>  >actually be handled at the search level rather than at the synonym level,
> >>  >though again we aren't set up for this yet).
> >>  >
> >>  >What Obol really cares about for these terms (and what I think we should
> >>  >care about) is consistency with the cell ontology.
> >>  >
> >>  >Note that the cell ontology has "eosinophil" (NOT "oesinophil"). If this
> >>  >is a genuine synonym then there should be a dialog with the cell ontology
> >>  >curators to introduce that synonym in the cell ontology. Synonyms for GO
> >>  >can then be computed automatically for this.
> >>  >
> >>  >Note also that the cell ontology has "chromophil cell" and "neutrophil".
> >>  >Why not "chromophil" and "neutrophil cell"? If this reflects actual usage
> >>  >then the term names should reflect this. If there are reasons to believe
> >>  >that the cell ontology naming conventions are arbitrary, then the cell
> >>  >ontology should be modified to be consistent with actual usage. The GO
> >>  >should then take the cell ontology as the authorative source, and use the
> >>  >corresponding naming convention.
> >>  >
> >>  >Cheers
> >>  >Chris
> >>  >
> >>  >On Mon, 21 Feb 2005, Jennifer I Clark wrote:
> >>  >
> >>  >>  Hi Chris,
> >>  >>
> >>  >>  I was thinking more about the cell differentiation things. I reckon your
> >>  >>  obol parser would be okay without the synonyms as long as I tell you a
> >>  >>  few more naming rules.
> >>  >>
> >>  >>  We have the following kinds of differentiation terms:
> >>  >>
> >>  >>  x cell differentation
> >>  >>  xblast differentation
> >>  >>  xphil differentation
> >>  >>
> >>  >>  where the word endings 'blast' and 'phil' are synonyms for the word
> >>  >>  'cell' in the first example. If I just tell you about these extra rules
> >>  >>  then could you add them to the pareser? That way I wouldn't need to to
> >>  >>  bother with the synonyms. I could add documentation to the website to
> >>  >>  say that where these endings are in a term we don't add the word 'cell'
> >>  >>  since it is redundant.
> >>  >>
> >>  >>  Thanks,
> >>  >>
> >>  >>  Jen
> >>  >>
> >>  >>  Chris Mungall wrote:
> >>  >  >
> >>  >>  >Whichever way we do it, I don't think we should mark up things in the
> >>  >>  >ontology just to suit a particular piece of software.
> >>  >>  >
> >>  >>  >We should only do this if it doesn't impact curation time and if the
> >>  >>  >notion of a 'formal' synonym for a term is useful to humans.
> >>  >>  >
> >>  >>  >Here are the criteria for a formal synonym:
> >>  >>  >
> >>  >>  >- it must be an exact synonym
> >>  >>  >  [ and would still be designated as such in the obo file, it
> >>  >>  >    would just have an additional qualifier ]
> >>  >>  >
> >>  >>  >- the existing term name is some kind of colloquial shorthand
> >>  >>  >  that is readily recognised by most biologists, yet the naming
> >>  >>  >  style of the shorthand is different from the consistent naming
> >>  >>  >  style employed for similar terms
> >>  >>  >
> >>  >>  >- the existence of the synonym is a genuine rephrasing of the
> >>  >>  >  term name, and does not exist by virtue of alternate spellings
> >>  >>  >  of words.
> >  > >>  >
> >>  >>  >- the formal synonym must employ canonical spellings
> >>  >>  >
> >>  >>  >Why would this be useful to humans too? Well, sometimes it
> >>may be useful
> >>  >>  >to organise a set of terms such that either the formal synonym (or just
> >>  >>  >the name if there is no designated formal synonym) is used as
> >>the primary
> >>  >>  >label. Thus the term with name "methanogenesis" will be
> >>listed as "methane
> >>  >>  >biosynthesis" and will have consistent labeling style with other "X
> >>  >>  >biosynthesis" terms.
> >>  >>  >
> >>  >>  >If this isn't useful then we shouldn't bother.
> >>  >>  >
> >>  >>  >If we don't, then what I do think will be useful (and what
> >>will be quite
> >>  >>  >sufficient for obol) is to separate the useful exact_synonyms from the
> >>  >>  >exact_synonyms that are derivable from the exact_synonyms of sub-terms.
> >  > >>  >
> >>  >>  >An example is "sulfite reduction/sulphite reduction". This is
> >>only useful
> >>  >>  >for the purposes of searching, and could be handled automatically.
> >>  >>  >Displaying this synonym is of no use to the user.
> >>  >>  >
> >>  >>  >Cheers
> >>  >>  >Chris
> >>  >>  >
> >>  >>  >On Fri, 18 Feb 2005, John Day-Richter wrote:
> >>  >>  >
> >>  >>  >
> >>  >>  >
> >>  >>  >>J Clark wrote:
> >>  >>  >>
> >>  >>  >>
> >>  >>  >>
> >>  >>  >>>If there was a way to mark the 'obol' synonym then I could easily
> >>  >>  >>>assign the synonym type to that as I went along. Might this be a
> >>  >>  >>>possibility with DAG-Edit John?
> >>  >>  >>>
> >>  >>  >>>
> >>  >>  >>We have synonym types in OBO-Edit... just hang on! A beta-version will
> >>  >>  >>be out soon (of course, you won't be able to save your changes to the
> >>  >>  >>repository for a while after that, because everyone else will still be
> >>  >>  >>using DAG-Edit).
> >>  >>  >>
> >>  >>  >>For now, I suggest you add a synonym dbxref to the OBO synonyms (like
> >>  >>  >>"syntype:obol"). Once we move to full-fledged synonym types,
> >>we can do a
> >>  >>  >>quick search-and-replace to change those dbxrefs into synonym category
> >>  >>  >>identifiers.
> >>  >>  >>
> >  > >>  >>    -John
>
>