Talk:Relevant for page

Great!

relevant_for up-propagates, so technically should do do this when making the slim (does the slim interface do this anyway?)

for example, biological_process is relevant for everyone

valid_for is the only-one to down-propagate. It may be controversial to create these links.

http://www.slideshare.net/cmungall/go-sensu-terms-and-taxonomy-apr-2007/

Note that mixed colors does not necessarily mean there is an error in the ontology. You can have a term relevant_for one taxon nested under a term relevant_for another. Sometimes the pattern of colors may give you a hint something is wrong, but even then it can't be guaranteed. If you want to be able to do error checking automatically you'll need specific_to (this may be more controversial, as it's hard to prove specific_to). Perhaps specific_to should be renamed only_in, or never_outside it's perhaps less confusing.

We were discussing treating relevant_for etc as annotations in their own right. But what you have done is best because you can take advantage of the existing OE subset/slim highlighting mechanism.

Hi Chris,

Yes I have been thinking through this a lot. I am working through actually making some of these links and John has given me a better structure for the file to hold the links so I will try that today.

I'm not sure I see much purpose in propagating the relevant_for links. It seem to me that the relevant_for links give us a nice clue to where curators can manually insert the specific_for and valid_for links which are much more useful. Also it's nice to have the relevant_for links be there as indicators of annotations as it allows annotation checking. Though maybe we could have a special way of indicating a relevant_for link that has come straight from an annotation rather than one that has been propagated. I have made some powerpoint slides to show at the consortium meeting. I will send them to you. They are still a work in progress, but I think the more we can hash out in advance, the better.

I am working on rigging up files to demonstrate what could be done and they are in go/scratch.

for this point:

Note that mixed colors does not necessarily mean there is an error in the ontology. You can have a term relevant_for one taxon nested under a term relevant_for another. Sometimes the pattern of colors may give you a hint something is wrong, but even then it can't be guaranteed. If you want to be able to do error checking automatically you'll need specific_to (this may be more controversial, as it's hard to prove specific_to). Perhaps specific_to should be renamed only_in, or never_outside it's perhaps less confusing.

I agree and understand that we can only have automated checking if we check relevant_for links from annotations against specific_to links. However, when I saw that mixed colouring I was able to look at it and see that the parent term must be 'specific_to' viridiplantae even though that relationship had not yet been inserted. The next step would be to actually insert the specific__to links.

For this point:

We were discussing treating relevant_for etc as annotations in their own right. But what you have done is best because you can take advantage of the existing OE subset/slim highlighting mechanism.

Yes it's okay, I do understand that. It's just that I had to mock up in this way as I didn't have a way to handle the annotation files in OBO-Edit yet. Since I did this work I have shown John and he suggested that I should have the GO and ncbi taxonomy hierarchy in OBO format and then the GO-taxon links as cross-product style intersections in a third OBO format file. I thought that sounded quite good and OBO-edit is already able to handle that. He said I should just give the stanzas with the cross-products a different namespace. I am going to try that.

I wonder if we can get the ncbi taxonomy hierarchy autogenerated in obo format? Would be nice not to have to maintain a copy of the ncbi taxonomy file manually.

Is there a protocol for keeping individual threads on a discussion page like this? I can see this could get very confusing very quickly. :-)

Thanks,

Jen