Epistemic formalism (was Re: [Phenoscape] Re: [go] evidence code ontology)

[Jim Hu]

On Feb 7, 2008, at 7:33 AM, David Hill wrote:

> Tanya Berardini wrote:
<snip>
>
>>
>>
>> I think this is a great example of curation as an art opposed to a
>> science.  It's easy to make hard and fast rules that are just waiting
>> to be broken:  "If you don't see it in a Northern, it's not there."
>> vs. "If you don't see it in a Northern, it could be there but you  
>> need
>> more tissue/more mRNA/longer exposure/more sensitive probe...."
>> Personally, I annotate the positives in Northerns and tend to stay
>> away from the negatives (don't make the NOT annoation) because, you
>> never know.   On the other hand, if the transcript isn't detected by
>> Q- or SQ-RT-PCR, I would make that NOT annotation.
>> Others may differ in that practice.
>>

It seems to me that when an author is making a strong assertion that  
something is not present and the reviewers accept it, then the curator  
should use it if its relevant to the annotation.  But I'm trying to  
figure out where a good not expressed would lead to a NOT annotation  
in GO.  The only example I'm coming up with would be comparing the  
Northern for gene Y in mutant vs wt for gene X, when you're annotating  
gene X as regulation of whatever Y does.  I can see using a negative  
Northern or Q-PCR to annotate to negative regulation of X process.   
But I'm missing how one would use a negative RNA measurement to assert  
a NOT in the absence of a different experiment that gives an  
affirmative assertion to the term you're attaching the NOT to.  No one  
would annotate NOT to regulation of all the things that don't change  
in a microarray, right?


>>
>
> I think Tanya really hits the nail on the head here. I don't think  
> any of us exhaustively annotate NOT. If the author seems to feel  
> that it's an important conclusion (maybe it was totally unexpected)  
> then we annotate with a NOT annotation. The idea behind it in it's  
> original context was to give people a 'heads up' that on the surface  
> we may have a gene product where it seems like it should be given a  
> GO annotation, but in fact it has been shown that this is not the  
> case. If my memory serves me correctly the original discussion of  
> NOT was for something like a peptidase that had great sequence  
> similarity to other known peptidases, but was shown to lack a key  
> residue and was therefore catalytically inactive.

Or in a real case that I think we'll be annotating at some point.   
Publication A claims that Protein X is a nuclease and this IDA  
annotation is used to transfer the GO association via ISS to other  
genomes.  A later publication, B, shows that X is not a nuclease; the  
authors of B have purified X more than the authors of A, and show that  
the nuclease activity is a contaminant.  In this kind of instance, I  
think it is not enough to drop annotations from A from the list of  
annotations made, because the annotation creep is likely to continue.   
I think that the MOD wants to make a proactive assertion of NOT as a  
cry of "Please correct those ISS annotations".  In the absence of a  
NOT, the absence of a nuclease annotation might be misinterpreted as  
the MOD hasn't got to that paper yet.

As I think I mentioned in another thread, this kind of thing is a  
special concern for us with E. coli based on the number of different  
groups doing GO on our genes, and also due to our quixotic attempt to  
make GO annotation something that can be done in the community.

Jim
=====================================
Jim Hu
Associate Professor
Dept. of Biochemistry and Biophysics
2128 TAMU
Texas A&M Univ.
College Station, TX 77843-2128
979-862-4054


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://fafner.stanford.edu/pipermail/go/attachments/20080207/070665a9/attachment.html