I think this is a good idea because the proposal addresses the Evidence Code Problem (TM) well. <br><br>
>1) When separate experiments are done in a paper that link a function
and a process, we capture the individual functional >data and process
data by annotating the individual experiments. Then, we make an
additional annotation to a <br>>'process- specific function' using the EXP
evidence code. In these annotations EXP would indicate that there were
>several types of experimental evidence in the paper that led to the
final conclusion.<br><br>As long as the individual MF and BP annotations are made with one of the experimental codes, it is completely consistent to use the EXP code for the 'function-specific process' annotation.<br>
<br>IDA + IDA = EXP<br>IDA + IMP = EXP<br><br>Example:<br><br>DNA binding IDA<br>regulation of transcription IDA<br>DNA binding involved in regulation of transcription (aka transcription factor activity) EXP<br><br><br><br>
>2) When the function of a molecule is very well known and the authors
do an experiment in a paper that indicates a >process. If it is a
well-established fact that the molecule only performs a certain
function and the authors clearly indicate >the function-process link, we
make a the process annotation in the usual way, and then make the
'process-specific >function' annotation using the IC evidence code with
the molecular function in the 'with' field.<br><div class="gmail_quote"><br>Annotations made in this fashion make explicit what is contained in the graph. <br><br>Example:<br><br>protein kinase activity IDA<br>protein amino acid phosphorylation IC, evidence_with =GO:0004672 (protein kinase activity)<br>
<br>At the meeting, we said that if we made the F-P link between 'protein kinase activity' and 'protein amino acid phosphorylation' in the graph, then anyone looking at annotations to the latter term using a browser that took into account the graph structure of GO would get any gene products annotated to 'protein kinase activity'. The concern was raised over the annotation not being EXPLICIT and not present in the gene association file if the BP annotation was not made. <br>
<br>David's proposal addresses this concern and makes use of the IC evidence code in a manner consistent with the current guidelines for using that code.<br><br>My two cents,<br><br>Tanya<br><br>On Fri, Apr 10, 2009 at 4:25 PM, David Hill <span dir="ltr"><<a href="mailto:dph@informatics.jax.org" target="_blank">dph@informatics.jax.org</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Hi Everyone,<br>
<br>
At the GOC meeting in Eugene, Tanya presented a proposal to create new 'process-specific' function terms so that we can create part_of links between the MF and BP ontologies. We decided to go ahead with this proposal, but in the discussion afterwards, there was some concern about how annotators would use these terms. Specifically, it was brought up that often a single experiment does not give evidence that a gene product has a specific molecular function AND a role in a specific biological process. Another case was brought up that the molecular function of some gene products are very well characterized and in most cases when a researcher studies the process that the gene product is involved in, it is a given that the product is performing its well-known function.<br>
<br>
It seems to me that in these cases and from the discussion in Eugene, we are not able to represent the conclusions from these papers not because the authors don't make strong cases for function-process information, but because we are limited to making separate annotations for each experiment and are not able to represent the collective data and keep with our evidence code paradigm. I think it's unfortunate that we can't capture these types of data, when they often represent the detail of knowledge that is a 'given' in the field. I propose that we begin to capture these data in the following way:<br>
<br>
1) When separate experiments are done in a paper that link a function and a process, we capture the individual functional data and process data by annotating the individual experiments. Then, we make an additional annotation to a 'process-specific function' using the EXP evidence code. In these annotations EXP would indicate that there were several types of experimental evidence in the paper that led to the final conclusion.<br>
<br>
2) When the function of a molecule is very well known and the authors do an experiment in a paper that indicates a process. If it is a well-established fact that the molecule only performs a certain function and the authors clearly indicate the function-process link, we make a the process annotation in the usual way, and then make the 'process-specific function' annotation using the IC evidence code with the molecular function in the 'with' field.<br>
<br>
I think that this might get around some of the frustration that annotators were feeling at the meeting.<br>
<br>
Fire away!<br>
<br>
David<br>
<br>
-- <br>
David P. Hill, Ph.D.<br>
Bioinformatics Scientist: Ontology Development<br>
Gene Ontology Consortium<br>
The Jackson Laboratory<br>
<a href="http://www.geneontology.org" target="_blank">www.geneontology.org</a><br>
<a href="http://www.informatics.jax.org" target="_blank">www.informatics.jax.org</a><br>
tel:207-288-6430<br>
<br>
_______________________________________________<br>
Annotation mailing list<br>
<a href="mailto:Annotation@geneontology.org" target="_blank">Annotation@geneontology.org</a><br>
<a href="http://fafner.stanford.edu/mailman/listinfo/annotation" target="_blank">http://fafner.stanford.edu/mailman/listinfo/annotation</a><br>
</blockquote></div><br><br clear="all"><br>-- <br>Tanya Berardini<br>TAIR Curator<br><a href="http://www.arabidopsis.org" target="_blank">www.arabidopsis.org</a><br>