ValWood
commented
7 years ago I think this deserved wider discussion at the GO meeting. Up to now there have been no guidelines on how to do this...it seems quite arbitrary when a term is accepted or refused, but we need to make annotations consistently.
In case its useful, this is what I wrote before ANtonia created this ticket:
We extend the “transcription factor activity” term in 3 main ways i) The gene(s) regulated by the transcription ii) The response to a specific environmental change iii) If this transcription directly regulates a specific process (NOTE: we only do this when the transcription factor is DIRECTLY CONTROLLING THE PATHWAY (i.e. for metabolic processes the pathway activation is often transcriptional. We would not link to regulation instances where the TF is only ensuring that a gene product present is present at the right time or is constitutively expressed, and the pathway the gene product is involved in are controlled post-transcriptionally in a “NORMAL” cell (e.g most genes regulating cell cycle transitions). This allows us to accurately represent cases where the regulation of the level of a specific gene product is directly regulating a specific process). However, to represent the information fully we often need to link annotations provided for i) ii) and iii);
There are 2 ways to capture the information fully:
- The make independent annotations and link them using extensions
- To request fully precomposed terms.
Historically we have done this by requesting a composite process term to link to, and would represent, for example:
MF GO:0001078 has regulation target gene y, in_response_to iron ion, part_of “negative regulation of transcription involved in negative regulation ferrichrome biosynthesis” MF GO:0001078 has regulation target gene x, in_response_to iron ion, part_of “negative regulation of transcription involved in negatively regulation of ornithine metabolism”
So in Antonias’s example:
transcriptional repressor activity, RNA polymerase II core promoter proximal region sequence-specific binding part_of negative regulation of transcription from RNA polymerase II promoter in response to iron (existing process)
this does not tell the full story so Antonia requested
Option1
negative regulation of ferrichrome biosynthetic process by negative regulation of transcription from RNA polymerase II promoter in response to iron
because it seemed neater than post-composing:
Option2
transcriptional repressor activity, RNA polymerase II core promoter proximal region sequence-specific binding part_of negative regulation of transcription from RNA polymerase II promoter in response to iron (existing process),negative regulation of transcription from RNA polymerase II promoter involved in negative regulation of ferrichrome biosynthetic process
option1 was vetoed. However, although clunky, but it does follow a previous precedent based on existing terms, and it captures everything completely.
So, if we don’t have these terms we need alternative guidelines of how to do this compliantly.
Option3 A proposal:
MF GO:0001078 transcriptional repressor activity, RNA polymerase II core promoter proximal region sequence-specific binding has_input sib1 part_of (GONEW negative regulation of ferrichrome biosynthetic process, in response to iron) (using a compound extension) has_input sib2 part_of (GONEW negative regulation of ferrichrome biosynthetic process, in response to iron) (using a compound extension)
This captures everything because the MF GO:0001078 already has an ontology encoded F-P link to GO:0000122 negative regulation of transcription from RNA polymerase II promoter
And a biologist reading this page would understand the statement without any loss of information, because the transcriptional repressor is inherently coupled to regulation of transcription from RNA polymerase II in their head (no need to repeat it in the extension, or a precomposed term explicitly).
Thus a DNA binding Transcription Factor like “GO:0001078 transcriptional repressor activity, RNA polymerase II core promoter proximal region sequence-specific binding” would end up with a bunch of Biological Process annotations which included:
- “negative regulation of transcription from RNA polymerase II promoter”
- “negative regulation of process blah” for any process directly transcriptionally regulated
paolaroncaglia
ukemi
vanaukenk
pgaudet
Antonialock
srengel
RLovering
So, Val and I think some terms need to be obsoleted in the ontology.
We started talking about it when I requested a long precomposed BP to describe fep1. Fep1 is a DNA binding transcriptional repressor that suppresses transcription of iron starvation genes when there is plenty of iron around. I requested: "negative regulation of ferrichrome biosynthetic process by negative regulation of transcription from RNA polymerase II promoter in response to iron"
Melanie told me that the term is too specific and I could either annotate to "unconnected" process terms, or string them together using annotation extensions. I definitely want to nest the terms because they "belong" together.
Val then pointed out to me that I was trying to incorporate the function of the enzyme in my process term - and I could make the term a lot less nasty by taking the "MF part" out. We connect MFs to BPs using extensions, and I shouldn't treat the transcription factors any differently.
This is an example annotation for one of our kinases:
The kinase is also annotated to these biological processes as separate annotations, but here they have been connected to the MF.
So, instead of annotating fep1 to "negative regulation of ferrichrome biosynthetic process by negative regulation of transcription from RNA polymerase II promoter in response to iron"
I should annotate it to the MF transcriptional repressor activity, RNA polymerase II core promoter proximal region sequence-specific binding has_input gene:sib1 part_of negative regulation of ferrichrome biosynthetic process in response to iron
and to the BP negative regulation of ferrichrome biosynthetic process in response to iron
I'm not focussing on transcripton factors at the moment, so it is very confusing that there are terms in the ontology that we "shouldn't" annotate to (if we can't request new sibling terms then I suppose we shouldn't really annotate to the existing terms anyway right?)
So, I'm asking if you could obsolete all the transcription regulatory terms that we "shouldn't" annotate to (and for which you won't create new sibling terms) - many are descendants of GO:0006357
@ValWood @mcourtot