Electron transport mf-bp Sept 2007 onwards.

From GO Wiki
Jump to navigation Jump to search

17th September 2008

Participants: Michelle Gwinn-Giglio, Debbie Siegele, Ingrid Keseler, Jennifer Deegan.

We considered two possible ways of making mf-bp links in the photosynthesis area, as below.

Version 1

In version 1, we would have a sensu-style arrangements, where a different process term was made for each different form of the given process. The different forms of the process would be distinguished by the different sets of constituent functions.

[i]photosynthesis chloroplast-type (oxygenic)
---[has_part]this chloroplast-type function
---[has_part]that chloroplast-type function
---[has_part]the other chloroplast-type function
---[xxx]a function that occurs in both types
[i]photosynthesis in bacteria (anoxygenic) 
---[has_part]this bacterial function
---[has_part]that bacterial function
---[has_part]the other bacterial function
---[xxx]a function that occurs in both types

Caveats: halobacterium have a different set of enzymes and are archaea. Different even from the anoxygenic bacteria, so we may need an Archaea type as well.

Would it make more sense to derive this information from the annotations? Putting it together without the annotations would take a long time.

Version 2

In Version 2, a single process term would be made to cover all forms of the given process, and all the functions that make up any form of the process would be placed under that single term.

---[part_of]this chloroplast function
---[part_of]that chloroplast function
---[part_of]the other chloroplast function
---[part_of]this bacterial function
---[part_of]that bacterial function
---[part_of]the other bacterial function
---[part_of]a function that occurs in both types

Version 3

During the meeting a third option was proposed in which the GO Consortium simply provides tools that enable users to derive their own mf-bp links by mining the annotations.

If a protein is annotated to both function and process then it is an indication that the function is part of the process.
How often is the protein co-annotated? If it's a lot then that is a clue. Use the amazon model -
'people who bought that also bought this'. We can also show that a function coincides with a
process in a given set of organisms.
Action: Jennifer to ask Chris is it is possible to try this just for this process using an SQL Query.

We tried to use OBO-Edit to find the functions that would be constituents of the photosynthesis process term, but this proved to be quite difficult, even when looking under just oxidoreductase.

Action: Debbie will try to find the E.C. numbers for the functions so we can find them easily in oboedit.

General thoughts

During the discussion many people felt that version 1 was a very labour intensive plan and that it was not a feasible proposition.
Version 2 was thought to be more feasible, and version 3 certainly the simplest approach.

Conversely, it was pointed out that is useful to distinguishing between oxygenic and anoxygenic photosynthesis if this could be done. The archaeal processes may complicate this.
We need a sometimes_part_of relationship. it's the always or never part_of that is very hard to deal with.

Questions to send back:
What are the use cases?
Who wants these links?
Why do they want them?
(Ask David and Suzi which BRCs and why.)
If it was just the NIAID BRCs then they may no longer want them.
BRC update:
They are all about to have funding reviewed.
JCVI group does GO annotations.
ERIC group does too.
Biohealthbase in texas also do them.
JCVI sends in annotations but the others do not. Some groups do
not do manual annotation at all. This is unlikely to change.

Q/ Can we invite Ingrid to the consortium meeting? She is in Peter Karp's group in ecocyc.

Next meeting: How about thursdays instead of wednesday. Jen to write to list.

22nd September

Chris has checked for function terms connected to photosynthesis, and found that there are very few annotations to the function terms. http://fafner.stanford.edu/pipermail/go-fp/2008-September/000051.html

Jen has written to Tanya and Donghui to ask if they have any idea of whether it is lack of research or lack of annotation that causes this. Tanya and Donghui have both been invited to the next meeting, which is currently planned for Thursday at 4pm BST. [Tanya asked Peifen and she says that most of the experiments on photosynthetic electron transport chain were done in spinach and cyanobacteria, which is why TAIR have few annotations on this subject.]

The PIs have extended an invitation to Ingrid to join us for the consortium meeting.

An agenda item has been added to the Managers' call for the 24th September, to ask who wants the mf-bp links and what the use case is.

25th September

Participants: Michelle Gwinn-Giglio, Debbie Siegele, Ingrid Keseler, Jennifer Deegan.

We met to discuss the feedback from the Managers' call. There was scepticism about the idea that other processes would be easier. Also the view was expressed that it would be dangerous to fill in annotations in bacteria using these links, as bacterial metabolism varies so widely that these links could not possible be considered a good quide.

We loaded the OBO file of mined reactome links that Chris had provided (go/scratch/fp-links/reactome-all.obo) along with the GO live file, and looked at the has_part links.

  • The biologists were unexpectedly comfortable with the has_part relationship, though they felt that sometimes_part_of would be much more useful in this case.
  • We viewed these links in the Tree Viewer, the Ontology Tree Editor, and the Graphviz Viewer, and people understood well.
  • We had a long detailed discussion about the different kinds of part_of, and wondered what others there might be. We went to the OBO relationship ontology but did not manage to find out more there because of technical problems. We hope to return to this at the next meeting.
  • We felt that some of the function terms would be is_a children of metabolic process (on the condition that is would be okay to call these functions single step processes.)
  • We noted that there were no mf-bp links from the reactome file in the area of photosynthesis and wondered if Chris might try mining KEGG or Metacyc for photosynthesis links.

2nd October

Participants: Michelle Gwinn-Giglio, Debbie Siegele, Ingrid Keseler, Jennifer Deegan; and Chris Mungall joined us for the second hour.

We discussed things in very great depth and the conclusions are presented in the form of a powerpoint file that we are developing for the consortium meeting in a couple of weeks: http://cvsweb.geneontology.org/cgi-bin/cvsweb.cgi/go/meeting/consortium/Montreal2008/FunctionProcessLinksPhotosynthesisJDeegan.ppt See revision 1.1 for the status at the end of this meeting.

The basic conclusion to this meeting was that we could only make links between functions and single step processes, and in other cases where the part_of rule holds, because of the massive complexity of the other links and the restrictive nature of has_part.

9th October

Participants: Michelle Gwinn-Giglio, Debbie Siegele, Ingrid Keseler, Jennifer Deegan, Chris Mungall and Harold Drabkin.

The new conclusions are presented in the same powerpoint file but in revision 1.3


We discussed the same issues further and reached the conclusion that we could mine all the other links from pathway databases and instantiate them only if we can use a new sometimes_part_of link. Using this link we would not have to do the manual curation step to double check the links, as the weaker relationship would certainly be true. Chris and Michelle have agreed that they are both happy to go ahead with this plan, and we have decided to propose this as a workable solution to the consortium meeting.

Action items:

  • Jennifer to update slides.
  • Harold to come up with glycolysis examples to illustrate different forms of relationship.
  • Jennifer to make road map analogy slides.
  • Chris to make relationship slides to start off talk at consortium meeting.
  • Jennifer to contact David and explain this compromise and ask if he would also like to see this go ahead, and if he would endorse it at the meeting.

21st October

As Chris was unable to attend the Consortium meeting Jennifer Deegan delivered both his technical explanation of the function-process links and the conclusions of the working group's practical investigation. Slides

The proposal was agreed: Conclusions