Talk:2010 GO camp downstream effect
Minutes from meeting 22nd April 2010
Rachael has put a survey on the wiki to see how much people agree at present as starting point.
How do we define downstream processes?
- indirect effects of a gp - have a effect but not directly affecting. IMP problem (phenotype observations obviously indirect)
(examples on wiki)
- when it's first characterisation of a gene it's hard to say what its direct process is. Later might be clearer. A case where defining processes with starts and ends if helpful.
- Wormbase conflicts when annotating genes with IMP to high-level terms. Example embryonic lethal for splicosomal proteins (i.e. where we know it's not really involved in embryonic dev). GO being used as phenotype annotation.
- What is the phenotype for essential gene? SGD has viable/inviable as their phenotype annotation.
- Wormbase does capture phenotype data separately to GO
- telomere maintenance annotations in SGD, lots. Many unrelated to telomeres e.g. trafficking proteins. These annotations were removed. Some proteins
Five problem areas:
- IMPs from screens
- few papers available
Ruth: heard talk where genes grouped by mouse phenotype - where GO terms over-used they become useless
Val: shouldn't ISS from IMP screens?
Karen: we used to annotate a lot more stuff to GO than we do now. Changes when you have a better view of a whole genome.
Ranjana: Wormbase do phenotype to GO term mapping for RNAi papers. Causes problems where using GO for analysis - every gene set will be shown to be enriched for embryonic development.
Not just a problem for large datasets - also the same issues for individual experiments in papers.
Do SGD remove annotations for screens when more annotations are known? Yes - added phenotype annotations and deleted GO annotations
TAIR don't go back and delete their annotations
Wormbase don't have the mechanism to remove these
Val - don't all groups go back through old annotations when a new paper is being annotated?
Not always (SGD), MGI no. RGD sometimes.
Minutes from meeting 10th May 2010
Present: Yasmin, Rachael, Pascale, Ruth, Fiona, Ranjana, Li Ni, Rama
Rachael started the meeting by saying that she will go through and discuss each point on the survey she set up:
Rachael: most groups capture what the author says, hence can include as a guideline - maybe put a condition of curator judgment and what you know about the gene - does everyone agree? but don't go overboard with downstream processes.
Ruth: Also use caution on experimental method used wherever possible
Rachael: don't use a NAS if at all possible.
Ruth: Can this be done for microarray high throughput data?
Pascale: use curator judgement
Rama: high throughput will get cleared up by new evidence code that will come into play soon.
Pascale: we shouldn't annotate to HTP?
Rama: yes you can if you are comfortable.
Rachael: these are not going to be hard and fast rules but just guidelines:
Ruth: curator judgment, what it says in paper and experimental method used
Rachael will send around documentation.
Rachael: all groups agreed that this what they would do for a newly characterized gene product but we don't all agree for removal of old annotations.
Rama: should we check for high level (parent) annotation or granular made and tell the curator that you may have a downstream process that you may want to annotate to
Pascale: this would be a good script to run : already in place in DictyBase - eg for improvement to IEP annotations.
Ruth: sometimes annotates to a parent term until she comes across amore granular term in another paper. Some dbs adding all refs for certain gp so will include broad and granular terms.
K?: in some cases may want to keep the high level terms:
Ruth: Could target it to GO slims
Fiona: at the moment annotate to response to but may need a better script
Pascale: some groups annotate all papers and some don't: discuss this further and
Rama: doesn't know what other groups are doing to capture all the annotations...maybe do a survey before or after camp
Pascale: maybe do before camp or at the camp.
Rachael: will draught a guidleine and send around
Rachael: - only annotaing to core processes As with Karens example. All people agreed and this example could be included Val had a comment - easier for groups with small proteomes. Better to annotate to a core process and not any downstream effect
Rachael:vals other example
Rama: SGD annotates to core process. eg:Nat5 n-term modification of half geneome therfore will affect many processes and genes. If there is a particular process that is modifying another then this is more specific eg: phosphorylation involved in G1 phase of cell cycle. Composite terms required - eg transcription involved in X_process.
Pascale - suggested to Rama that this should be present at camp to see what people think.
Li Ni - how do you know that it is a core process or downstream process eg: this gene is also expressed in....
Rama: sometimes you don't know - sometimes it is hard to tell
Pascale: examples are easier to follow:
Yasmin: example of PMID 20413580 ADA3 being involved in downstream effects affecting gene expression that raralpha targets, shown by ADA3 knockdown - hence annotated to the downstream effect (regulation of gene expression) by IMP:
Ruth: Statement about a balance : something involved in a core process and function: general transcription form RNA polII promoter shows where that transcription factor is placed. Maybe take into account more than in the paper: is a gene a general transcription factor therefore have parent term.
Rachael: If it has activity then there should be a core process?
Ruth: if the transcription factor is involved in more processes? Would a general transcription factor be annotated to (for example) GO:0032568 general transcription from RNA polymerase II promoter which provides a similar downstream statement as the annotation of a specific TF to 'regulation of transcription involved in X process'
Rachael: Val puts info in col16
LiNi: MGI already do this
Rachael: Vals exmple to add more specific processes to col.16 e.g. heme: this gene has a specific role in a specific downstream process: Need examples
Rachael - next point people don'tt go back to remove downstream process annotations Pombe add to their phenotype but not other groups so could suggest that people could leave in for their users
Ruth - should be a system in place to remove dowstream effects
Rachael - might be useful for users eg Yasmin and kidney annotations
Pascale - have a check system in place: Going back to replace a more granular annotations? eg kidney development initial annotation and then later a paper says it is a transcription factor - would you then remove kidney development?
Ruth: what the community need is information that is phenotype related therefore need to go to phenotype db eg: MGI: look at genes that mess up kidney development may not originally be involved in kid dev. process.
Pascale: GO captures the normal role but not the phenotype mutant info. Pascale proposes not to capture IMP and capture that info somewhere else.
Ruth: All genes involved in kid dev use MGI:phenotype db - very useful.
Ranjana: Phenotype db? Users get a sense of what the dbs are using. GO users don't necessarily know how to use the phenotype databases.
Rachael: GO has many development terms, so users will expect to be able to pull out genes related to a particular process/development by using GO, we get a lot of queries to the gohelp desk about how to find all genes involved in [X] process.
Pascale eg:homeobox tf will have same development phenotye to another gene: capture which genes are involved in eg:kidney development.
Ruth: A gene(s) overexpressed in 1 tissue that disrupts kidney development - other organs might be immune to the effects of the over-expn of this particular gene but it doesn't mean that that partcular gene is involved in kidney development. The effect on kidney development is a downstream effect.
Pascale: 3 effects on downstream effects: 1. core process - need guideline to discuss this 2. real downstream effect eg ubiquitination 3 phenotype - to use with caution - need guidelines
Ruth: would be nice to have an overall decision.
Rachael: last point: Doug says the more annotations doesn't give wrong impression but actually also gives an overall picture of the functioning of the gene.
Rama: need gene examples
Pascale: phosphate kinase example annotated - will add here
Rachael: new doodle everyone agreed to make the more specific terms
Li: add a bunch of terms to ontology eg Tf involved in X process term?
Rama: request the appropriate term on a case by case basis
Ruth: Request that if they are creating regulation of : have a check - to speed up requesting the terms Consider "regulation of transcription involve in"
Rama: this is in the process of happening
Ruth: Over-annotated insulin: gene specific regulation of transcription from RNA polymerase: should this be gene regulation in X to insulin:
Fiona: Is this a response to and not a regulation?
Ruth: any nuclear factor involved in regulation of RNA polymerase II.
Rachael: concluded that she will send around guidelines and asked that everyone fill in the recent doodle survey and then will send around another doodle poll for setting up the next meeting.
Minutes from meeting 1st June 2010
1. Definition of downstream process "For any gene product, a downstream process is any biological process which that gene product is not an essential component of or does not directly regulate, but whose action results in a series of steps leading to the positive regulation of a downstream biological process"
Karen - Not just reg, e.g. is is you mess up splicing, you mess up ribosomal proteins, you mess up translation.
Pascale – not to annotate to translation Agree not to annotate to that, but needs to include a broader view of
Ruth – need example, insulin will pos reg movement of glucose transporter
Revised definition; "For any gene product, a downstream process is any biological process which that gene product is not an essential component of or does not directly regulate, but whose action results in a series of steps leading to a downstream biological process."
2. Ligand receptor signaling pathway slide a) Is a ligand part of the signaling pathway or does it regulate the signaling pathway?
David confirmed that ligand is part of signaling pathway.
Ruth - some ligands regulate the signaling pathway, if the ligand is a rate limiting factor
Pascale – does the ligand production gps regulate the pathway, rather than the ligand reg pathway.
Ruth – so far we say
Pascale – we agree that a rate limiting factor regulates the pathway.
b) The job of an intracellular signaling molecule is to regulate transcription, which means that it will regulate a large number of cellular processes e.g. SMADs regulate transcription of specific genes when activated by TGF-beta or BMP receptor signaling. Should we then annotate the intracellular signalling molecules to the regulation of all types of downstream cellular processes, even though they take multiple signals from the multiple receptors?
Val? – people want to know what processes trigger transcription factors
Rach – do you also want to annotate to what processes the TFs go on to reg?
Yas – would not assume, would want it to be reported in the paper
Pascale – how many steps up and down stream do we want to go? Can get easily too far away from main term. Can we say only one step up and down?
Karen – specificity of effects. Mess up some TFs, you mess up lots of processes. But for some specific TFs, happier to annotate to specific processes, as it was very specific.
Rama – Request speicifc terms like TF in repose to nitrogen catbliite ? if very general just annotate to transcritiopn, if speicifc, annotate to a specific term
Ruth – I agree
Pascael – the term is transcriuption in reponse to ----- so good and specific
Rach – all good, but we have agree on this already.
Rama – nat5 acetylation of half of yeast genome. So annotate to aceyltayion, but more specific annotate to specific process.
Pascale – worried that only annotating to parent means the general GO term has two meanings. It is usedfor proteins with lots of subs and also for pros where only aceyltation is know
Ruth – need for global/general aceyltaion term?
Pascale – yes good idea
Rama – uncomfortable with idea of global ‘terms’
Ruth – MAP kinase annotate to reg of transcription of spficic genes? Or reg as part intracellular signaling pathway
Val – depends on what is talked about in paper.
Pascael – redrw signaling molecules into slide
Tim – can only annotate to what is in paper
Pascael – MAP kinase, but then paper talks aboit effect on a downstream process, tougher to annotate when have only a little info on a protein
Karen – hard to annotate when lack of knowlege
Rach – ok to annotate to downstream when only a little info, aiming to create guidelines of ideal situation of lots of info.
Ruth – RNA pol gene specufic transcrption should only be annlotated to proteins actually inv in recruiting RNA pol to DNA
Karen – Agree can be confusing when you get to far from the downstream process
Ruth – SGD annotates differently to MGI
Rach – Karen – Pascale – if something is inv in core process that is fine. If you have a mutant screen
Karen – HTP mutant screens, SGD do not annotate any longer
c) Are gene products involved in the production of ligand i) Regulate the signaling pathway that the ligand regulates/is part of?
Rachael – Do we consider the production of ligand to regulate signaling pathway
Karen – depends on whether ligand acts on multiple pathway, do we want to annotate to multiple processes?
Pascale – consider annotating to reg of pathway, although that may not always be the case?
Rach – if it regs multiple pathways?
Pascale – yes, annotate to multiple pathways in line with annot of single pathways. Maybe multiple TFs control the production of the ligand?
Karen – the production of the ligand may not be what regulates the signaling pathway, maybe some other protein sequesters it and that would regulate the signaling pathway.
ACTION: need examples
ii) regulate the downstream processes that result from the signaling pathway? Example - if the action of a gp inv in the production/secretion of insulin is measured by whether or not the insulin-signaling pathway is activated, or by whether glucose uptake is increased.
ACTION: Needs discussing by WG with aim of finding a consensus of how far downstream a gp can be said to act. Possible solution is to annotate in line with paper, so annotate to whatever the paper measures.
3. Legacy annotations a) We have agreed to always remove incorrect annotations b) Do we agree to aim for the removal of high-level NAS/TAS/IC/ISS once evidence to support more granular annotations has been annotated? c) Do we need a guideline to deal with annotations from groups that remove annotations vs those that don't for the purposes of ISS annotations?
Rama – b should be the aim across the board.
Li – not sure about ISS, but all the others MGI does automatically
Pascale – some people aim to capture all data. And keep all expt data
Rach – next one deals with this, some groups want to keep all expt data, do we need a guideline for this
Pacael – more data the better
Rama – more data in not always better, don’t want to dilute data
Pascale – increading confidence in a paper,
Rama – other issue is comprehensively annotating all possible pocrsses.
Pascale – first ev is general, seond is more specific, do you remove first one?
Karen – yes we do, as we give more info to users, SGD would remove old annots in same branch
Rach – what about when old annots are to a different branch
Rama – absoulutely
Rach – so some groups do and some groups don’t
Pascel not sure it is ok for all groups to do their own thing, need to be more formal about the decision
Rama – we should remove very downstream processes once we know that it not true
Rama – we aonly annotate to what is in the paper, we don’t extrapolate beyond that
Rach – is it reasonable to say ok to keep annots in if you have no where else to store them?
Li – most mods don’t remove old annots not a good time to try to get everyone on the same line
Pascael – good idea to get down an argument down for each side to help make descision
Summary of Action Items;
ACTION: Need example to go along with definition of Downstream Process
ACTION: We need to document what the different groups do?
ACTION: Rach – summarise why groups remove annots
ACTION: All comment on presentation slides when Rach send them round