2010 GO camp Use of Regulation issues
David Hill's slides Notes extracted from this presentation:
- annotators should annotate to the most granular term possible - in the context of understanding the biology represented in the curated paper as best as they can, and decide which term is the best.
- this isn't straightfoward; we don't understand all of biology. It is not always clear from a phenotype if a gene product is carrying out a regulatory process or is carrying out a function that is an integral part of the process.
- And the regulates relationship is hard to fully define. This is partly due to human nature: people like to show that what they are working on is important, and biology is a series of regulatory events; almost everything that happens is regulatory.
Current guidelines about using the regulation terms in GO: - Use when a gene product is modulating the rate, frequency or extent of 'X'
--- 'X' is something that happens, a process or a function
--- If 'Y' regulates 'X', then it stops it, starts it, slows it down, or speeds it up.
--- The key is defining 'X'. If we can define 'X' with respect to where it starts, stops and its parts, then it should be easy to tell if 'Y' regulates 'X'.
--- If 'Y' regulates 'X', then it stops it, starts it, slows it down, or speeds it up. Gene product 'Y' will have this effect on the beginning, end or some part of the process. Therefore it becomes important that molecular functions are linked with biological processes (all biological processes can then be thought of as a series of molecular functions; one that starts it, one or more that are in the middle, and one that stops it. If another process modulates the activity of one of those molecular functions, then it regulates the process.
- However, the definition of a process is subjective, and GO needs to reflect the community consensus about what functions are part of a process and what functions are not. With certain metabolic pathways, working out what molecular functions are core to a process will be easy. In contrast, some developmental pathways - such as metamorphosis - this will be impossible.
- In addition a gene product that is centrally involved in a pathway can also be involved in a feedback loop that regulates the same pathway Insert example here. In such a case, the enzyme in the pathway negatively_regulates the pathway and would also be part_of the pathway.
- Even for the easy biological processes - this is a huge amount of work for ontology developers - as different pathways will involve different molecular functions in different organisms, and the same molecular function will contribute to different pathways.
1. Use your biological knowledge. If it is a well-known pathway and hasn't been fully represented in GO, then background knowledge is needed to decide if the function is part of the pathway, regulates the pathway, or does both.
2. If the gene product or pathway is not fully described, then try to reflect what the paper you are reading is saying. The author should give hints about what is happening, and will be the experts in this field.
3. If there is still doubt - annotate to the less specific parent (e.g. to the biological process term instead of the regulation of BP term).
Discussion on 14-04-2010
Present: Jane Lomax, Rachael Huntley, Emily Dimmer, Jennifer Smith, Val Wood, Kimberly Van Auken, David Hill, Fiona McCarthy, Maria Costanzo, Jodi Hirschman, Varsha Khodiyar, Ruth Lovering, Pascale Gaudet. Chair: Jane
Ruth - with regards to annotation of the least granular term when the curator is uncertainly - doesn't this contradit the statement that regulation terms should not be considered as part_of a process
David - the regulation relationship is a restriction on the process term - it is wrong to use the part_of relationship between 'regulation of BP X' and 'BP X' terms, however the regulation relationship carries more information as to the aspect of the process it is involved with.
Jane - this means that it is not possible to distinguish between those gene products which are known, core components of a biological pathway, and those that are peripheral regulation members. Users will be unable to distinguish between these two categories of members of a BP.
David - this is true. It is difficult to distinguish between integral and regulation members. However to annotate to terms that describe regulation when you are not sure that contribution a gp provides is worse.
Val - however, results from a HTP screen are likely to implicate many gp's which are involved in upstream processes - I wouldn't like to annotate all of these gp's to the BP. Should we indicate if regulation is direct or if it is indirect?
David - this is the caviate that comes with analysing results from such screens. Direct annotation to BP terms are more appropriate as you can say that the gp is involved in the process, however we do not know how.
Kimberly - to confidently annotate to a regulation term, a curator may sometimes take into account the moelcular identity of the gene product - not only the phenotype. For example when looking at a range of genes that are implicated in the locomotion process, if I know which gp is a transcription factor and which is a myosin chain, I may be able to make a more informed annotation decision. E.g. Transcription factor - regulation of locomotion; Myosin - locomotion.
David - Agree. This is a central job for curators - we need to include curation knowledge when determining the correct annotation.
Ruth - when talking with a researcher - he found that when analysing a microarray dataset - the mouse phenotype data was useful in certain analyses, and the GO annotation data more useful in other circumstances. I have a concern regarding the overlap of phenotype and biological process annotation. Aren't we in danger of overcurating when we only have phenotype information?
Pascale - agree. Sometimes it is just phenotypic data we annotate to.
David - however we should represent this information from papers. It is important to annotate to some downstream effects, esp for biomedical users - who are looking at diseases. We need to be able to say if a protein .e.g a insulin receptor has effect x.
Ruth - Regulation terms are defined as 'Any process that modulates the frequency, rate or extent of a biological process....'. Shouldn't we define such terms more discretely?
David - BP terms should be primarily defined via the functions, using BP->MF links
Pascale - how much should we care at the distinction between a 'Regulation of BP' and a 'BP' term?
David - the distinction is going to be nebulous if the field is still not fully characterised.
Val - how far from the beginning of a Biological Process should you annotate to a regulation term? How far upstream should a curator go? Should direct/indirect regulation be indicated?
David - this should be possible eventually with the ontology, its difficult to accurately determine which process is direct and which is indirect. We'll get closer to this through MF->BP links
Jane - could curators contribute BP-specific MF's?
David - yes. the editors would then need to ensure these are correct.
Looked at Serenella's example (see below) - David - we have to reflect what the community defines as a process. A signalling pathway consists of multiple regulation events. We need to know where the community considers the beginning/end of the process - and then decide if a MF is part of, or regulates the pathway. In Serenella's example I would consider that the protein is a part of the pathway.
Emily - how to increase consistency, through Qcs. Could PAINT help identify where closely related organisms have been differentially annotated to regulation of BP/BP terms for the same ortholog?
Val - in adidtion the Matrix tool available from AmiGO labs could also help to look at intersections between processes.
Val - I have an example that I discussed with SGD regarding autophagy and the gene ATG1 - this gp is involved in an on/off switch - it was not regulation a process, although the authors stated it did have a regulation role. It starts the process of - it does not regulate rate, therefore would consider part of the process and not regulating it. David - we need to ensure that we have information on what is considered part of the process - if ATG1's MF is not included in that definition, then it should be annotated as regulating the process. ACTION: Val will circulate example
- Jane to send Doodle to settle on a date for the next meeting (in approx. 2 weeks time)
- Val to send ATG1 example
- Between now and the next meeting, curators in the WG should look at the examples below. And make a decision as to how they would annotate the data - using the regulation of BP or the BP term directly.
- Curators should then vote (and make short comments) using the Doodle poll that Jane will email.
- Val to send round URL for AmiGO Labs matrix and example query.
Longer term action items proposed:
- Guidance for curators who would like to suggest MF->BP links to editors, or BP-specific MFs.
- Ensure documentation described above is generally accepted.
- Annotate the documentation with examples drawn from a range of species/biological processes
- Any QC methods possible to help locate inconsistent usage of terms? (identified by PAINT/ AmiGO Labs matrix tool?)
Discussion on 05-05-2010
Kimberly Van Auken
Discussion of annotation examples
Example 1: receptor internalization vs regulation of receptor internalization
Looking at the Doodle poll (http://www.doodle.com/eyhpiduamtfum65v), a number of curators initially thought that the regulation term would be chosen, but the term definition makes that annotation difficult.
Does the regulation term include the activation step? Doesn't seem to.
Is this process (receptor internalization upon ligand binding) considered transport or signaling?
Does receptor internalization regulate a signaling pathway?
Ruth: Does ligand internalization get used similarly to a protein getting phosphorylated, i.e. a target of the process?
Pascale: My interpretation of the experiment is that receptor internalization is a way to downregulate the pathway. First the ligand binds, then internalization. Two functions of ligand: activate kinase, activate signaling pathway.
Rama: Is receptor internalization a consequence of the kinase activity?
Ruth: Receptors are recycled. Is the ligand involved in the process of receptor internalization or is it the target of internalization?
One possible conclusion: Annotate to neither term, create a new term.
Kimberly: Would ligands be annotated to both activation and down regulation?
Need to modify the receptor internalization definition. Where does the process of receptor internalization start?
Check Chris' SourceForge item.
Ruth: By analogy, should cholesterol be annotated to endocytosis?
Jane: Receptor internalization is a child of transport.
Does receptor internalization always have an effect on signaling?
Jane: are we talking about a single receptor or a population of receptors?
Pascale: We agree that the ligand shouldn't be annotated to either of the terms in the doodle poll, but are not yet clear on what new term(s) would be needed
Ruth: if you look at the ontology, you couldn't put ligand annotations under transport there's a whole machinery that effect endocytosis, regardless of what's being internalized could consult experts on endocytosis, consult Alex? is it useful to have cytokines and receptors annotated to receptor internalization terms?
Pascale: send this issue around to the GO list for more opinions.
Action Item: Rachael and Ruth will send this example out to the GO list.
Rama: is the gene product making a decision for that pathway? for Atg1 - the 'go' or 'no go' has already been conveyed the upstream gene is actually regulating the pathway
Pascale: this makes sense this may be a case where authors use regulation which can mislead curators
Rachael: look at the negative feedback data
Kimberly: the definition of the regulation terms can fit many, many phenotypes
Rama: the definitions may need some work
Pascale: this can be difficult for IMP annotations
Pascale: do we agree that Atg1 is required for the process or does it regulate the process?
Kimberly: annotation to process term fit with paper that SGD annotated
Rama: is a kinase always a regulator?
Kimberly: what are the general principles?
Pascale: be careful about using regulation terms with IMP
Rama: what is the assay, are the key players known, is this gene new/outside of the pathway David had said, when in doubt annotate to the parent term
Ruth: it'd be useful to add to the doodle pool, it'd be useful to add another row to see what conclusion we reached
Pascale: looking at the paper, authors say involved in regulation of the autophagic process
Rama: there is another paper that came later, the existing annotation reflects combined knowledge
Kimberly: should annotations be changed if the view of the pathway changes?
Emily, Pascale: annotations should reflect the most up-to-date view of a gene's role
Jane: definition is vague, should try to pin down definitions on where they start and finish
Rama: the gene product detecting the stimulus regulates the pathway
Rama, Pascale: TOR is upstream of autophagic vacuole assembly
Rama: some regulation terms are not part_of the process, they just have regulation relationships to the parent process. Are there any that are part_of the process? Rama will try to find an example.
Jane: trying to break the process down into individual steps
Pascale: look at the work the signaling group is doing - try to standardize with them?
Jane: yes, this makes sense
Pascale: review the signaling definition - very broad
Rama: looking at definition of autophagy - thought that process started when nutrient starvation is detected
Pascale: should autophagy be a child of a signaling term?
Rama: it is a child of response to stress
Jane: will need to add start and stop to as many definitions as possible
Kimberly, Rama: will be easier for molecular pathways, not so easy for more general processes like body morphology
Action Item: Rama will check negative feedback info, but it looks like the parent process annotation is the preferred annotation; SGD check the beginning of autophagic vacuole assembly
Review of current GO annotation practices
Question 1: How do groups decide on when to annotate to 'regulates process x' or 'process x'? For instance does SLIT regulate axon guidance or involved in the process of axon guidance? Val - this depends on the defined start/end of a process and somethings can be annotated to BOTH the regulates term and also directly to the process term. Some groups decide that if removing the activity of a gene product produces an all/nothing event - then they define it as being part of the process. When should annotations be inherited up the regulates relationship?
Question 2: Annotations could be checked more efficiently if GO term definitions could include the beginning and end of a process.
Question 3 (from Ruth): Should we also consider how the 'activity' of a protein is defined when making these regulation/process decisions. The definition for transcription factor activity is: The function of binding to a specific DNA sequence in order to modulate transcription.... This suggests that a transcription factor is involved in the process of regulating transcription rather than involved in the process of transcription itself. However, some of us in the UK feel that many transcription factors are involved in the transcription process itself definition: The synthesis of either RNA on a template of DNA or DNA on a template of RNA. Does the definition for transcription factor activity need modification to enable this annotation, or are only the polymerases involved in the synthesis?)
Question 3: Regulation of processes, especially with IMP;
- example 1: 
Proposed annotation policy
Examples (papers) and discussion of GO annotation issues
- Vote here for how you would annotate in each of these examples:
(from Rachael): PMID:19424712 - Ang1 is the ligand for TIE2, so it activates TIE2 which is then internalized and degraded. (Fig. 5a) HUVECs were incubated with Ang1 and the amount of Tie2 remaining at cell surface was decreased, therefore it has been internalized - question is whether Ang1 should be annotated to the process 'receptor internalization' or the regulation of? Or is GO:0007171 'activation of transmembrane receptor protein tyrosine kinase activity' enough?
(from Val) I queried with SGD why ATG1 was annotated to "autophagic vacuole assembly"rather than a "regulation of" (see 3 abstracts below). I think that the general conclusion was that the annotation was OK, because researchers considered that ATG1 was part of the autophagy process.
- The general consensus was that this is an integral part of autophagy and the arguments for this were convincing.
- This seems OK, but I still felt that something is missing. Maybe, if this is an example of where the definition of the process needs to be updated to say exactly which step is the beginning of the process. Not knowing much about this process, with these papers I know I would have annotated to "regulation of", mainly bacause it's a kinase and because of the use of "regulation" in the titles and abstracts by the authors. Abstracts:
Scott RC, et al. (2007)
Direct induction of autophagy by Atg1 inhibits cell growth and induces apoptotic cell death. Curr Biol 17(1):1-11
BACKGROUND: To survive starvation and other forms of stress, eukaryotic cells undergo a lysosomal process of cytoplasmic degradation known as autophagy. Autophagy has been implicated in a number of cellular and developmental processes, including cell-growth control and programmed cell death. However, direct evidence of a causal role for autophagy in these processes is lacking, resulting in part from the pleiotropic effects of signaling molecules such as TOR that regulate autophagy. Here, we circumvent this difficulty by directly manipulating autophagy rates in Drosophila through the autophagy-specific protein kinase Atg1. RESULTS: We find that overexpression of Atg1 is sufficient to induce high levels of autophagy, the first such demonstration among wild-type Atg proteins. In contrast to findings in yeast, induction of autophagy by Atg1 is dependent on its kinase activity. We find that cells with high levels of Atg1-induced autophagy are rapidly eliminated, demonstrating that autophagy is capable of inducing cell death. However, this cell death is caspase dependent and displays DNA fragmentation, suggesting that autophagy represents an alternative induction of apoptosis, rather than a distinct form of cell death. In addition, we demonstrate that Atg1-induced autophagy strongly inhibits cell growth and that Atg1 mutant cells have a relative growth advantage under conditions of reduced TOR signaling. Finally, we show that Atg1 expression results in negative feedback on the activity of TOR itself. CONCLUSIONS: Our results reveal a central role for Atg1 in mounting a coordinated autophagic response and demonstrate that autophagy has the capacity to induce cell death. Furthermore, this work identifies autophagy as a critical mechanism by which inhibition of TOR signaling leads to reduced cell growth.
Evolution of Atg1 function and regulation. Chan EY, Tooze SA.
The serine/threonine kinase Atg1 plays an essential role downstream of TOR for the regulation of autophagy. In yeast, where Atg1 was first identified, a complex regulatory mechanism has been described that includes at least seven other interacting proteins and a phosphorylation-dependent switch. Recent findings confirm that the mammalian Atg1 homologues ULK1 and 2 have autophagy regulatory roles. However, we and others have also demonstrated mechanistic differences with the yeast model and between these two Atg1 family members. Here, we elaborate on our growing understanding of Atg1 function, incorporating findings from yeast, C. elegans, D. melanogaster and mammalian cells. We propose that through evolution, Atg1 proteins have adopted additional cellular functions and regulatory mechanisms, which could involve multiple gene family isoforms working within multifunctional protein complexes. The gene family expansion observed in higher eukaryotes might reflect an increased functional diversity of Atg1 proteins in cell growth, differentiation and survival. PMID:19411825
The role of the Atg1/ULK1 complex in autophagy regulation. Mizushima N.
The Atg1/ULK complex plays an essential role in the initiation of autophagy: receiving signals of cellular nutrient status, recruiting downstream Atg proteins to the autophagosome formation site, and governing autophagosome formation. Recent studies of mammalian Atg1 homologs (ULK1 and ULK2) have identified several novel interacting proteins, FIP200, mAtg13, and Atg101. FIP200 and Atg101 are not conserved in Saccharomyces cerevisiae, despite the high conservation rates of other downstream Atg proteins between the yeast and mammals. Furthermore, through studies of the Atg1/ULK1 complex, the molecular mechanism by which (m)TORC1 regulates autophagy is now being clarified in detail. Copyright 2009 Elsevier Ltd. PMID:20056399
Extra info added (Ruth) Figure ATG1 and autophagy  Within the yeast cell, nearly all Atg proteins collect at the pre-autophagosomal structure (PAS) near the vacuolar membrane, and the PAS is thought to be the site of formation of the Cvt vesicle and autophagosome. Function of the Atg1/ULK1 complex in autophagy: Although the Atg1/ULK1 complex receives information regarding the cellular nutrient status from (m)TORC1, it is also essential for mTOR-independent autophagy. Lithium induces autophagy through the inhibition of inositol monophosphatase in an mTOR-independent manner..... Thus, the Atg1/ULK1 complex plays a general role in autophagy regulation. ... A systematic analysis performed in yeast showed that the Atg1 complex plays an essential role in recruiting other Atg proteins to the PAS.
[Comment from Rachael] Example 2: at the least I would use regulation of autophagy because of Abstract 1 stating that Atg1 negatively feedsback to TOR thereby regulating autophagy. However, as Val points out I think the definition should be updated as to the start of the process. Since the authors use the phrase 'initiation of autophagy' quite a lot, I would side on 'regulation of autophagy' from what is in David's slides - 'If ‘Y’ regulates ‘X’, then it stops it, starts it, slows it down, or speeds it up.'. However, this is not clear in the definition of some regulation terms e.g. regulation of autophagy 'Any process that modulates the frequency, rate or extent of autophagy...' - no mention of starting or stopping.
- Protein: Q9Y696 human CLIC4
Upregulation of CLIC4 is required for Ca2+-induced keratinocyte differentiation
To determine if upregulation of CLIC4 is essential for Ca2+-induced differentiation in keratinocytes, we reduced CLIC4 level by expressing CLIC4 antisense in an adenoviral vector (AS-CLIC4 Ad) in keratinocytes exposed to 0.5 mM (Hi) Ca2+ medium. We have previously shown that this vector reduces CLIC4 expression in a dose-dependent manner (Suh et al., 2005 ). Under conditions in which AS-CLIC4 adenovirus (Ad) reduced CLIC4 and prevented the Ca2+-induced increase in CLIC4 expression, K1 and K10 proteins were not elevated at 24 hours after Ca2+ addition, and the expression of the late differentiation marker filaggrin was inhibited after 48 hours (Fig. 3A). Expression of AS-CLIC4 did not alter the endogenous level of PKC (Fig. 3A), suggesting that CLIC4 upregulation is an important component of the differentiation response to elevated Ca2+. To verify this observation by an independent differentiation marker, cell cycle parameters were monitored to determine if the rapid G1 cell cycle arrest associated with Ca2+-induced differentiation, was altered (Dotto, 1999 ). Within 24 hours of 0.5 mM Ca2+ exposure, BrdU pulse-labeled cells were absent when assayed by flow cytometry whereas the number in G0-G1 increased substantially (Fig. 3B,C). However, these changes were prevented by expression of CLIC4 antisense, and the proliferating population was maintained up to 48 hours as in the 0.05 mM Ca2+ medium (Fig. 3C). To verify CLIC4 is a specific CLIC family member that mediates the keratinocyte differentiation response, non-specific (NS) and specific shRNA constructs were developed for CLIC1 (sR-1), CLIC4 (sR-4) and CLIC5 (sR-5) and used for a transient knockdown of CLIC members (Fig. 3D,E). When transfected cells are induced to differentiate by 0.5 mM Ca2+, CLIC4 shRNA prevented upregulation of K10 but the other shRNAs did not. This result also distinguishes CLIC4 from other CLIC proteins that might have been influenced by AS-CLIC4 expression (Suh et al., 2005).
Which GO term should be associated with Q9QYB1 mouse Clic4:
- 'GO:0030216 ; keratinocyte differentiation' IMP
- GO:0045618 positive regulation of keratinocyte differentiation IMP?
The term GO:0071277 cellular response to calcium ion was also associated with this gene
- David said that 'if another process modulates the activity of one of those molecular functions, then it regulates the process'.
- This makes me wonder what if I have a protein A which is a regulatory subunit annotated as 'molecular function regulator' .
- Lets say that the 'molecular function' is part of a 'Biological process'.
- Should this protein A be annotated as 'Biological process regulation' or 'Biological process'?
- P85A_HUMAN (P27986)
- GO; GO:0035014; phosphoinositide 3-kinase regulator activity.
- DR GO; GO:0060396; growth hormone receptor signaling pathway.
- DR GO; GO:0008286; insulin receptor signaling pathway.
- DR GO; GO:0048009; insulin-like growth factor receptor signaling pathway.
- DR GO; GO:0014065; phosphoinositide 3-kinase cascade.
- DR GO; GO:0046854; phosphoinositide phosphorylation.
Suggestions for Quality Control procedures
Back to 2010_GO_camp_Meeting_Agenda