View source for Annotation Conf. Call, July 28, 2015

[[Category: Annotation Working Group]]

==Agenda==

===Overview by ontology editors===

* Nothing has been obsoleted, even if QuickGO webservice tests and their usage by Protein2GO have given that impression. The details of these tests and how they are used by protein2GO and GOA have [only just been documented](https://github.com/geneontology/annotation_extensions/wiki#checking-pipelines). We need to work with Tony S to provide a mechanism for suspending usage of potentially problematic relations pending review.

* Annotation extension relationships are meant to extend the meaning of a GO term to create a new, more restricted class. Where the relation in question is used in GO, or is related logically to one that is, OWL translation of annotation extensions can be used to group annotations more accurately and potentially in useful novel ways (for example, grouping annotations to process term by the cell-type in which they occur). A small number of relations appear instead to qualify when the relationship between a GP and a GO term applies, (e.g. dependent_on, independent_of, in_presence_of, in_absence_of).  OWL translation/interpretation of the resulting annotation extensions is problematic.  

From reviewing usage of these relations, some are used with extremely variable semantics it seems there may be other more exotic uses (see review of [dependent_on usage](https://github.com/geneontology/annotation_extensions/issues/17).)

Given these issues, it is important to review usage of these relations to discuss what we are trying to achieve with respect to whether/how the annotations can be integrated into tools that support our users needs.

AI:  Make ticket for AMIGO to display (go-plus) relations plus usage in GO.   Links to these pages can be added from the GO usage docs on GitHub.
   DONE: https://github.com/geneontology/amigo/issues/221

  RAMA notes that those not using protein2go may not be getting warnings about incorrect usage due to violations of local domain and range or subset constraints. 
  AI: Rama to check with Tony about who is getting reports of failure to load into QuickGO (all same checks are used there).

===Example papers===
=== 1. Obsoleted relationship:localization_dependent_on ===
<br />
PMID: 22842922<br />
DSF2 relocalizes from budneck to cytoplasm upon DNA replication stress<br>
We would capture this as: cytoplasm, localization_dependent_on: GO:0006974 (cellular response to DNA damage stimulus) or<br>
cytoplasm, in_presence_of: CHEBI: MMS/Hydroxyurea<br>
(there are few papers in yeast where they have show localization changes in response to certain stimuli)

dependent_on:  "Identifies a requirement under which the gene product is located at the specified cellular component." <br>
exists_during (no def in gorel.obo): Identifies a process or stage during which a cellular component is present. 

'''[Response from Rachael]''' Here I would say cytoplasm exists_during cellular response to DNA damage stimulus

'''[From discussion:]'''  The combination of GO term plus AE needs to make sense as a class on its own - ignoring the gene product. The result should be a subclass of the primary GO term used.  In this case, the extension should define some subclass of cytoplasm.  The original annotation doesn't make sense when parsed in this way:  "cytoplasm that localization dependend on cellular response to DNA damage stimulus" is a nonsense class and doesn't match the original intention: localization of DFS2 to the cytoplasms depends on cellular response to DNA damage stimulus'.  Rachael's annotation does make sense when parsed in this way:  The subclass of cytoplasm that (exists) during 'cellular response to DNA process'.  

The exists_during may be causing some confusion here though - some cytoplasms exists before and after this process.  So perhaps we should try to rename the relation.  'during' might work better.  This was previously used as a more general relation (covering happens_during), but we could use it with the more restricted meaning here (only applying to continuants (CCs).

=== 2. Obsoleted relationship:requires_substance ===
<br />
PMID: 11006294<br>
ISC1 has inositol phosphosphingolipid phospholipase activity (http://amigo.geneontology.org/amigo/term/GO:0052712) which is dependent on phosphatidylserine. We captured it with 'requires_substace' (in_presence_of would work too). Both these relations are slated for obsoletion. The definition of the term doesn't include phosphatidylserine as a required cofactor.

requires_substace: The annotated gene product participates in a Biological Process or executes a Molecular Function only in the presence of a substance" <br>
in_presence_of: "Identifies a chemical, gene product or complex in the presence of which an ontology term is observed to apply to the annotated gene product."

'''[Response from Rachael]''' Phosphatidylserine is the cofactor here, should we be capturing cofactors? If so, how? Would activated_by be the correct relation (Identifies a chemical substance that increases the activity of the gene product)?
'''Ruth:''' could 'has_agent' be used here?

=== 3. Obsoleted relationship:independent_of === 
'''independent_of:''' "Identifies an entity that was shown not to be required for the ontology term to apply to the annotated gene product."
<br />
PMID:25215947<br />
Typically mitophagy is dependent on PARK2 / parkin (Uniprot O60260). However, in PMID:25215947, authors show that AMBRA1 (Q9C0C7)  (in presence of chemical inducers of mitochondrial depolarisation such as FCCP) initiates mitophagy independent of PARK2 / parkin.
What to do now that independent_of relation is obsoleted? Simplest to create new term instead? e.g. “parkin-independent mitophagy in response to mitochondrial depolarization”. <br />
NB Examples of parkin-independent mitophagy:<br />
PMID: 25349190 "Parkin-independent mitophagy requires Drp1 and maintains the integrity of mammalian heart and brain."<br />
PMID: 24176932 "Loss of iron triggers PINK1/Parkin-independent mitophagy."<br />
Paul Denny, UCL Functional Annotation Group.
<br /><br />

'''[UCL discussion]'''In order to not create 'reverse annotations' (as agreed in Bar Harbor), UCL agreed to request the GO term parkin-independent mitophagy. 
* This would not be a suitable approach for all cases, e.g. not just to capture a specific chemical. 
* The new terms would only be created if it has been demonstrated that a major pathway exists that is different from the typical/common pathway, e.g. equivalent to canonical and non-canonical wnt signaling. 
* We should not be trying to capture assay conditions, curators should try to understand the role of the chemical/protein in the physiological process. 
* If the entity is a gene product, curators should annotate this in the forwards direction, i.e. use the gene product as the primary entity being annotated to the process. See example 4 below for when we would suggest it is appropriate to capture chemical dependency within a GO term.

DOS: I'm happy to create a class for Parkin independent mitophagy

=== 4. Obsoleted relationship:independent_of === 
<br />
PMID: 12482963<br />
ISW1a complex binds to nucleosome independent of ATP. We annotated to the subunits of ISW1a to contirbutes_to nucleosome binding and included the ATP independent detail in col-16. <br>

'''[Response from Rachael]''' Could we have a new term ATP-independent nucleosome binding? We have several ATP-dependent terms, suggesting there are ATP-independent equivalents?

=== 5. Several examples here for Obsoleted relationship:dependent_on ===

'''dependent_on''': Identifies a requirement under which the ontology term is observed to apply to the annotated gene product."
===== i) PMID: 12482963 =====

<br />
Both Isw1a and Isw1b complexes show nucleosome-dependent ATPase activity. How do we capture the nucleosome dependency part. Both dependent_on and in_presence_of are slated for obsoletion.

'''[Response from Rachael]''' I'm not sure I would capture this. The complexes interact with nucleosomes to restructure the chromatin, so I think there are other terms you can use to describe this. Just skimming the paper, there are several annotations you could make (note: I haven't read in detail, so these are just thoughts depending on the evidence);

a. nucleosome binding part_of ATP-dependent chromatin remodeling (or part_of nucleosome positioning)

b. ATPase activity part_of nucleosome positioning

c. ATPase activity part_of nucleosome mobilization

===== ii) examples on the wiki how to use the dependent_on relationship. =====
These examples were created to guide curators, if the relationship is being obsoleted it is important to give guidance on how this information can be captured, or state that it is out of the scope of GO. <br/>

'''IAP2-dependent regulation of NF-kB activity:'''<br/>
Statement from paper:

To test whether RIP3 and RIP4 have to be ubiquitinated by cIAP1/2 in order to mediate NF-kB activation, we compared RIP-mediated NF-kB luciferase reporter activity when ectopically expressed in HEK293T cells in the presence or absence of the IAP inhibitor BV6, a treatment that induces rapid auto-ubiquitination and degradation of endogenous cIAP1/2[41]. As shown in Figure 4A, BV6 treatment greatly impaired TNF and RIP1 RIP4-induced NF-kB activation but had no impact on TAK1-mediated NF-kB induction (Figure 4A). Those results, which indicate that cIAP1/2 act upstream of TAK1, are consistent with a role for cIAP1/2 as E3 ligases regulating RIP14-mediated activation of NF-kB.

Expressed in LEGO: http://go-genkisugi.rhcloud.com/seed/model/gomodel:taxon_9606-5408ded30000003

'''''previous recommendation''''' RIP3: GO:0051092 positive regulation of NF-kappaB transcription factor activity: PMID:21931591: dependent_on(UniProtKB:Q13490 IAP2)

Ruth comment: this highlights the problem of consistent annotation, the annotation should actually have listed not just IAP2 but also TNF! And in theory could list all proteins involved in the signaling pathway upstream of RIP3. What is being captured is the idea that it is unexpected that IAP2 is needed for regulation of NF-kappaB.
The IAP2 protein was not annotated in the forwards direction as having any involvement in the regulation of NF-kappaB nor in the regulation of the kinase activity nor an involvement in the cellular response to TNF. 
An improved annotation would be: 
IAP2: positive regulation of protein K63-linked ubiquitination: IMP: PMID:21931591: part_of cellular response to TNF, causually_upstream_of positive regulation of protein serine/threonine kinase activity, has direct input RIP3 <br/>
IAP2: ubiquitin-protein transferase activity: IMP: PMID:21931591: part_of cellular response to TNF, causually_upstream_of positive regulation of protein serine/threonine kinase activity, has direct input RIP3 <br/>

Also 
RIP3: GO:0051092 positive regulation of NF-kappaB transcription factor activity: PMID:21931591: part_of cellular response to TNF

These annotations do not however capture that it is the ubiquitinated RIP3 which is active. But should GO capture impact of protein modifications (eg phosphorylation, ubiquitination, glycosylation) on activity?

'''A protein's function is required for its involvement in a process:'''

Statement from paper:

ACTEIII/PTE-1 mutants with reduced enzymatic activity or with the defect in peroxisome localization did not induce peroxisome proliferation

Expressed in LEGO: http://go-genkisugi.rhcloud.com/seed/model/gomodel:taxon_9606-5437882f0000021

'''''previous recommendation''''' ACOT8: peroxisome fission: PMID:15194431: IDA: dependent_on(GO:0047617 acyl-CoA hydrolase activity)

Ruth: again this should be a forwards annotation <br/>
ACOT8: acyl-CoA hydrolase activity: PMID:15194431: IDA: part_of peroxisome fission

'''Two proteins required for an activity'''

Statement from paper:

To test the RuvA and RuvB proteins for DNA helicase activity, we used a simple gel electrophoretic assay which measures the displacement of a short 32P-labeled oligonucleotide (66nt long) from single-stranded circular omegaX174 DNA. Purified RuvA and RuvB catalyzed the unwinding of the oligonucleotide from the single-stranded circle (Fig.1,lane e). Neither RuvA (lanes c and d) nor RuvB (lanes f and g) alone were capable of unwinding, even at much higher protein concentrations.
'''''previous recommendation''''' 
RuvA: ATP-dependent DNA helicase activity: PMID:8433990: IDA: dependent_on(RuvB)<br/>
RuvB: ATP-dependent DNA helicase activity: PMID:8433990: IDA: dependent_on(RuvA)<br/>

Ruth: what has happened to the use of contributes_to and could this use IGI? <br/>
RuvA: contributes_to: ATP-dependent DNA helicase activity: PMID:8433990: IGI: RuvB<br/>
RuvB: contributes_to: ATP-dependent DNA helicase activity: PMID:8433990: IGI: RuvA<br/>
I think the use/creation of a specific complex term would enable all of this to be captured
But how can curators capture that 2 or more proteins form a complex that is necessary for a specific function, is contributes to and IGI sufficient, or should IPI be used to demonstrate that the activity requires the other protein to be within the complex?

=== 6. in_absence_of ===
'''absence_of:''' Identifies a chemical, gene product or complex in the absence of which an ontology term is observed to apply to the annotated gene product." [GOC:rph]
<br />
Q75JZ1 gtaC<br />
annotation: nucleus '''in_absence_of''' cAMP<br />
PMID:24653039<br />
Title of paper: Nucleocytoplasmic Shuttling of a GATA Transcription Factor Functions as a Development Timer. 
“We report here that GtaC exhibits rapid nucleocytoplasmic shuttling in response to periodic cAMP waves.” p1329
<br />
COMMENT: This was annotated in conjunction with a current ‘cytosol in_presence _of cAMP’ annotation. No other annotations would describe the shuttling as well as these two annotations together and in further conjunction with the annotation: ''regulation of aggregation involved in sorocarp development dependent_on cellular protein localization''<br />
NOTE: we already deleted this annotation as it was flagged in P2GO, but I really disliked deleting that one.
<br /><br />


'''[Response from Rachael]''' I would capture this annotation as nucleus exists_during 'cellular response to cAMP' (or even 'cAMP-mediated signaling'?) and a second annotation to cytosol exists_during 'cellular response to cAMP'. For your second annotation, I believe what this is interpreted as is: gtaC is involved in regulation of aggregation involved in sorocarp development and this is dependent on the process of cellular protein localization - I don't think this is what you intended? I think what you are trying to say is the regulation of aggregation... is dependent on where gtaC is localized? If so, I would capture this as gtaC is involved in regulation of aggregation... occurs_in cytosol or nucleus (whichever is applicable).

'''[Response from DOS]''' Framed as classes of GO term =>  "The class of all nuclei in the absence of cAMP'.  Besides being a very oddly contingent class, its also very ambiguous.  The absence of cAMP from where.  The nucleus?  The cytoplasm?  The environment of the cell?  Rachael's is better in that it makes this clear we're talking about during response to cAMP.

=== 7. examples for in_presence_of ===
'''in_presence_of:''' "Identifies a chemical, gene product or complex in the presence of which an ontology term is observed to apply to the annotated gene product

===== in_presence_of example Q550R2 ctxB ===== 
<br /> 

annotation: positive regulation of gene expression has_regulation_target carA, acaA, rasG,rasC, '''in_presence_of''' cAMP<br />
PMID:22114350<br />
FIGURE 6: Inhibition of cAMP-activation of Ras in ctxA−/B− cells. (A) cAMP activation of RasC and G. <br />
FIGURE 5: Cortexillin-null cells have delayed and diminished expression of cAR1 and ACA. Suspensions of WT and cortexillin-null cells in starvation buffer were pulsed with cAMP…. ctxA− and ctxB− cells had delayed and reduced expression of ACA and cAR1, and ctxA−/B− cells had almost no expression of either ACA or cAR1.<br /> 
COMMENT: Even though in Fig. 6 authors mention ‘cAMP activation, I like in_presence_of’ as it implies the correct thing; dependent_on implies that the regulation of the gene expression is dependent of cAMP, but it’s really dependent on the mutated gene(s) shown in this study. The addition of cAMP is mimicking natural starvation induced behaviour and in the assay they used cAMP. A more interpretive annotation would be:<br />
''positive regulation of gene expression has_regulation_target carA, acaA, rasG,rasC, happens_during response to starvation''. Here the user must know what happens during starvation and it is more my interpretation than annotating what actually happens where users can interpret.
<br /><br />

'''[Response from Rachael]''' If the authors are studying starvation but using cAMP to mimic this you should really add the starvation term to the extension, as this is the physiological process. In this case I would annotate to positive regulation of gene expression has_regulation_target x,y,z (,) part_of cellular response to starvation. This is saying (I think!) that as a result of starvation, ctxB regulates the expression of x,y,z. I'm not sure if there is an easy way to determine when to use part_of and when happens_during, but I would expect some regulation of gene expression to occur as a result of starvation, therefore I would say reg. of gene expression is part of starvation, rather than it occurs at the same time as starvation (in which case it would be happens_during).

'''Ruth:''' I would consider the role of the genes being regulated. Did the authors look at these genes because of their known/expected role in the cell's response to starvation? If these genes play a role in what the cell does in starved conditions then use 'part_of', if the genes are not known to have any role in starvation conditions,  then use 'happens_during'.

===== in_presence_of example Q54BD4 dstC =====
<br />
annotation: response to cation stress happens_during sorocarp development '''in_presence_of''' potassium chloride<br />
PMID: 22944283<br />
There are four such annotations made for all 4 tested stress factors that resulted in developmental defects.<br />
p.793: we investigated the development under stress of the dstC− (STATc) mutant. The dstC− mutant showed severe defects with fewer and/or smaller fruiting bodies under all of the stressing conditions (KCl, NaCl, MgCl2, LiCl)
FIGURE 6. TacA is involved in the stress response during development. <br />
COMMENT: Here it feels not appropriate to replace in_presence_of with ‘dependent_on’ and I would have lots of gripes against using ‘has_input’. Not clear what ‘has_agent’ means, never used that and wouldn’t like it either in this context.<br />
Petra Fey, dictyBase
<br /><br />

'''[Response from Rachael]''' Here it sounds like you would want a new term 'response to potassium cation stress', in which case I think you should use has_input so that the OWL interpretation of a new term would be correct - but an editor should confirm this.

==== 9&10. examples on the wiki how to use the in_presence_of relationship. ====
These examples were created to guide curators, if the relationship is being obsoleted it is important to give guidance on how this information can be captured, or state that it is out of the scope of GO. <br/>

===== 9. Catabolism of hyaluronic acid (HA) by Hyal-1 only when CD44 antigen is present =====

Statement from paper:

In this study, CD44, a receptor for HA, and hyaluronidase-1, -2, and -3 (Hyal-1, -2 and -3) were stably expressed in HEK 293 cells and the mechanism of HA catabolism was systematically investigated using fluorescein-labeled HA. Without CD44 expression, none of Hyal-1, Hyal-2, and Hyal-3 affected catabolism of fl-HA. Most of the fl-HA in the medium remained intact and was eluted in the void volume of the Sepharose CL-2B column (Fig. 2, A)

'''''previous recommendation''''' Hyal-1: hyaluronan catabolic process: PMID:17170110: in_presence_of(UniProtKB:P16070 CD44)

Ruth: could this be captured using IGI and including CD44 in the with column, and also creating the forwards annotation?: <br/>
Hyal-1: hyaluronan catabolic process: PMID:17170110: IGI: CD44
CD44: hyaluronic acid binding: PMID:17170110: IDA: causually_upstream of hyaluronan catabolic process

===== 10. Activation of caspases by ACER2 in presence of 4-hydroxyphenyl retinamide =====

Statement from paper:

We found that ACER2 knockdown significantly inhibited both the 4-HPR-induced cleavage of PARP (Fig. 5E) and activation of caspase-3 (Fig. 5F), supporting that ACER2 knockdown inhibits the 4-HPR-induced cell death in tumor cells... ACER2 overexpression markedly augmented both 4-HPR-induced cleavage of PARP (Fig. 6D) and activation of caspase-3 (Fig. 6E), supporting the view that ACER2 overexpression sensitizes tumor cells to 4-HPR-induced cell death.<br/>
'''''previous recommendation'''''  ACER2: activation of cysteine-type endopeptidase activity involved in apoptotic process: PMID:20628055: IMP : in_presence_of(CHEBI:42588 4-hydroxyphenyl retinamide) <br/>

Ruth: looking again at the abstract: Increased generation of dihydrosphingosine (DHS), a bioactive sphingolipid, has been implicated in the cytotoxicity of the synthetic retinoid N-(4-hydroxyphenyl)retinamide (4-HPR) in tumor cells. However, how 4-HPR increases DHS remains unclear. Here we demonstrate that 4-HPR increases the expression of ACER2, which catalyzes the hydrolysis of dihydroceramides to generate DHS, and that ACER2 up-regulation plays a key role in mediating the 4-HPR-induced generation of DHS as well as the cytotoxicity of 4-HPR in tumor cells. ACER2 converts of DHC into DHS. <br/>

The abstract suggests that important process is the generation of DHS, and the other side is that I think currently discussing whether or not to include drugs in column 16, or whether this drugs should just be added to the WITH field, currently this would not be possible as this is not a binding expt and WITH cannot be included with IDA. Plus I think this annotation should be removed as I am not convinced that ACER2 directly activates caspases. (but it might be good to have the discussion assuming that it is possible to make the annotation).

Do we need a term to capture role of proteins which produce DHS which then induces apoptosis?