Spreadsheet of 'response to' term requests

slaulederkind commented 6 years ago

Response to requests 9-2017 B.xlsx

ukemi commented 6 years ago

I didn't look at this, but what is it? If it is a long list of terms, perhaps it could be done programmatically.

slaulederkind commented 6 years ago

I forgot to add: Where's the new TermGenie?

pgaudet commented 6 years ago

Hi @slaulederkind Would it be possible for RGD to use annotation extensions for these ? We would keep a number of parent classes that make sense for grouping purposes (the exact ones to be discussed), and specify the exact chemical in the 'has input' field.

Thanks, Pascale

ValWood commented 6 years ago

Personally I think that "response to" terms should ONLY be allowed in extensions, and blocked in primary annotation (similar to phase terms currently).

This would prevent their use for describing a change in viability in response to a specific chemical (a viability or growth phenotype).

I was going to recommend this soon, as a follow on from the slim talk (50% of mouse proteins already have 'response to' annotations, and often we would be better to class these as "unknown physiological role"). Sometime no annotation is better......."response to should only be applied as an extension to certain pathways (mainly signalling).

slaulederkind commented 6 years ago

RGD doesn't use column 16 at the moment, but maybe in the future.

pgaudet commented 6 years ago

Would it be possible to use Noctua for these annotations for now ? And then upload the annotations generated by Noctua back into RGD ?

cmungall commented 6 years ago

Personally I think that "response to" terms should ONLY be allowed in extensions, and blocked in primary annotation (similar to phase terms currently)

good suggestion. We could go even further and ditch the terms from GO and use an exposure event ontology...

RLovering commented 6 years ago

I maybe misunderstanding this thread but the UCL team is not convinced about removing the 'response to' terms. These enable grouping of many different processes together and when I have conducted functional analysis I have found it useful to see the response to terms enriched as they often help confirm the area of biology being disregulated, or relevant to the samples. If this is a serious proposal please let me know and I will get examples together to explain this further. But I do not want to do this unless there is a suggestion that all 'response to' terms are going to be removed. Would it be possible to do a text search for 'response to' and transcriptomic/proteomic/functional analysis to get a sense of how often papers find this term useful in their analysis?

ValWood commented 6 years ago

My comment was more of an opinion than a proposal. I do think we need to address the use of "response to terms"

As @ukemi said during my slim presentation at GOC, everything occurs in respsonse to something. Some response to terms are probably OK in GO, but we should really think what we mean when we use them.

Response to hormone, or other receptor ligand, probably OK, but would this not always be used in the context of a signalling pathway as an extension?

response to drugs, like " response to vasoconstrictor agent" seems outside of the scope of GO?

At PomBase we annotate response to x using the phenotype ontology, because all that is usually shown is an effect on growth in a mutant.

Where you say "often help confirm the area of biology being disregulated" indicates that these are not normal processes, and are therefore outside of the scope of GO.

We already almost half of our protein coding genes annotated to "response to some chemical" , and this isn't really often indicative of any particular process ( It is frequently just a pathological/toxic effect of the chemical). Similarly for mouse, over 11,000 gene products annotated to response to something terms.

When the response is part of an identifiable biological program/process, we can make an annotation to the process with "response to" in the extension (with occurs during, or part_of).

This is probably a good topic for discussion on a future annotation call, or at a future GO meeting. It would be good to have published examples of usefulness. I'm sure we have some good published examples of the phenotypes of drug sensitivities and resistance. They may be useful, but that doesn't necessarily mean that they should be GO terms. A generic cell level phenotype ontology would help to capture these....

mah11 commented 6 years ago

At PomBase we annotate response to x using the phenotype ontology ...

For this reason, it's convenient for us to have the "response to" terms in GO -- we use them in logical definitions for FYPO terms. That usage would be unaffected by flagging some or all of the GO response terms as not for direct annotation.

In principle we could rejig the FYPO logical defs to use the exposure event ontology that Chris alluded to. In practice, the ExO listed at OBO Foundry isn't remotely suited to our purposes at present (and I haven't found any others), so we would have to inundate its maintainers with requests as well as do the rest of the work of updating our defs. Our lives will be easier if we don't have to do all that.

So at present I favor keeping the terms, for admittedly selfish reasons. On whether to annotate using them, I'm approximately neutral, with no objection to some restrictions.

ValWood commented 6 years ago

A good example of where they should not be used is in changes in gene expression in response to chemical (for example arsenic). Arsenic causes problems for 100s of enzymes. Does it interfere with iron biosynthesis? (I'm not sure).

But if a chemical is toxic a good annotation would be either signalling pathway x occurs_during response to chemical y or

detoxification of chemical y

If you can't be this specific an effect might just be due to the toxicity.

A good first move would be to ban IEP for these terms perhaps?

RLovering commented 6 years ago

just wanted to also state that response to terms do not just cover the signaling pathways but the whole response of a cell to a specific signal, which might lead to for eg secretion of another signaling molecule

pgaudet commented 6 years ago

Hi @RLovering Can you give some examples of useful enrichments ? By definition enrichments should be groups, so perhaps having grouping terms such as 'response to antibiotic' and 'response DNA damage' + the with column for the specific treatment would be sufficient ?

Thanks, Pascale

ValWood commented 6 years ago

just wanted to also state that response to terms do not just cover the signaling pathways but the whole response of a cell to a specific signal, which might lead to for eg secretion of another signaling molecule

So in these cases wouldn't you want to couple? So exocytosis/signalling etc occurs_during response_to x

These "response to" terms are used in many unintended ways. For example (and these are the first 2 I looked at randomly!)

http://amigo.geneontology.org/amigo/reference/PMID:20978158 Title A novel human dynactin-associated protein, dynAP, promotes activation of Akt, and ergosterol-related compounds induce dynAP-dependent apoptosis of human cancer cells.

used for response to ergosterol? This isn't really saying anything about a biological process.

http://amigo.geneontology.org/amigo/term/GO:0036275 http://www.informatics.jax.org/reference/MGI:5636128 used for "response to flouro uracil.

These are not processes, they are phenotypes.

There is a reason to use some terms probably like "response to certain stresses, Oxidative, osmolarity, DNA damage" but in these cases it should be possible to say what the process is (signalling, or detoxification or whatever) with the occurs_during/part_of "response to x as an extension.

Otherwise- how do you know that it is really a valid organismal response?

ValWood commented 6 years ago

I would challenge all of these if I had time...... http://amigo.geneontology.org/amigo/term/GO:0060992

ValWood commented 6 years ago

response to methyamine? http://amigo.geneontology.org/amigo/reference/PMID:7440563 for Glycine dehydrogenase (decarboxylating), mitochondrial? methyamine is an inhibitor....

ValWood commented 6 years ago

http://amigo.geneontology.org/amigo/term/GO:1905705 cellular response to paclitaxel (from IEP) from http://www.sciencedirect.com/science/article/pii/S0014299908003968

Many of these terms are used only once or twice and they are not biological processes they are phenotypes. Perhaps to start with we should restrict the usage?

ValWood commented 6 years ago

Surely we need to be able to specify which process the "response to thing" is affecting?

vanaukenk commented 6 years ago

@pgaudet and @ukemi and I discussed the 'response to' issue on a recent call.

We think this issue merits a more in-depth discussion akin to what is currently being done for signaling pathways.

We'd like to propose a 'response to' session or workshop at the GOC meeting in NY in May. The main goals would be to:

Understand better what groups who use the 'response to' terms are hoping to capture with these annotations
Determine what level of granularity is appropriate for 'response to' terms in the ontology
Devise annotation SOPs for annotating 'responses'

We will discuss more on a future annotation call.

vanaukenk commented 6 years ago

Moved from #15103

We should take a good look what we are really using these terms to annotate.

Many appear to represent conditions. Some are phenotypes. Some are the target of a drug. Some are requested to construct logical definitions of phenotypes and should not be used in annotation (maybe we could do this differently?). It often isn't clear which BP the response is related to (when it usually could be). etc. You can throw any chemical at an organism and most times there will be some "response"...what does it mean?

Some examples: cellular response to topoisomerase inhibitor? (drug target?) cellular response to gold(3+)???? cellular response to alkane? cellular response to boron-containing substance levels? GO:1901593 response to GW 7647 ????? GO:0072757 cellular response to camptothecin (camptothecin is A topoisomerase inhibitor see above) GO:1904842 response to nitroglycerin? hmmm.... GO:0070301 cellular response to hydrogen peroxide?? isn't that just the oxidative stress response? It's a condition used to produce oxidative stress (response to oxidative stress is a useful one, if qualified).

It would be better if these terms were only used as refinements to true process annotations:

"signalling pathway x" occurs_during "response to y" or "cellular detoxification" in response to y.

It is difficult to assess what is "unknown" with so many uninformative "response to x " annotations...

Comment from @mah11 Re "response to" terms: quite a few of the strange-looking ones are there to allow other ontologies (including FYPO) to use them in logical definitions or other cross-references. Any such terms should be flagged as not for gene product annotation (I thought they were, but it's certainly possible that some were missed).

https://mailman.stanford.edu/pipermail/go-ontology/2016-April/007432.html

cmungall commented 6 years ago

For the FYPO owl definition use case, do you actually need a response-to-X GO term, or do you need the X? You could use your SPECO ontology here. Or in many cases, CHEBI. For other species we are developing a multispecies exposure ontology https://github.com/EnvironmentOntology/environmental-exposure-ontology/

mah11 commented 6 years ago

@cmungall

For the FYPO owl definition use case ... [questions]

I did a quick review of FYPO terms, and they fall into a few categories:

In most cases, usually cell population growth terms, we do use ChEBI for X. Presumably these are of no concern here.
Some terms refer to GO "cellular response to" terms that haven't been called into question (responses to calcium ions, iron ions, pheromone, ROS, or various stresses that have reasonably well-characterized responses).
Among the remainder, there are responses to chemicals, of which I think at least some could be converted to annotations using conditions+a chemical-agnostic FYPO term, but we would have to review the annotations to make sure. For some chemicals, there are both a lot of terms and a lot of annotations, so changing them would be a big, tedious task unless there's a way to automate it (applies to hydrogen peroxide, hydroxyurea, methyl methanesulfonate).
- easy example: abnormal chromosome morphology during cellular response to acetaldehyde
- harder example: abnormal mitotic cell cycle regulation during cellular response to hydroxyurea
The rest are the potentially problematic ones - the stimulus isn't presence of a chemical, and I don't know whether there's really a specific and biologically meaningful cellular response.
- examples: increased RNA level during cellular response to adenine starvation, decreased protein degradation during cellular response to UV

Could the exposure ontology be used regardless of whether a GO "response to" BP does or should exist? If so, we could switch to it in principle, but I would greatly appreciate guidance for suitable design patterns.

In the attached list I've ferreted out all the "response to" terms we're using, and put them in categories as above. responses_in_fypo.txt

mah11 commented 6 years ago

updated attached list (new term requested using "cellular response to leucine starvation")

mah11 commented 6 years ago

... and would the exposure ontology include "exposure to x starvation"?

geneontology / go-ontology

Spreadsheet of 'response to' term requests #14303