Closed gocentral closed 9 years ago
Original comment by: cmungall
Logged In: YES user_id=254723 Originator: NO
CHEBI:16865 gamma-aminobutyric acid is_a neurotransmitter
neither CHEBI:14321 glutamate(1-) nor CHEBI:29987 glutamate(2-) are
Mike Bada has made some GO-ChEBI cross products (available soon) that will help spot inconsistencies here.
However, it's not as simple as this. As you point out, the key question is whether the molecules play the role of neurotransmitter.
ChEBI makes a good start, separating role from structure at the top level
However, ChEBI then mixes everything in an is_a hierarchy (attached). The link between GABA and neurotransmitter is "is_a" - I think it would be fantastic if ChEBI were to change this (and other such links) to "has_role". Ideally the role hierarchy would be pure roles, such as "neurotrasmitter role".
This wouldn't answer David's question in itself, but I think it's a good step to clarifying these kinds of issues.
I think GO could do with some clarification here itself; for example:
GO:0042165 neurotransmitter binding [DEF: "Interacting selectively with a neurotransmitter, any chemical substance that is capable of transmitting (or inhibiting the transmission of) a nerve impulse from a neuron to another cell."]
GO:0030594 neurotransmitter receptor activity [DEF: "Combining with a neurotransmitter to initiate a change in cell activity."]
In binding, it does not seem the molecule has to be playing the role - it's just any substance capable of playing the role.
In receptor activity, it is unspecified.
As for David's final question - we could use the GO-ChEBI cross products find GO functions or processes that have chemicals defined by certain roles as participants; for example, all the terms below represent functions or processes that have a CHEBI:hormone as participant:
GO:0006697-ecdysone biosynthesis GO:0006701-progesterone biosynthesis GO:0006704-glucocorticoid biosynthesis GO:0006705-mineralocorticoid biosynthesis GO:0006708-ecdysone catabolism GO:0006709-progesterone catabolism GO:0006712-mineralocorticoid catabolism GO:0006713-glucocorticoid catabolism GO:0006716-juvenile hormone metabolism GO:0006718-juvenile hormone biosynthesis GO:0006719-juvenile hormone catabolism GO:0008205-ecdysone metabolism GO:0008211-glucocorticoid metabolism GO:0008212-mineralocorticoid metabolism GO:0009683-indoleacetic acid metabolism GO:0009684-indoleacetic acid biosynthesis GO:0009685-gibberellic acid metabolism GO:0009686-gibberellic acid biosynthesis GO:0009690-cytokinin metabolism GO:0009691-cytokinin biosynthesis GO:0009692-ethylene metabolism GO:0009693-ethylene biosynthesis GO:0009694-jasmonic acid metabolism GO:0009695-jasmonic acid biosynthesis GO:0009696-salicylic acid metabolism GO:0009697-salicylic acid biosynthesis GO:0009721-detection of auxin stimulus GO:0009722-detection of cytokinin stimulus GO:0009752-detection of salicylic acid stimulus GO:0009754-detection of jasmonic acid stimulus GO:0009823-cytokinin catabolism GO:0009850-auxin metabolism GO:0009851-auxin biosynthesis GO:0009852-auxin catabolism GO:0009926-auxin polar transport GO:0010184-cytokinin transport GO:0030072-peptide hormone secretion GO:0032341-aldosterone metabolism GO:0032343-aldosterone catabolism GO:0042359-vitamin D metabolism GO:0042368-vitamin D biosynthesis GO:0042369-vitamin D catabolism GO:0042437-indoleacetic acid catabolism GO:0042448-progesterone metabolism GO:0042457-ethylene catabolism GO:0042701-progesterone secretion GO:0043400-cortisol secretion GO:0045443-juvenile hormone secretion GO:0045487-gibberellic acid catabolism GO:0046244-salicylic acid catabolism GO:0005499-vitamin D binding GO:0005500-juvenile hormone binding GO:0010011-auxin binding GO:0035100-ecdysone binding GO:0051740-ethylene binding
File Added: gaba.png
Original comment by: cmungall
Logged In: YES user_id=631592 Originator: YES
Hi Chris,
Thanks for the comment. I've also been thinking about this in the context of function-process links. On the one hand, we could have a term like 'dopamine binding' and then create children like 'neurotansmitter binding, dopamine' and 'hormone binding, dopamine'. On the other hand, we could have one term and relate it to two new biological process terms like 'postsynaptic reception of a signal' and 'reception of a hormone signal'. An instance of the execution of 'dopamine binding' could possibly be involved in an instance of either process. If we knew which process (for example the experiment was done in culteured neurons) then we could also annotate to the process type.
For either way of representing this, we should be consistent in the ontology because we eventually would like to reason about the possible processes that an execution of a molecular function can be involved and the process/es for which we have experimental evidence that the functioning is involved.
On another note, this all gets back to the big nasty discusiion of how far to take things like binding terms. Do we want to include things like orgnaismal origin, specific types of gene products, details about roles that molecules play? A very interesting discussion for the next GOC meeting.
David
Original comment by: ukemi
Logged In: YES user_id=1285648 Originator: NO
Dear David and Chris,
Chris has point out an important thing: whether the molecules which we are looking at play the role of neurotransmitter. I have a look at the child terms of GO:0030594 neurotransmitter receptor activity. As you can see, there are several terms such as neuropeptide receptor activity GO:0008188. The definition of a neuropeptide from Dorland Medical dictionary (http://www.mercksource.com/pp/us/cns/cns\_hl\_dorlands.jspzQzpgzEzzSzppdocszSzuszSzcommonzSzdorlandszSzdorlandzSzdmd\_n\_08zPzhtm\#12570541) is: any of several types of molecules found in brain tissue, composed of short chains of amino acids; they include endorphins, enkephalins, vasopressin, and others. They are often localized in axon terminals at synapses and are classified as putative neurotransmitters, although some are also hormones. So here is the same case of GABA vs glutamate: some neuropeptides are hormones that have also the neurotransmitter function (for instance neurotensin). I hade this problem when I started to work on neurotransmitter secretion/release and in particular on catecholamine. Norepinephrine secretion GO:0048243 is_a neurotransmitter secretion GO:0007269 and also a is_a of hormone secretion GO:0046879. I did not understand very well if we would like to follow the classification of ChEBI molecular role ontology or not.
Chris wrote: CHEBI:16865 gamma-aminobutyric acid is_a neurotransmitter neither CHEBI:14321 glutamate(1-) nor CHEBI:29987 glutamate(2-) are.
Glutamate 1 and 2 are not directly is_a child of neurotransmitter but they are conjugate base of L-glutamic acid that has a is_a with neurotransmitter. So are we sure that the conjugate base L-glutamate 1 and 2 are not neurotrasmitter for ChEBI? I am working on the "response to drug" terms and we decided to follow the molecular structure ChEBI ontology and it is very useful. Uses of cross products between ChEBI and GO could be a solution.
I have forwarded this discussion to an expert in Glaxo here in Italy because I thought he would be helpful.
Cheers Erika
Original comment by: arike
Logged In: YES user_id=631592 Originator: YES
Erika,
Thanks. I was hoping you would get involved in this discussion. I think that there are problems in both GO and Chebi. I think that Chris correctly points out that these molecules play a role as neurotransmitters or actually they can play a role as neurotransmitters. This is different fro them seing structurally is_a something else. It is important to keep in mind as you discuss this with the neuroscience experts; in GO if 'serotonin binding' is a 'neurotransmitter binding' then every time a molecule functions in 'serotonin binding' is has to function as 'neurotransmitter binding'. So when we ask the experts we need to distinguish between 'always is a neurotransmitter' and 'can be a neurotransmitter'. If we could get a list of the chemicals in GO categorized this way it would be a great start.
David
Original comment by: ukemi
Logged In: YES user_id=1285648 Originator: NO
Hi,
>If we could get a >list of the chemicals in GO categorized this way it would be a great >start.
Good idea. I could help to get this list complete if you agree.
Cheers Erika
Original comment by: arike
Logged In: YES user_id=802919 Originator: NO
Dear all
we are a bit reluctant to add new types of relationships to ChEBI (even though two more are in the "pipeline"). Instead of "has role" we'd rather put the entity in the "biological role" ontology. In this respect, if GABA is_a neurotransmitter which is_a biological role, then it is equivalent to the "GABA has_role neurotransmitter"
Note however that roles ('biological role' ontology) and uses ('application' ontology) are not absolute (while the 'molecular structure' and 'subatomic particle' are). They depend on the system (which could be a cell, organism, or some artificial in vitro system). What is a hormone in one organism, is a food in another. Therefore I would advise against cross products of GO and ChEBI 'biological role' and 'application' nodes. Then situations like 'serotonin binding' is_a 'neurotransmitter binding' will not arise.
As for L-glutamate(2-)/L-glutamate(1-)/L-glutamic acid: I do not know for sure which form is a real neurotransmitter; thus the advice of neurobiology expert is needed. I know that neutral but not charged GABA is a neurotransmitter.
Original comment by: kiri11
Logged In: YES user_id=1744794 Originator: NO
Hi all,
I agree with David that I think there's a conflation between what is happening for a given instance versus what is capable of happening for the instance. If substances are to be classified under role chemicals (as they currently are in ChEBI), then I think a role should be regarded as a functionality that a substance reasonably can assume, as no substance is doing anything continually: Serotonin molecules hanging around in a benchtop solution are not neurotransmitting, just as, though I'm a bioinformatician, I don't continually function in that capacity. And, as Kirill pointed out, a substance can play different roles in different in vivo conditions.
With this in mind, it seems ChEBI uses the "capable" definitions for classification under roles, at least implicitly. As for the composite GO terms that use these ChEBI terms, I really feel that the GO terms should be aligned with the ChEBI terms; I'd find it odd otherwise. Using the same "capable" criterion then, a serotonin-binding event would be a binding in which an entity binds to serotonin (which can act as a neurotransmitter), and a neurotransmitter-binding event would be a binding in which an entity binds to a substance that can act as a neurotransmitter. Therefore, it's not necessary that a serotonin-binding event is always a neurotransmitter-binding event, and a serotonin binding could be classified under neurotransmitter binding (just as serotonin is classified under neurotransmitter), even though serotonin may not always function as a neurotransmitter.
Following up on Chris's post, all of the terms under "biological role" are chemicals, so perhaps "biological role" could be renamed "chemical with biological role"?
Cheers, Mike
Original comment by: mbada
Logged In: YES user_id=254723 Originator: NO
kirill - I understand your reluctance to add new relations - these things should not be undertaken lightly, especially something like changing the links between the structural classification from is_a to has_role - this would be an absolutely huge change. I'm just trying to think long term.
it's interesting you take of the chebi "biological role" ontology and the "molecular structure ontology". In fact, there is no computable way to make this distinction with the chebi right now. Let me explain:
There are some terms, such as CHEBI:16865 "gamma-aminobutyric acid" that have is_a parentage to both role (neurotransmitter) and structure. A computer can use this to figure out that the term is in both (sub)ontologies, reflecting the dual aspect of this term.
So perhaps we can figure out what the "pure" roles are in chebi, by finding terms that have is_a parentage only to role.
This would include some terms, like CHEBI:18472- (-)-12-hydroxy-9,10-dihydrojasmonic acid, which doesn't sound like the name of a role.
Put another way, there is no way to computationally differentiate in chebi "pure" roles such as hormone, xenobiotic, food, from these jasmonic acid terms and the like, that represent chemical entities that can be classified as having a certain role.
So when you say you would advise against cross products of GO processes and biological or application roles, we can't actually computationally enforce this rule.
Anyway, I'm not sure we would want to! "neuropeptide receptor" gets some 40,000 hits on google, pubmed gets 111, so it seems the term is in use!
In response to Erika's question, I think GO's is_a structure should follow ChEBI's is_a, except in those cases where the ChEBI is_a is really more of a has_role. In the has_role cases, some judgement would be applied with a solution along the lines David outlines below.
Clarification of the distinction in chebi would help enormously, although as we are currently doing this manually anyway there is not such a huge hurry. However, I hope this will feature on the chebi roadmap at some point in the future! Some possible short-term workarounds could include Chebi-slims for "pure" roles like hormone, xenobiotic, etc.
Original comment by: cmungall
Logged In: YES user_id=802919 Originator: NO
cmungall wrote:
> This would include some terms, like CHEBI:18472- (-)-12-hydroxy-9,10-dihydrojasmonic acid, > which doesn't sound like the name of a role.
No it isn't. Nor I think that
jasmonic acids (CHEBI:24937) is a phytohormone (CHEBI:26158)
is a correct statement. This branch of ontology is virtually untouched. But eventually we'll get to it. The checked compound jasmonic acid (CHEBI:18292) belongs to oxo monocarboxylic acids (CHEBI:35871) so I presume most of "jasmonic acids" will go there. (I am not sure that the term "jasmonic acids" will stay however:m it is not a recognised class name).
> So when you say you would advise against cross products of GO processes > and biological or application roles, we can't actually computationally > enforce this rule.
Why not? Isn't it possible just to ignore everything that has no parent in molecular structure ontology?
Kirill
Original comment by: kiri11
Logged In: YES user_id=254723 Originator: NO
Kirill - OK, this makes sense, I hadn't accounted for the fact that ChEBI like all ontologies is in flux.
Nevertheless, I presume there will be more terms like the jasmonic acids that using your algorithm appear to be roles but are in fact strucural entities. If we had has_role rather than is_a in the appropriate places then we wouldn't need to worry about this.
But let's presume that all terms like this get assigned their proper structural parentage. I still do not think we want the rule that no CHEBI terms lacking a strcutral parent should get a GO cross-product. GO likes having neurotransmitter terms.
So if we retain these, we have to have some kind of rule like:
Process1 involves Chebi1
IMPLIES
Process1 involves Chebi2
WHEN
Chebi2 is_a+ Chebi1 AND Chebi1 is_a+ Chebi:molecular_entity
It would be far simpler if we could drop the second clause; and we can do this if Chebi uses has_role to link structural entities to the roles they play
Cheers Chris
Original comment by: cmungall
Logged In: YES user_id=802919 Originator: NO
> 1 year later...
There was a suggestion [1] during this week's ChEBI user group meeting [2] to introduce the new relationships such as has_biological_role and has_application. In other words, there should not be any is_a crossing over from molecular structure to biological role and application ontologies. I am happy to say that there is consensus (among both ChEBI people and ontologists, that is) that this eventually has to be done. Note the word "eventually". There are plenty of tricky cases such as "protein-derived cofactors" or "tricyclic antidepressants" - it is difficult to imagine how these can not be is_a linked to both molecular structure and role/application. But that is what makes it interesting.
[1] Several times and rather loudly repeated. [2] http://www.ebi.ac.uk/industry/Workshops/chebi-users-workshop-190508.html
Check this space in couple of years...
Kirill
Original comment by: kiri11
Logged In: YES user_id=1473024 Originator: NO
Hi Kirill,
You can start now with the trivial cases. This was identified as a priority on Monday because it will sort many of the problems GO is having with ChEBI.
Please list the non-trivial cases you find and put them up for discussion on the ChEBI tracker.
Best wishes, Colin.
Original comment by: batchelorc
Logged In: YES user_id=802919 Originator: NO
Hi Colin
I can not start now because it is Friday afternoon and we have release next Wednesday.
We have a lot of things to discuss post workshop before starting any implementation.
Cheers, Kirill
Original comment by: kiri11
I hopefully have addressed all of these. I have moved them up to the more generic parents.
Original comment by: ukemi
Original comment by: ukemi
There are a number of receptor activities in the molecular function graph that have been called into question as I have looked at the neurotransmission parts of the biological process graph.
Should GABA receptor activity and glutamate receptor activity be neurotransmitter receptor activities? Note that to make this relationship, every time they function the GAB and glutamate should be playing the role of a neurotransmitter.
If dopamine is indeed a hormone as well as a neurotransmitter, then its receptor activity should be removed as a child of neurotransmitter receptor activity.
I think the same goes for adrenocorticotrophin activity.
Can anyone spot any others?
David
Reported by: ukemi
Original Ticket: "geneontology/ontology-requests/4063":https://sourceforge.net/p/geneontology/ontology-requests/4063