Drug Terms in the GO

[gocentral]

We currently have the following "drug" terms in the GO:

8 Molecular Function terms:

drug binding drug transporter activity drug:hydrogen antiporter activity multidrug efflux pump activity multidrug endosomal transporter activity multidrug transporter activity multidrug, alkane resistant pump activity xenobiotic-transporting ATPase activity

11 Biological Process terms:

drug catabolism drug export drug metabolism drug transport endogenous drug catabolism exogenous drug catabolism induction of synaptic plasticity by chemical substance multidrug transport positive regulation of synaptic plasticity by chemical substance regulation of synaptic plasticity by chemical substance response to drug

Based on discussion on the GO mailing list, it is proposed that response to drug ; GO:0042493 will be obsoleted in favor of specific "response to" terms for particular chemical substances which will be grouped under response to chemical stimulus ; GO:0042221, to be organized in parallel with the ChEBI molecular structure ontology. This change removes the burden of deciding what is or is not a "drug" from the GO.

Based on similar reasoning, I would further propose that all GO terms that specifically include the word "drug" or "multidrug" in the term name be obsoleted for similar reasons, and that the child terms of such terms that refer to specific chemical substances be rehomed appropriately. As this will affect lots of annotations, I would welcome additional discussion of this issue.

-- Alex

Reported by: addiehl

Original Ticket: "geneontology/ontology-requests/3378":https://sourceforge.net/p/geneontology/ontology-requests/3378

[gocentral]

Logged In: YES user_id=735846

Longer version of proposal to show to users:

'response to drug ; GO:0042493' proposal

The GO Process ontology currently has a number of terms grouped under 'response to drug'. These terms include 'response to caffeine', and 'response to antibiotic', 'response to cocaine' and so on. These terms are of interest to pharmaceutical researchers for the annotation of gene products that respond to the administration of these drugs.

Child terms such as 'response to cocaine' make sense in the context of the process ontology, but grouping them under 'response to drug' does not. 'Drug' is a very subjective term. For example an antibiotic may be a drug to a human, but to a bacterium it is not.

In order to solve this problem the GO Consortium has produced the following proposal:

1) The grouping term 'response to drug ; GO:0042493' will be obsoleted.

2) The child terms will be retained and will be moved to be under the existing high level term 'response to chemical stimulus ; GO:0042221

3) The child terms will be grouped according to the ChEBI ontology. http://www.ebi.ac.uk/chebi/

4) Synonyms will be added to aid searching, as required. For example 'response to benzodiazepine' could have a related synonym 'response to antidepressant'.

This proposed change removes the burden of deciding what is or is not a "drug" from the GO.

Original comment by: jenclark

[gocentral]

Logged In: YES user_id=436423

also see SF 1586006 -- 'response to herbicide ; GO:0009635' and 'response to insecticide ; GO: 0017085' will also have to become related synonyms instead of terms.

https://sourceforge.net/tracker/index.php? func=detail&aid=1586006&group_id=36855&atid=440764

Original comment by: mah11

[gocentral]

Logged In: YES user_id=436423 Originator: NO

One more SF: 1658374, on toxins/toxic substances:

https://sourceforge.net/tracker/index.php?func=detail&aid=1658374&group\_id=36855&atid=440764

Also, those who have access to the GO internal wiki can see the more recent plan from Erika Feltrin:

http://gocwiki.geneontology.org/index.php/Response\_to\_drug

Original comment by: mah11

[gocentral]

Logged In: YES user_id=254723 Originator: NO

Where are we with this plan?

'response to drug' GO:0042493 has no is_a children as it stands. Yet GO has GO:0042220 'response to cocaine'. The terms are unrelated. This seems very odd.

Furthermore, GO:0015904 tetracycline transport and many other drug transport terms are part_of response to drug. Is this a TPV? When is an "X transport" a "response to X"? Is it only when X is non-endogenous.

(there is no cocaine transport)

It seems response to drug can be safely obsoleted.

Note that if we do obsolete people can still find 'response to drug' using oboedit and the chebi xps and the chebi 'drug' hierarchy (chebi should really use a has_role relation here rather than overloading is_a)

-- 3) The child terms will be grouped according to the ChEBI ontology. http://www.ebi.ac.uk/chebi/

Original comment by: cmungall

[gocentral]

Logged In: YES user_id=254723 Originator: NO

The wiki says:

---

14th November, 2007

I have a thought: proposal for "drug transport" Try to follow ChEBI structure to move all child terms of "drug transport" out of this node.

What does it mean to follow ChEBI structure? ChEBI overloads is_a, and has is_a links from e.g. cocaine to both 'drug' and structure. We should not duplicate this mistake and instead just use the structural classification (if chebi used a has_role relation we could easily use OE to fix the GO hierarchy).

Most of the proposals seem good.

---

To be asked ChEBI curator WAIT before doing that!!!!

* amiloride

• acriflavine
• azole
• benomyl
• cycloheximide

[edit]

Why wait?

Also, why is bicyclomycin not on the list?

It would be great if these were requested through the chebi tracker & the link placed here.

---

Move "bicyclomycin transport" as is_a "antibiotic transport"

should we not strive for a structural classification here?

Is antibiotic not a relational term?

Original comment by: cmungall

[gocentral]

Logged In: YES user_id=254723 Originator: NO

I can't find chebi terms for the following:

acriflavine transport nalidixic acid transport organomercurial transport amiloride transport aminotriazole transport benomyl transport cycloheximide transport bicyclomycin transport microcin transport microcin B17 transport polymyxin transport fosmidomycin transport azole transport

Original comment by: cmungall

[gocentral]

Logged In: YES user_id=1285648 Originator: NO

Dear all,

Alex summarized correctly and clearly which is the plan about “response to drug” terms. It has been proposed to follow the ChEBI molecular structure ontology and classify the “drug” terms on the basis of the structure of the drug and not its role as a drug. I have done some of the steps in the “response to drug” proposal but there is still a lot of things to do. I followed, as said, the ChEBI molecular structure ontology and move terms as methotrexate, cocaine, nicotine out of “reposne to drug” and to be child of “response to organic cyclic substance” GO:0014070 I have just move GO:0046677 response to antibiotic out of response to drug because “antibiotic” is classified in ChEBI biological role and application ontologies so I cannot follow ChEBi in this case and we have to be very careful about that.

Below is an example:

GO:0014070 : response to organic cyclic substance --[i] GO:0014071 : response to cycloalkane --[i] GO:0014072 : response to isoquinoline alkaloid ----[i] GO:0043278 : response to morphine --[i] GO:0031427 : response to methotrexate --[i] GO:0035094 : response to nicotine --[i] GO:0014074 : response to purine ----[i]GO:0031000 : response to caffeine --[i] GO:0014073 : response to tropane --[i] GO:0042220 : response to cocaine

The idea is to move all terms children of “response to drug” and obsolete the term and I agree with Alex that we should also obsolete all terms with the word “drug” in the term name.

There are some obstacles to be solved.

** Unfortunately these terms have annotations For instance drug export GO:0046618 has just 12 gene associations and also response to antibiotic have 296 associations.

** multidrug transport GO:0006855 has almost the same definition of drug transport GO:00415893 but it has 318 gene associations. So we could merge it with "chemical substance transport". We need to discuss about that.

I would like to clarify some things written on the wiki page: I did changes on “drug transport” on a scratch file that I had submitted to the cvs but then, Jen and I thought it was better to contact the mailing list before doing more changes (just after that, I started to become mad writing my PhD thesis and delay the work to better time. Apologies for this!).

Then, the sentence “To be asked ChEBI curator WAIT before doing that!!!!” is just to remember me that we need to ask also for these terms. Cyclohexemide is present in the ChEBI ontology but this term need to be manually curated before be visible on the ChEBI browser. If you search for “cyclohexemide” in .obo file, you can find it. [Term] id: CHEBI:27641 name: Cycloheximide alt_id: CHEBI:23484 alt_id: CHEBI:4015 is_a: CHEBI:23452

I know that this topic is very tricky and a lot of people might be involved in it so any comments and suggestions are very very welcome.

Thanks, Erika

Original comment by: arike

[gocentral]

Logged In: YES user_id=482944 Originator: NO

Hi all,

I realize I'm chiming in very late here - but I've been thinking about this kind of thing a lot lately and this occurs to me:

I totally agree that it is impossible to classify compounds as either toxins or not toxins and drugs or not drugs so that it will be true for all cases. However, I don't agree that it follows from this that all of the toxin or drug terms need to be obsoleted. What I do think follows from this is that no compound-specific terms should be children of the toxin or drug terms. But the concept of response to toxin or response to drug is still quite sound. The GO would not assert that any given compound is always a drug or toxin, but annotators could make that assertion through dual annotations to (for example) response to drug and response to compoundX. Why isn't this ok?

Michelle

Original comment by: mlgwinn

[gocentral]

Logged In: YES user_id=735846 Originator: NO

Hi Michelle,

This is very good timing indeed, as Erika has moved all the specific response to drug terms out of under response to drug but did not have time to do the obsoletions before she stopped working to write her PhD thesis. I see your point, and can see the utility of this term. I'd be interested to know what others think.

Jen

Original comment by: jenclark

[gocentral]

Logged In: YES user_id=436423 Originator: NO

The problem is that if we were to keep the term, we would need a definition of "drug" that is unambiguous and works for all species

As we noted some time ago, in the email thread that led to this item, the designation "drug" is arbitrary: it has no relationship to any chemical or functional characteristics of substances; both naturally occurring and synthetic substances can be considered drugs; a given substance can be regarded as a drug in some contexts but not others. These issues mean that defining "drug" satisfactorily would certainly be difficult, and may be impossible. Alex also pointed out that many (though not all) of the circumstances in which responses to drugs are observed in fact represent abnormal biological situations.

Similar concerns apply to the 'toxin' terms.

m

Original comment by: mah11

[gocentral]

Logged In: YES user_id=482944 Originator: NO

Hi,

I agree that what is a drug is arbitrary when you are looking across species. However for a given species it is generally clear if something is used as a drug or not. I think the drug case is harder than the toxin case since toxins are generally defined by the ability to cause harm, while drugs are meant to do many things other than harm something. So I agree it is difficult, but I don't think it is impossible. I think it must be defined so that in any given organism it is possible to determine when something is a drug response and when it is not.

As for these things being abnormal, I don't think it is abnormal at all. I consider drugs a class of xenobiotic that are produced by humans with the intent of modifying the physiology of a target organism. Now from the standpoint of the organism that has been targeted, the organism doesn't really care what the source of the xenobiotic is (natural or human made) but the targeted organism sure does care about responding to it. And in fact responding to such substances is a very normal part of biology - without such activities, most organisms would die pretty fast.

So why have a term for drug responses - because humans really care about which genes are involved in responding to drugs and which are not. And for a particular target organism one should be able to pretty easily tell which compounds are used as drugs and which are not. Drug resistance mechanisms, drug interactions, drug side effects - all of these areas of biology are very important - should we not have terms to capture the processes involved in these areas in GO?

I suggest this definition for drug: "a xenobiotic used in a medical application to intentionally modify the physiology of a target organism"

Michelle (a very similar comment to this one is in SF item #1658374)

Original comment by: mlgwinn

[gocentral]

Logged In: YES user_id=254723 Originator: NO

At the very least this should be coordinated with CHEBI.

Ideally there would be 3 orthogonal ontologies:

[1] GO-BP [2] CHEBI [3] Chemical roles and applications

The relations between terms in these ontologies would be participation (process to chemical entity) and has_role (chemical entity to role). Anyone wanting to know genes involved in transport of drugs would query annotations + [1] + [2] + [3].

CHEBI in fact includes a sub-ontology of roles, including "drug". Unfortunately CHEBI overload the is_a relation rather than using a has_role relation. Still, it is a start.

I would still rather keep decisions of what is a drug and what is not out of GO and do it dynamically using external ontologies. Ideally this external ontology would have knowledge such as "foomycin is used by humans as a drug against M.snuffalococcus and is produced by S.flabagabagus" encoded in its DAG. I think this may satisfy Michelle's species-specific role requirements.

Michelle's alternate solution is to have dual annotation to drug and compound GO terms;

On 2008-02-19 09:42 Michelle said:

"What I do think follows from this is that no compound-specific terms should be children of the toxin or drug terms"

But would this not lead to mass confusion if we have a term like "foomycin transport" not an is_a child of "drug transport"? Especially if CHEBI classify foomycin is_a drug.

Perhaps another solution to Michelle's requirements is to use the new annotation properties column to indicate that the context is a compound produced for clinical purposes.

Original comment by: cmungall

[gocentral]

Logged In: YES user_id=436423 Originator: NO

The main thing is that I agree completely with Chris.

I'd also like to note that not all drugs are xenobiotics. Even using the 'foreign to a particular organism' definition of 'xenobiotic' (as opposed to 'foreign to all organisms), some drugs don't fit. Recombinant human insulin or growth factor would be considered drugs in the clinical settings in which they're administered. Drugs also aren't all produced my humans ...

m

Original comment by: mah11

[gocentral]

Logged In: YES user_id=482944 Originator: NO

Hi,

First let me say that I never included "produced by humans" in my def of drug because I realized humans don't produce them all.

Second, I agree that any kind of replacement therapies like insulin or HGF would be problematic under the proposed def. I'll have to think about how to handle that. But just because defining something is hard does not mean that we shouldn't do it.

Third, I agree that it would be great to have an, and I quote, "external ontology [that] would have knowledge such as "foomycin is used by humans as a drug against M.snuffalococcus and is produced by S.flabagabagus" encoded in its DAG." But we don't have that - we are no where near having that - CheBI is certainly not that - is CheBI planning on doing that? It will be a HUGE amount of work as it is basically encoding annotations in a DAG. I'm not sure that's even desirable - it should be handled as multiple annotations I think.

Fourth - Chris thinks it will be confusing for users to see "response to foomycin" is not a child of "response to drug" - perhaps it will, but as I said, a comment on the term explaining why this is so will perhaps help with that.

Fifth - if you remove the response to drug/toxin terms you will remove the ability to query out all proteins involved in these processes and that will be a big loss since that kind of query is exactly the kind of thing that GO should be doing. And that kind of question is what GO should be able to answer for biologists interested in these processes. If biologists come looking for something and its not there - we risk losing them as GO users forever.

:)

Michelle

Original comment by: mlgwinn

[gocentral]

Logged In: YES user_id=803460 Originator: NO

Michelle,

Is it possible to use synonyms (possibly, narrow) to address this?

Original comment by: pgaudet

[gocentral]

Logged In: YES user_id=865072 Originator: YES

For what it's worth, I still agree with my earlier arguments as well as those of Midori and Chris. I would prefer to see an outside ontology of drugs or CHEBI to identify these substances. However, I also appreciate the utility argument of Michelle and think that Pascale's proposal to use synonyms might address the issue as a compromise.

-- Alex

Original comment by: addiehl

[gocentral]

Logged In: YES user_id=482944 Originator: NO

Hi again,

I had actually assumed that synonyms would be added if you went forward with the obsoletion. But I do not think they are a solution or even a compromise. You will still not be able to serve the biologist who wants to collect every gene product that an annotator knows is involved in a response to a drug. Even if that biologist could collect annotations to every term that has a synonym of response to drug that will not be the same thing as an annotator asserting that a gene product is involved in response to drug.

I've had some interactions with biologists out in the world trying to use GO and I tell you they are frustrated by it. Some of the "rules" just don't make sense to them. I try to explain about why its important from the ontology structure perspective and stuff, but they don't really care about that. They want to find the concepts that are important to them as biologists.

As for the possibility of confusing biologists by not having the individual drug terms under response to drug - the same kind of situation exists for pathogensis. When we in the PAMGO group started trying to make terms for host interactions with pathogens we quickly saw that there was no way to draw a clean line between pathogenic processes and other non-pathogenic symbiotic processes. So we made set of terms to describe all the ways symbionts and hosts interact with each other and then we made a separate term for pathogenesis which has no children. We depend on dual annotation to tell the whole story. Now, will a biologist wonder why we don't have children under pathogenesis, probably, but we need to explain why to them, which we do whenever we do trainings on these terms. (But we should add to the comment I think). I don't see this drug situation as any different. Sometimes processes are pathogenic sometimes not depending on the organisms and environment, sometimes things are drugs and sometime not depending on the organisms and circumstances. But the concept of pathogenesis is still valid as is the concept of response to drug.

Michelle

Original comment by: mlgwinn

[gocentral]

Logged In: YES user_id=436423 Originator: NO

superseded by discussions at SLC GOC meeting, April 2008

Original comment by: mah11

[gocentral]

• status: open --> closed-out-of-date

Original comment by: mah11