Closed LilyAndres closed 3 years ago
@ddooley, @CropStoreDb, @kaiiam, @laurenechan, @LuciaSegovia
I think the confusion comes from the way vitamins are modeled in CHEBI vs what we're trying to do in CDNO. CHEBI's vitamin
and fat-soluble vitamin
classes are roles
whereas the actual compounds like vitamin A
are molecular entities:
CDNO wants to maintain a nutritionist's subset of CHEBI (and additional) terms to make it easier for them to annotate their data and link to concepts from standards organizations like USDA. As such CDNO has the dietary nutritional component
hierarchy which consists of material entities with the exceptions of (plant secondary) metabolite
and vitamin
.
Unfortunately CDNO's assertion of the CHEBI vitamin hierarchy within a material entity hierarchy causes the discrepancies noted above.
I'm not sure what would be the best course of action but I see two possibilities:
1) Separately maintain role
and material entity
versions of terms like vitamin
(in either CHEBI or CDNO if desired)
2) Alternatively have CDNO modify it's framework to differentiate between material entity
and role
dietary nutritional components. The latter perhaps being similar to what @LilyAndres is calling dietary_function
classes?
I'm not sure where vitamers would fit in.
Primarily I'm advocating for a solution that makes clear whether an entity is a role or a chemical bearing a role just by looking at its label. Then in CDNO we can name the chemical compounds in a vitamin group in a clear way.
1) Ideally CHEBI would add the word "role" to those entity labels that are roles to avoid misuse by the many users that are unfamiliar with the role vs chemical entity distinction.
Just looking at the language in the definition of "vitamin B" illustrates this tug-of-war of language in CHEBI itself:
A role played by any of the group of eight water-soluble vitamins originally thought to be a single compound (vitamin B). The group comprises of vitamin B1, B2, B3, B5, B6, B7, B9, and B12 (Around 20 other compounds were once thought to be B vitamins but are no longer classified as such). B vitamins play important roles in cell metabolism.
"The group" is a group of chemical entities bearing a "has role" relation to Vitamin B, right? "20 other compounds were once thought to be B vitamins" ... implies an "is a" relation. "B vitamins play important roles in cell metabolism." If "B vitamins" are roles, then translation: "[roles] play important roles in cell metabolism"
2) With CHEBI offering terms like "B vitamin role", then we would have space to create a class "B vitamin" which is defined as a compound which has a "B vitamin role".
But if CHEBI won't add "role" to "B vitamin" etc., then we need a workaround to avoid duplicate terms in OBOFoundry, OLS etc.
3) Perhaps "B vitamer" can be defined as a compound which has a "B vitamin" role? Is this acceptable to common language use?
I am not 100% sure if I understand the has a role
and is a
properly, so I apologize on beforehand if what I'm saying here makes no sense.
First, I think we need to keep in mind that vitamins are very heterogenous group of substances and there is still a lot of research going on, so it might be the case that decisions should be taken base on individual vitamin groups. Would that be an option?
Then, following @ddooley example, indeed, vitamins are groups of compounds that has a role
in human bodies. Within those vitamin groups, each compound ('is a' ?) contributor to the vitamin activity often with different bioactivity.
For example, there is no compound called "vitamin A" as such, some compounds are responisble of the vitamin A activity in the human body. For illusitration, this is how it's normally calculated Vitamin A according to FAO/INFOODS:
Total vitamin A activity expressed in mcg retinol activity equivalent (RAE) = mcg retinol + 1/12 mcg ß- carotene + 1/24 mcg other pro- vitamin A carotenoids
So, maybe for this example, we could say that vitamin A has a role
and retinol is a
? 😅
I agree with @ddooley that label clarity (role vs entity) is essential in whatever approach is taken. That being said, I think attempting to model roles in any meaningful fashion may need to take place in CDNO as opposed to ChEBI due to the vast amount of different functions various vitamers can serve. And I do hope that we plan on modeling those mechanisms of action, albeit that may be a long term goal... Would it be reasonable to propose only is_a relationships in ChEBI to merely represent the vitamers and their grouping classes (e.g. B vitamins) and then include specific biochemical functions/roles within CDNO's developing dietary function branch?
I am not 100% sure if I understand the has a role and is a properly, so I apologize on beforehand if what I'm saying here makes no sense.
No worries BFO (the top level classification hierarchy of OBO ontologies) is a little complicated but basically roles and materials entities are two higher level concepts in BFO. For example a person is a material entity, a tangible physical thing on it's own, whereas a role needs something else to exist. E.g., a person could have a teaching role. So the role of teacher depends on the person for its existence. Is a
is the regular subclass hierarchical relation between classes for example a hammer is a
tool.
For the most part CHEBI's terms are material entities aka there are specific chemicals (or group of chemicals) that CHEBI terms such as vitamin A
represent. It's also true that one could see vitamins as having a role in metabolism, but they themselves are just chemicals therefore material entities. Things like fat soluble vitamin
I guess could be modeled either way?
I think @laurenechan and @ddooley and I are saying similar things, would it be possible for CHEBI to either have both role and material entity terms, or at least have a material entities version of all vitamin terms (with subclass relations between them)?
Thank you all for your inputs.
@mateolan thoughts?
as suggested by @laurenechan , I agree ChEBI is best placed to define the 'is a' molecular relationships that can then be reused within the CDNO nutritional component class. I also think it could be appropriate to maintain the definitions and relationships for dietary role within CDNO 'dietary function' branch. Just to clarify, there is likely need for a 'functional attributes' class (ie objective quantifications) distinct from evidence-based assertions of 'dietary function'.
I think @laurenechan and @ddooley and I are saying similar things, would it be possible for CHEBI to either have both role and material entity terms, or at least have a material entities version of all vitamin terms (with subclass relations between them)?
I totally agree that vitamins are chemicals therefore material entities and so they would all be considered first as entities (and then we can assign them a role)
Would it be reasonable to propose only is_a relationships in ChEBI to merely represent the vitamers and their grouping classes (e.g. B vitamins) and then include specific biochemical functions/roles within CDNO's developing dietary function branch?
I think this could help to clarify the "coverage" of respective hierarchies in CheBi and CDNO
As far as I understand the previous discussion, the "dietary function branch" is like a hierarchy for the different roles the chemical substance (or entity) can have in the body (for example, a vitamin role) The term vitamin A refers to retinol and its metabolic derivatives (see figure below). I also want to add that carotenoids are not considered to be vitamins so far but this could evolves in the future (notably as lutein and zeaxanthin seems to have a protective role towards the retina).
As far as I know, we speak of "vitamer" in the specific case of vitamin E and vitamin B.
I totally agree that vitamins are chemicals therefore material entities and so they would all be considered first as entities
👍
As far as I understand the previous discussion, the "dietary function branch" is like a hierarchy for the different roles the chemical substance (or entity) can have in the body (for example, a vitamin role)
Yes perhaps a dietary function branch (or similar) is future addition to CDNO. But the priority for now is to have a consistent set of imported CHEBI terms to make the nutritional component hierarchy (a simplified representation of the chemicals in CHEBI that are of interest to nutritionists).
Indeed. As outlined in the original scope: “Additional knowledge relating to interactions of food components with human physiological function is embodied within the CNDO dietary function class” We would also anticipate a ‘physical attribute’ class (for terms such as energy, specific gravity) and ‘functional attribute’ class (for eg calculated activities)
From: Kai Blumberg @.> Sent: Monday, 24 May 2021 8:39 PM To: Southern-Cross-Plant-Science/cdno @.> Cc: Graham King @.>; Mention @.> Subject: Re: [Southern-Cross-Plant-Science/cdno] Definition of vitamins and vitamers in CHEBI and CDNO (#57)
I totally agree that vitamins are chemicals therefore material entities and so they would all be considered first as entities
👍
As far as I understand the previous discussion, the "dietary function branch" is like a hierarchy for the different roles the chemical substance (or entity) can have in the body (for example, a vitamin role)
Yes perhaps a dietary function branch (or similar) is future addition to CDNO. But the priority for now is to have a consistent set of imported CHEBI terms to make the nutritional component hierarchy (a simplified representation of the chemicals in CHEBI that are of interest to nutritionists).
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/Southern-Cross-Plant-Science/cdno/issues/57#issuecomment-846952406, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AKTZCMCAK2UO3RPAQJUVMTLTPIUDZANCNFSM44YOXNNQ.
Thanks a lot everybody for the inputs.
I have been trying to contact other collaborators just to have more backup, I will post their responses here once I get their permission.
Meanwhile, I have been thinking about the proposals for CHEBI and what could work best for us, I have to mention though that within the CDNO we need to think about the definition we have for the class 'dietary nutritional component'
which is: "A material entity that is ingested and contributes to survival, growth and development". So, this means we can't have a role in the hierarchy, in case we decide that vitamin A, B, C, D, E and K are roles, right?
Here are the proposals according to what we have discussed so far:
Proposal from CHEBI
Proposal 1 is what everybody is agree, I just have to ask CHEBI team about the last relationship between fat and water soluble vitamin
to vitamin
class, so that's pending, at the moment the relationship is 'has role'.
Proposal 2 is a new way to create the relationships within CDNO, as we can't have roles in this specific class, so we would move vitamin A, B, C, D, E, and K
and the vitamin
class to the dietary function class (in development). We would add instead, the term 'vitamin C vitamer'
for example.
in case we decide that vitamin A, B, C, D, E and K are roles, right?
As @maweber-bia said it would be better to have a material entity version of all vitamin terms. I agree with that, but I think at the very least everything in the DNC hierarchy needs to be a material entity to be correct following BFO. If we only model vitamins as roles they'd need to be in a separate role hierarchy which we could call dietary function
or similar.
I don't understand the above proposals. the has role
relation only applies between a material entity
and a role
class so if you have the axioms L-ascorbic acid has role vitamin C
and vitamin C is a water soluble vitamin
then vitamin C
and water soluble vitamin
both have to be roles. Therefore water soluble vitamin
can't have role vitamin
. Instead water soluble vitamin
would have to be subclass to vitamin
, which it already is in CHEBI:
@LilyAndres you can apply the same logic to figure out what the relationships would be but the basic question is again do you model vitamins as material entities, roles or both. Answering that question is what is needed, then based on it the appropriate terms and relationships can be derived.
pretty sure there will be two branches: one in chebi, for roles, and one in CDNO for vitamers having a vitamin role. Vitamers can still be organized in an is-a hierarchy which more or less mirrors the Chebi vitamin role hierarchy.
Looks like @ddooley is advocating for a CDNO to have a set of vitamers or similar which are the material (or chemical) entities that we think of e.g., L-ascorbic acid
which would have has role relations to the CHEBI vitamin terms. This would make sense if all the CHEBI vitamin terms were in the role hierarchy. Unfortunately at the moment it's a mix. For example vitamin K
and vitamin D4
are entities whereas vitamin
and fat-soluble vitamin
are roles. I think the first order of business is getting/helping CHEBI straighten this out. Then we can fix CDNO accordingly, perhaps offering to host the role classes if desired.
+1 for straightening out ChEBI first. ChEBI needs to decide whether they take seriously the "Biological Interest" portion of their name. Roles would seem to me to be the most appropriate way to model these, and as Kai notes, these are a crap-shoot inside of ChEBI at the moment. It does make sense to me however, that "vitamin K" and "vitamin D4" are chemical entities (defined as entities having precise molecular compositions) whereas "vitamin" and "fat-soluble vitamin" are roles (defined as having classes of biological outcomes that prevent nutritional deficiencies)--we have to allow for the fact that the word "vitamin" means something slightly different in each of these contexts.
I think the goal is to build an evidence-based understanding for when to use each term, I would prefer to see a generalized "bottom-up" approach to modeling all nutrients/bioactive molecules (vitamins, minerals, essential amino acids, essential fatty acids, oligosaccharides, etc) and their roles in terms of the GO molecular functions and GO biological processes in which they engage-- (including annotated for each biological organism consuming these chemical entities, because the same chemicals perform different roles in different organisms, for example: cats require vitamin A in its pre-formed state, but can't make it from beta-carotene the way humans or dogs can).
We generally know enough about these terms that this modeling should be feasible. Disease states for nutrient deficiencies (which is currently how "nutrient essentiality" is established), could and should be defined in terms of the aberrant GO Biological Processes and GO Molecular Functions that result from dietarily deficient states and/or cellular lack of these chemicals (not all deficiencies are a result of a lack of dietary input, e.g. malabsorption).
Currently "nutrients" are only roughly cast within the Disease Ontology: "nutritional deficiency disease (DOID:5113 http://www.disease-ontology.org/?id=DOID:5113)". Related to the current discussion: I believe that vitamin and vitamer roles should be defined only as "exhibiting biological activity against a specific vitamin deficiency". Other roles that these chemical entities (that may have "vitamin" as part of their name) *may play in living organisms are not their role as a vitamin. For example, in the disease hypervitaminosis A https://disease-ontology.org/?id=DOID:9972, Vitamin A is no longer functioning in the vitamin role https://www.ebi.ac.uk/chebi/searchId.do?chebiId=CHEBI:33229, but in a toxin role https://www.ebi.ac.uk/chebi/searchId.do?chebiId=CHEBI:27026. Again, this points to the need to be specific. The role of the chemical entity may play* depends on dosage, timing, organism, absorption, and other factors.
This paper written by an international group https://pubmed.ncbi.nlm.nih.gov/31769838/ can also provide some guidance on material entity vs. roles and how we use the more general terms.
On Tue, May 25, 2021 at 1:09 PM Kai Blumberg @.***> wrote:
Looks like @ddooley https://github.com/ddooley is advocating for a CDNO to have a set of vitamers or similar which are the material (or chemical) entities that we think of e.g., L-ascorbic acid which would have has role relations to the CHEBI vitamin terms. This would make sense if all the CHEBI vitamin terms were in the role hierarchy. Unfortunately at the moment it's a mix. For example vitamin K and vitamin D4 are entities whereas vitamin and fat-soluble vitamin are roles. I think the first order of business is getting/helping CHEBI straighten this out. Then we can fix CDNO accordingly, perhaps offering to host the role classes if desired.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Southern-Cross-Plant-Science/cdno/issues/57#issuecomment-848225919, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAMZGI7UGZU4ZIHJ7CTC5V3TPP7X3ANCNFSM44YOXNNQ .
I fully agree with the views of @mateolan that we should define different roles separately from the chemical entities as the role is a function of the context (species, dose, health status, age, etc.)
I do not see any trouble in defining "vitamin" or "water-/fat-soluble vitamins" as roles if you consider the more specific terms "vitamin X" or "vitamin Y" as entities.
For a wider viewpoint, I also think that we have to consider all the different kind of molecule "superfamilies", namely proteins, lipids, carbohydrates, and polyphenols (that is, main plant secondary metabolites) in a compositional hierarchy rather than a nutritional hierarchy (please see my suggestion in issue # 61
👍 to @mateolan and @maweber-bia's comments. Just reiterating the issue in chebi, several of these vitamin X
terms e.g. vitamin C
are modeled as a role not just the top level classes like water-soluble vitamin
.
I'm not the domain expert on nutrition so I can't comment if role or entity is more appropriate but we need to disentangle this. @maweber-bia is proposing the vitamin X
terms be modeled as chemical entities. That makes sense to me but it's on CHEBI and the nutritionists to agree on something.
I would prefer to see a generalized "bottom-up" approach to modeling all nutrients/bioactive molecules ... in terms of the GO molecular functions and GO biological processes in which they engage.
I think connecting to GO would be a great future direction but not the immediate focus here in CDNO. Although it is good to think about how we can allow for this in future applications importing both CDNO and GO.
Yes, I know of the ChEBI mess, and have been frustrated by it previously. Happy to discuss this with others in live convo. Probably best to get this modeled in a way we are all comfortable, and then propose change rather than waiting on ChEBI for action.
I highlighted the need for GO Molecular Funtion/Biological Process mappings because we need to be sure that whatever model we come up with here will support rather than stymie the coherency of those future mappings...but I do understand that the mappings themselves are outside the scope of current project.
On Wed, May 26, 2021, 07:57 Kai Blumberg @.***> wrote:
👍 to @mateolan https://github.com/mateolan and @maweber-bia https://github.com/maweber-bia's comments. Just reiterating the issue in chebi, several of these vitamin X terms e.g. vitamin C are modeled as a role not just the top level classes like water-soluble vitamin.
[image: image] https://user-images.githubusercontent.com/12255688/119679493-b17ee580-be40-11eb-99dd-e3e7e6a90afb.png
I'm not the domain expert on nutrition so I can't comment if role or entity is more appropriate but we need to disentangle this. @maweber-bia https://github.com/maweber-bia is proposing the vitamin X terms be modeled as chemical entities. That makes sense to me but it's on CHEBI and the nutritionists to agree on something.
I would prefer to see a generalized "bottom-up" approach to modeling all nutrients/bioactive molecules ... in terms of the GO molecular functions and GO biological processes in which they engage.
I think connecting to GO would be a great future direction but not the immediate focus here in CDNO. Although it is good to think about how we can allow for this in future applications importing both CDNO and GO.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Southern-Cross-Plant-Science/cdno/issues/57#issuecomment-848841183, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAMZGI7DPLVS42YSUJBZOLLTPUD7HANCNFSM44YOXNNQ .
Note that in response to a letter outlining the problem by Liliana to CHEBI, their curators have proposed a design revision in CHEBI as noted at start of thread. So here we discuss whether the CHEBI proposal is sufficient, and what the vitamin/er material entity will look like in CDNO. Would it be useful to detail a bit more what CHEBI's response was? One more telecon from the CDNO side sounds appropriate to summarize and finalize a response.
oh, Thanks Damion for the clarification, I did not see the link in the opening comment from Liliana
I now see that the Chebi proposal seems ok
assign 'is a' vitamin to both fat and water soluble vitamin
assign the 'has role' relationship to all vitamers (in their case vitamin) molecules.
but in that case, will the single molecules also be considered as chemical entities in another is-a branch ?
Only those molecules who have an evidenced vitamin role would then have 'has role' relationship with a subclass of fat-soluble or water-soluble vitamin depending on their chemical nature (by the way, I do not think the term 'vitaminer" is appropriate for all the molecules in the list)
On the other hand, can we then have the CHEBI:33229 - vitamin also a subclass of "micronutrient" as I mentioned in my proposal in #61 ? This is to refer to the need for the consumer perspective (nutrition facts on the packaging of the product), as discussed earlier
Thanks @ddooley, here is the original email from Adnan with my suggestion and question:
Lili: We think that all vitamins should be chemical entities, is CHIRO still on the scope to add roles for the chemical entities?
Adnan (CHEBI team): I am not sure about CHIRO since its something that Chris Mungall has been working on and I don't think it will be integrated into ChEBI anytime soon.
With regards to your vitamin issue, clearly it would be best if both 'water-soluble vitamin' and 'fat-soluble vitamin' could be classified in the same way. After discussing this issue with some former colleagues, our proposed solution would be to change the current ontology relationships of Vitamins A, D, E and K from 'has role fat soluble vitamin' to 'is a fat soluble vitamin' and the child classes of these vitamins which are the relevant chemicals to have 'has role' relationship to the vitamins (and the relevant chemicals be classified based on their structures).
E.g. all-trans-3,4-didehydroretinoic acid (CHEBI:133794) has role Vitamin A. Vitamin A (CHEBI:12777) is a fat-soluble vitamin.
If you have any objections to this, then please do let me know.
I think in this email CHEBI team is keen to hear other proposals.
Thanks, @mateolan really good points that we will consider.
@maweber-bia
but in that case, will the single molecules also be considered as chemical entities in another is-a branch ?
Exactly this is what I thought we were discussing, it is not clear in the email from the CHEBI team if they will have vitamins in both branches (roles and chemical entity). I will send an email to Adnan to clarify this, I will let you know how it goes and maybe after that we can have a call?
On the other hand, can we then have the CHEBI:33229 - vitamin also a subclass of "micronutrient" as I mentioned in my proposal in #61 ? This is to refer to the need for the consumer perspective (nutrition facts on the packaging of the product), as discussed earlier
@maweber-bia thanks, we have reached Adnan before, he actually helped to update the definition of 'micronutrient' [CHEBI:27027], but micronutrient is a role in CHEBI as well, so we just have to see how the vitamins would be arranged and then we can suggest other relationships?
About "it is not clear in the email from the CHEBI team if they will have vitamins in both branches (roles and chemical entity)" - what is certain is that the same entity can't be in both. So a naming convention should visually distinguish respective entities.
Ah...this is making more sense to me now...My email chain started with "Thank you all for your inputs. @mateolan https://github.com/mateolan thoughts?" so I was unclear from the get-go, and only started chiming-in once I had a general sense of approach. That will teach me to reach back into the Issues link next time I get pinged. Going forward with this issue, I think it might be most efficient to have a call with ChEBI, and discuss these issues real-time. Now that we have laid out the issues, I think we could reach a resolution in a discussion faster than we could type more emails back and forth...side-note, it looks like CHIRO is importing GO terms to be used as roles--which is pretty much what I was saying...I do fear that having vitamin roles within ChEBI is potentially confounding because as I said, vitamins roles are organism specific and as far as I can tell, these roles are not being associated with organisms. ChEBI advertises itself as having info about "...products of nature or synthetic products used to intervene in the processes of living organisms." If others have a way for parsing/signifying organism specific roles, I am all ears. Aside from adding organism specificity, I don't have any problem with your modelling Lilli. I am happy to be part of that ChEBI call if you think it would be helpful.
On Thu, May 27, 2021 at 5:51 PM LilyAndres @.***> wrote:
Thanks @ddooley https://github.com/ddooley, here is the original email from Adnan with my suggestion and question:
Lili: We think they all vitamins should be chemical entities, is CHIRO still on the scope to add roles for the chemical entities?
Adnan (CHEBI team): I am not sure about CHIRO since its something that Chris Mungall has been working on and I don't think it will be integrated into ChEBI anytime soon.
With regards to your vitamin issue, clearly it would be best if both 'water-soluble vitamin' and 'fat-soluble vitamin' could be classified in the same way. After discussing this issue with some former colleagues, our proposed solution would be to change the current ontology relationships of Vitamins A, D, E and K from 'has role fat soluble vitamin' to 'is a fat soluble vitamin' and the child classes of these vitamins which are the relevant chemicals to have 'has role' relationship to the vitamins (and the relevant chemicals be classified based on their structures).
E.g. all-trans-3,4-didehydroretinoic acid (CHEBI:133794) has role Vitamin A. Vitamin A (CHEBI:12777) is a fat-soluble vitamin.
If you have any objections to this, then please do let me know.
I think in this email CHEBI team is keen to hear other proposals.
Thanks, @mateolan https://github.com/mateolan really good points that we will consider.
@maweber-bia https://github.com/maweber-bia
but in that case, will the single molecules also be considered as chemical entities in another is-a branch ?
Exactly this is what I thought we were discussing, it is not clear in the email from the CHEBI team if they will have vitamins in both branches (roles and chemical entity). I will send an email to Adnan to clarify this, I will let you know how it goes and maybe after that we can have a call?
On the other hand, can we then have the CHEBI:33229 - vitamin also a subclass of "micronutrient" as I mentioned in my proposal in #61 https://github.com/Southern-Cross-Plant-Science/cdno/issues/61 ? This is to refer to the need for the consumer perspective (nutrition facts on the packaging of the product), as discussed earlier
@maweber-bia https://github.com/maweber-bia thanks, we have reached Adnan before, he actually helped to update the definition of 'micronutrient' [CHEBI:27027], but micronutrient is a role in CHEBI as well, so we just have to see how the vitamins would be arranged and then we can suggest other relationships?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Southern-Cross-Plant-Science/cdno/issues/57#issuecomment-850035660, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAMZGI5MKH3P55XWHOTT2ADTP3SJXANCNFSM44YOXNNQ .
what is certain is that the same entity can't be in both. So a naming convention should visually distinguish respective entities.
Yes @ddooley, I think this is why you mentioned we could have:
2) With CHEBI offering terms like "B vitamin role", then we would have space to create a class "B vitamin" which is defined as a compound which has a "B vitamin role". But if CHEBI won't add "role" to "B vitamin" etc., then we need a workaround to avoid duplicate terms in OBOFoundry, OLS etc.
@mateolan I'm sorry I probably didn't explain well, thanks a lot for your comments. We just have to be all on the same page so it's easier for CHEBI and CDNO to represent the information.
I will ask Adnan from CHEBI team if he can have a call with us, otherwise, I'm afraid that the interaction with CHEBI has to be via email. CDNO is happy to have a call though, but I want to be prepared depending on what Adnan says, so we can all work on a final proposal based on what CHEBI would do or would not do
Dear all, thanks for the clarification.
I will wait and see, but I do think that we have to make the distinction clear between roles and the molecular entities arranged by chemical structures. As proposed by Damion, a naming convention should visually distinguish respective entities, and perhaps 'vitaminic compound' or 'vitaminic component' could be used for that purpose in the compositional/nutritional branch ?
Just to follow this up, this is the email I sent to Adnan Malik from the CHEBI team:
Dear Adnan
We have one question about the example that you elaborated:
E.g. all-trans-3,4-didehydroretinoic acid (CHEBI:133794) has role Vitamin A. Vitamin A (CHEBI:12777) is a fat-soluble vitamin.
At present, it appears that vitamin C and B are defined under biological role. Are you proposing that vitamin A, B, C, D, E and K also be under the branch "role"?
Are you proposing that the terms "water soluble vitamin' and 'fat soluble vitamin' remain under the branch biological role?
The consensus that is emerging suggests that "water soluble vitamin' and 'fat soluble vitamin' be associated with chemical entity along with specific vitamin related chemicals such as L-ascorbic acid and the terms vitamin and e.g. vitamin C be retained under role.
I think it will help if we can have a video call with you, including the domain experts and the ontologist helping in the CDNO project. Do you think you are able to join, so we can discuss further? We would really appreciate it.
Here is Adan's reply:
Yes, ChEBI's proposal is to move all of the vitamins (A, B, C, D, E, and K)
under the biological role branch and fat/water soluble vitamin
will also remain under the role branch.
Each vitamer (a particular chemical) will have its own set of structural terms under the molecular structure branch along with the _hasrole relationship to the relevant vitamins.
E.g. L-ascorbic acid is_a ascorbic acid which is_a ketoaldonic acid
L-ascorbic acid has_role vitamin C which is_a water-soluble vitamin which is_a vitamin
L-ascorbic acid will also have additional roles such as has_role cofactor, has_role food antioxidant etc.
The major issue that we will encounter if we decide to move all of the vitamins under the molecular structure branch is that all of the chemicals which fall under B vitamin (CHEBI:75769)
are structurally diverse set of compounds so it would be difficult to group them together unless we use a very vague term such as 'organic molecular entity'. On the other hand chemicals which fall under the class 'vitamin E' can easily be grouped together since the chemical structures are quite similar and belong to the class called 'chromanols'.
I am happy to discuss this in a call. I am free this Friday or anytime next week, let me know what day and time suits you best?
Knowing this might be a bit easier to have the meeting with Adnan
Great ! The answer of Adnan seems engaging!
However, I am wondering why
if we decide to move all of the vitamins under the molecular structure branch is that all of the chemicals which fall under
B vitamin (CHEBI:75769)
are structurally diverse set of compounds so it would be difficult to group them together
As far as I understand, if each vitamer (a particular chemical) will have its own set of structural terms under the molecular structure branch along with the has_role relationship to the relevant vitamins, we do not need to group them together ??
We will just need to add them the role(s) according to their biological activity ??
Thank you for the clarification!
I'm sorry for this long comment but I think it's important to have this email that Jayne Ireland and Anders Møller from the European Food Information Resource (EuroFIR) sent on 05/17/2021 to the CDNO team.
@maweber-bia maybe what Jayne mentions about vitamin B in the comment below is what Adnan refers to "are structurally diverse set of compounds"?
In our opinion this issue stems back to the original definitions of “vitamins” being “something” having an activity, i.e. a (beneficial) physiological impact, on a living body. As such, vitamins are not necessarily chemical compounds in their original sense, but concepts describing certain effects.
This means that for nutritional components – as we call them in EuroFIR – we have a mixture of concepts/lumps of compounds, actual chemical entities and calculated activities of these components, e.g. the vitamins. CHEBI is, as its name says, a database of chemical entities of biological interest, and for this reason we did have some issues when creating the EuroFIR Component Thesaurus as many of our nutritional components did not have a counterpart in CHEBI (see above).
As indicated above, we would say that vitamins, like dietary fibre and protein, are not specific compounds, but rather concepts based on nutritional roles. Vitamers, on the other hand, correspond to the specific compounds that are measured by analysis; their values are used to calculate vitamin activity in food products. The CHEBI database contains information about specific compounds, thus its definitions can be used for vitamers, but not necessarily for vitamins. The vitamins – or components for that matter - that appear in the human/animal body and in food/feed, are very/most often not the components, determined analytically, but rather calculated.
Some examples to show the vitamin concepts can be as follows:
vitamin A <is a> fat soluble vitamin
vitamin A vitamer <has role> vitamin A
All-Trans-Retinol <is a> vitamin A vitamer
Carotenoid <has role> Vitamin A
– but we would not call carotenoid a vitamer, because it is not a specific molecule; it is a class of molecules
Beta-carotene <is a> carotenoid
Beta-carotene <is a> vitamin A vitamer
There is no such thing as vitamin B, it should either be called “vitamin B complex” as a group – or better, just listed as the vitamin B1, vitamin B2, vitamin B3 (term does not exist/is not used in many countries – it is usually named niacin), etc.
vitamin B1 <is a> water soluble vitamin
vitamin B1 vitamer <has role> vitamin B1
thiamine(+1) <is a> vitamin B1 vitamer
– however, it does not exist as such, as it is an ion; it is analytically usually determined as thiamine(chloride) hydrochloride, but usually named “thiamine”
thiamine monophosphate< is a> vitamin B1 vitamer
thiamine pyrophosphate <is a> vitamin B1 vitamer
etc.
Good info I agree with most of it, however here are points of disagreement for me:
we would not call carotenoid a vitamer, because it is not a specific molecule; it is a class of molecules
A class of molecules is something that belongs within an ontology, this is what enables us to have a subclass hierarchy where specific molecules are children to more general classifications like protein
.
The CHEBI database contains information about specific compounds, thus its definitions can be used for vitamers, but not necessarily for vitamins.
That's not the case, CHEBI contains information about 1) specific molecules with a formula, e.g. thiamine(1+) monophosphate, 2) more general classes of molecules e.g., protein def~ A biological macromolecule minimally consisting of one polypeptide chain synthesized at the ribosome.
, and 3) role
classes which are dependent on other compounds to be realized e.g., fat-soluble vitamin.
@ddooley and Adan have been arguing that vitamins should be CHEBI roles which to me seems to fit with the Jayne Ireland and Anders Møller's email.
thanks @kaiiam, I think @mateolan and @laurenechan were in favour for this?
@maweber-bia, @LuciaSegovia what you think?
I think the vitamin roles could be in CDNO 'dietary function/dietary role' class.
I think the vitamin roles could be in CDNO 'dietary function/dietary role' class.
I though Adan was proposing to have them in CHEBI? Do you mean we import them from CHEBI into the CDNO 'dietary function/dietary role' hierarchy? Or do you want to try and get CHEBI to seed this material to CDNO. I'm move in favor of CHEBI having it (to be more generally applicable) but CDNO importing and simplifying it for nutritionists. CHEBI is still upstream of, and will likely always be more widely used than CDNO.
@LilyAndres Hopefully I'm interpreting this correctly, but I am most interested in having chemical entities with is_a relationships used across the board for vitamins (e.g. L-ascorbic acid is_a vitamin C), primarily because I am concerned with vitamin C (or other broader vitamin classes) being considered a role. Vitamins (e.g. vitamin C) are likely currently and will continue to be referenced in other ontologies and term development patterns as a chemical entity. If we make it exclusively a role, I foresee misuse and headaches for users (e.g. someone tries to make a term for a vitamin C supplement containing unknown vitamin C compounds, they include the role vitamin C as part of the logical axiom as opposed to a material entity of vitamin C). I think @ddooley had mentioned previously an interest in having both a branch with is_a relationships, and then a separate branch with designated role terms with has_role relationships. Is this still in discussion or are we just looking to maintain one branch with a mixture of is_a and has_role relationships?
Vitamins (e.g. vitamin C) are likely currently and will continue to be referenced in other ontologies and term development patterns as a chemical entity.
Probably true and I can see this being easier from a developers perspective e.g. for DOSDP patterns etc.
are we just looking to maintain one branch with a mixture of is_a and has_role relationships?
We definitely don't want to mix roles and entities in the same hierarchy that'll create really nasty reasoning errors for anyone importing CDNO. I think the idea currently is to have the material entities link off to the role hierarchy via the has role relations e.g., L-ascorbic acid has role vitamin C
.
We definitely don't want to mix roles and entities in the same hierarchy that'll create really nasty reasoning errors for anyone importing CDNO. I think the idea currently is to have the material entities link off to the role hierarchy via the has role relations e.g.,
L-ascorbic acid has role vitamin C
.
Great, so then seconding @ddooley again, I think we still need ChEBI to differentiate the role terms (e.g. vitamin C role) from the entity terms (vitamin C) to make it blatantly clear what use cases each term type can satisfy.
@LilyAndres I liked what you suggested and it makes complete sense to me, but I can see also @laurenechan point about the "current use" of the terminology. And people in the nutrition field will be using vitamin C instead of L-ascorbic acid. Again, it might be different if you take other vitamins on the list.
But, I think I support the idea of having them as role and as entity, but I have some questions. What would it happen with "vitamin B"?, could we subtitute it by the vitamins names (thiamin, niacin, riboflavin, vit.b12, etc.)? Maybe, it's my poor understanding of the ontology world but, seeing vitamin B side-by-side with the vitamin C makes me uncomfortable.
Thank you for sharing all these views
I've just been realizing the complexity of the topic when I read the email from Jayne Ireland and Anders Møller with respect to vitaminers.
It is not clear to me what a vitaminer is, but I think this term only applies to some compounds of the vitamin B group and vitamin E group but not to all vitaminic groups - and it is clear that vitamin B is a complex group!
I am ok for these statements :
1a) L-ascorbic acid
2a) Beta-carotene has_role vitamin A which is_a fat soluble vitamin which is_a vitamin...which
but I do not agree with this statement because not all carotenoids have necessarily a vitamin A role: Carotenoid has_role Vitamin A
I supported the idea of considering vitamins as chemical entities, but this seems complicated given the current use in CheBI.
I am afraid that we will have the same difficulties with protein being both a molecular entity and a role:
3a) protein
Definitely, all roles need to be clearly distinguished from entities.
It seems that this fits with CheBI structuration after moving all of the vitamins (A, B, C, D, E, and K) under the biological role branch and fat/water soluble vitamin will also remain under the role branch. But again, if the nutritionists have another point of view, I am not ultra qualified to make the choice.
Ok I think there are 4 issues we can discuss with Adnan during the meeting.
Great, so then seconding @ddooley again, I think we still need ChEBI to differentiate the role terms (e.g. vitamin C role) from the entity terms (vitamin C) to make it blatantly clear what use cases each term type can satisfy.
vitamin C
for vitamin C role
as shown in the image below. I'm not quite sure how would it work for vitamin B, so at the moment I just did what Jayne from EuroFIR suggested and added 'vitamin B complex role'
. What you think?'dietary role'
class in CDNO, I made an issue #65 to show a definition for the new class, please have a look.
Does everybody agree on this? What would it happen with "vitamin B"?, could we subtitute it by the vitamins names (thiamin, niacin, riboflavin, vit.b12, etc.)?
vitamin B
or vitamin B1
? I can see that 'thiamine(1+)' [CHEBI:18385]
has the synonym 'vitamin B1'
, but as Jayne mentions:"There is no such thing as vitamin B, it should either be called “vitamin B complex” as a group – or better, just listed as the vitamin B1, vitamin B2, vitamin B3 (term does not exist/is not used in many countries – it is usually named niacin), etc." in this case thiamine or 'thiamine(1+)' instead of vitamin B1. What you think?
'vitamin C'
, 'vitamin K'
, 'vitamin B1'
, 'vitamin B2'
, etc as material entities, could you please make a suggestion on how can we incorporate them in CHEBI? I'm struggling a bit on how could we add this in CHEBI. Hello, we had a meeting with Adnan from the CHEBI team, here are the agreed actions from CHEBI.
Actions in CHEBI
For vitamin molecules (vitamers)
CHEBI will include the vitamin molecules under the chemical entity hierarchy
-such as L-ascorbic acid
(see image below)
-e.g.vitamin C
, vitamin D
, vitamin K
... also included in the class and/or as a synonym? (Adnan will discuss with CHEBI team and clarify)
For each of these molecules CHEBI will include a “has role” relationship to the vitamins (see image below).
For vitamins as roles
L-ascorbic acid
has role vitamin C role
(see image below)B vitamin group
A B vitamin role
for each specific B vitamin
would be made available in ChEBI (e.g. vitamin B1 role
, vitamin B2 role
…). These would likely be made exact synonyms with the terms 'Thiamin role'
and 'Riboflavin role'
for example (see image below).
An overarching “B vitamins” or “B vitamin complex” term would be made available in ChEBI as a grouping class, having category groups for each specific vitamin as well, with the distinct molecular entities see example:
B vitamin complex
└───vitamin B1
│ │ thiamine(1+)
└───vitamin B2
(not integrated in the image below)
Please note that the image above doesn't shows the classification of vitamins A
, D
, B
, C
, etc. as chemical entities because Adnan will clarify if they will be as a class and/or as a synonym.
I hope this is clear, but if not please let me know so we can explain better.
Now, here are the actions in CDNO
'dietary nutritional component'
class
-Under this have ‘vitamin C’ etc from CHEBI (see example as if it's determined as a class)dietary nutritional component
└───vitamin molecule
│ │ vitamin C (if it is determined as a class)
│ │ │ L-ascorbic acid
‘dietary role’
class (ongoing in #65)
-Under this include ‘human dietary role’
-Then ‘vitamin C role’ etc inherited from CHEBI
dietary role
└───human dietary role
│ │ vitamin role
│ │ │ water soluble vitamin role
│ │ │ │ vitamin C role
I hope this is clear, if not please let me know.
Thanks for the update.
Hello, here is an update for this issue:
'vitamin' [CHEBI:33229]
has now changed to 'vitamin (role)' [CHEBI:33229]
'vitamin C, B, D, etc'
have been integrated as material entities which have a 'has role' relationship to 'fat-soluble vitamin (role)' [CHEBI:24020]
and 'water-soluble vitamin (role)' [CHEBI:27314]
'fat-soluble vitamin (role)' [CHEBI:24020]
and 'water-soluble vitamin (role)' [CHEBI:27314]
are now roles and have a "(role)" in their label.'vitamin B5, B1, B12'
, etc. have a 'has part' relationship with 'vitamin B complex'
Here is a small example:
The terms 'vitamin C (role)', 'vitamin A (role)', 'vitamin K (role)', etc. are not described in CHEBI but we can probably create these terms in CDNO, either in the 'dietary role'
class issue #65 or 'nutritional functional attribute'
issue #66.
What do you think about these changes? Is it something that will work for you? @ddooley, @mateolan, @LCCarmody, @maweber-bia, @laurenechan
Ok with these changes, with the caveat that these roles are limited to the human organism, or otherwise organismically specified.
On Tue, Aug 31, 2021, 06:30 Liliana Andres @.***> wrote:
Hello, here is an update for this issue:
- The label 'vitamin' [CHEBI:33229] has now changed to 'vitamin (role)' [CHEBI:33229]
- 'vitamin C, B, D, etc' have been integrated as material entities which have a 'has role' relationship to 'fat-soluble vitamin (role)' [CHEBI:24020] and 'water-soluble vitamin (role)' [CHEBI:27314]
- Notice that 'fat-soluble vitamin (role)' [CHEBI:24020] and 'water-soluble vitamin (role)' [CHEBI:27314] are now roles and have a "(role)" in their label.
- 'vitamin B5, B1, B12', etc. have a 'has part' relationship with 'vitamin B complex'
Here is a small example:
[image: image] https://user-images.githubusercontent.com/25215773/131441008-2daf85f9-3404-41df-a761-e205f5862151.png
The terms 'vitamin C (role)', 'vitamin A (role)', 'vitamin K (role)', etc. are not described in CHEBI but we can probably create these terms in CDNO, either in the 'dietary role' class issue #65 https://github.com/Southern-Cross-Plant-Science/cdno/issues/65 or 'nutritional functional attribute' issue #66 https://github.com/Southern-Cross-Plant-Science/cdno/issues/66.
What do you think about these changes? Is it something that will work for you? @ddooley https://github.com/ddooley, @mateolan https://github.com/mateolan, @LCCarmody https://github.com/LCCarmody, @maweber-bia https://github.com/maweber-bia, @laurenechan https://github.com/laurenechan
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Southern-Cross-Plant-Science/cdno/issues/57#issuecomment-908890226, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAMZGIZ55GLFKUAG4MEX64DT7RLG3ANCNFSM44YOXNNQ .
CHEBI has now updated its vitamin branches - role and material entity. Today OntoBee/OntoFox did a refresh of CHEBI. Can CDNO redo its import of CHEBI and then FoodOn can import that and see where we stand with changes for some role-based items.
@ddooley thanks, I will be importing the rest of the vitamins by the end of the week and making a new release. I just need to touch base on the discussion we had during the last IFOW meeting, I think we will create the 'dietary role' class and then if some other ontology wants to have those terms (maybe ONS), they can incorporate them in the other ontologies, how that sounds?
Just ran the CDNO's chebi import see https://github.com/Southern-Cross-Plant-Science/cdno/pull/77 for @LilyAndres to proceed and help fix @ddooley's issue.
Yes, sounds good. "dietary role" overlaps a few ontologies, and no surprise that it could be introduced via CDNO.
I added the major classes for the vitamins, from here I will be adding more and more vitamins. Thanks everyone for your help and knowledge, should I close the issue now?
Sounds good! Any new issues can have their own threads.
This issue has been created to discuss the relationships of vitamins and their vitamers in CDNO and CHEBI.
1. We have ask CHEBI team, specially @amalik01 for clarification about the relationships described in CHEBI. One issue raised is the different relationships assigned to fat and water soluble vitamins. For example:
The
‘B vitamin’ [CHEBI:75769]
and‘Vitamin C’ [CHEBI:21241]
have the ‘_isa’ relationship with ‘water-soluble vitamin’[CHEBI:27314]
.In contrast
‘Vitamin A’ [CHEBI:12777]
,‘Vitamin D’ [CHEBI:27300]
,‘vitamin E’ [CHEBI:33234]
and‘Vitamin K’ [CHEBI:28384]
have the ‘has_role’ relationship with‘fat-soluble vitamin’ [CHEBI:24020]
.The term vitamer is not defined as such in CHEBI. However, some vitamer molecules have been assigned to specific vitamins with:
‘B vitamin’ [CHEBI:75769]
and‘vitamin C’ [CHEBI:21241]
and the rest with
See the following link:
The proposal from CHEBI is to
assign 'is a' vitamin to both fat and water soluble vitamin
assign the 'has role' relationship to all vitamers (in their case vitamin) molecules.
2. Within CDNO we propose to refine/extend the relationships, re-using where possible the CHEBI definitions. Thus: (reading the relationships right to left)
3. We propose establishing definitions/relationships for vitamin roles within the
‘dietary_function’
class of CDNO.A separate thread/issue will be generated to discuss this class.
What do you think?