Integration of ONS, FOBI and ONE

[cyang0128]

Dear @FrancescoVit @pcastellanoescuder, after a nice discussion with @ddooley, we plan to introduce the ontology hierarchy shown above. Both 'dietary survey method' (ONE) & 'food biomarker analysis' (FOBI) are research methods used in nutritional research. in another word, 'nutritional research method (ONS)' is a parent class of both terms. Looking forward to your comments regarding how to arrange the terms in our ontologies. thanks!

[FrancescoVit]

Hi @cyang0128 and hi all. I had problem joining the call on last Wednesday, so I do not know if that matter was covered or not.

Anyhow, thank for thinking of ONS. I agree that the term would fit in ONS, as a way of connecting FOBI and ONE with also the "root" of it all. I'm going to open an issue on the ONS tracker, and discuss a possible definition, and best super-class.

[pcastellanoescuder]

Hi all,

I think this hierarchy is great! However, neither the term "food biomarker analysis" nor any similar term appears in FOBI. I’m not sure if FOBI should include this term or not, as FOBI is not intended to describe or define food biomarker assays (or assays in general), but provides the terms used in these type of analyses.

So I’m not quite sure where that term should belong... Any thought on that?

Thanks!

[ddooley]

Chen had initially proposed "biomarker analysis" as the process FOBI was generally covering, I had mentioned that was a bit broad, as FOBI was more limited to food related biomarker analysis. I see in OLS that NCIT has a term "Biomarker Analysis" that could be seconded (reappropriated) for this, or we just go ahead and create "Food biomarker analysis" in ONS directly if FOBI wants to avoid it!

[FrancescoVit]

I had the occasion of looking a bit to FOBI and ONE and in general to the issue (for reference purpose between repos https://github.com/enpadasi/Ontology-for-Nutritional-Studies/issues/10),

I see the point of @pcastellanoescuder . It is quite hard to fit all FOBI under a single top-term, and surely this would not fit under the "assay" or "invetigation" class. For example, i see that the two top-classes are "biomarkers" and "foods"; none of them could go as sub-classes of some sort of assay. Moreover, most of the subclasses of "food" are taken from FOODON, so you would also state that those classes in FOODON are "assays" (so, going to the root, they are "planned process"), while they are in fact "continuant" and not "occurrent". So like this we risk of basically mess quite some axioms.

If I got it right, FOBI has the scope of 1) classification of food biomarker (connection to CHEBI), 2) connection of the chemical biomarker chemical entity to the food containing it (using "Biomarkerof" property); 3) avoid re-definition of food, using FOODON. There seems to be no classes pertinent to "occurrent" or generally process/assays, all of FOBI has its scope and range of action in the "continuant", and specifically "material entity", domain. Please, correct me if I got it wrong. Iit does not cover the technological aspect of a biomarker analysis, but we could think that classes in FOBI under the "biomarker" class could be connected to some assay like HPLC or maybe GC-MS on serum/urine/whatever.

If I got it right, ONE has the scope of 1) classification of the methods for nutritional surveys; 2) quality evaluation of the surveys and method, but also quality and minimum reporting for publication

The only way I see, is eventually to work on connecting the terms via axioms, and maybe the integration should be at the "data item" level. I'll try to elaborate a bit on that:

• On a nutritional study perspective, multiple assays can be performed on the same set of subjects. To some extent, ONS exist also to provide a way of representing this (but likely we are far away from completeness). In this regard, I think that the integration between ONE and FOBI has possible application, and here it may be make more sense: on the same set of subjects a nutritional survey was performed (ONE) and biomarker of the food consumed (FOBI) were assessed in the assay XXX (ONS/OBI) . The purely conceptual integration, trying to find a coordinated hierarchy other than the BFO, however, is difficult and potentially wrong.

Some generally nice classes for integration, for example, could be ONE:0002019 ("biomarkers section") or ONE:0002012 ("data measurement section"), but not in the current form, as the classes are pertinent to "document part" under the "information content entity". The integration would be at the analysis level, or at least at the assay level. At any rate, I don't think that hierarchical relation of super and sub-class is the best here.

At the very end, conceptually, OBI/ONS tell that a certain study was done (i.e. a group of subject has received an apple) and optimally how. Among the various possible assays, ONE is telling how the nutritional survay or other nutritional epidemiology was performed, and how this should be reported in a manuscript. Among the various possible assays, FOBI is telling us that if you find the biomarker XYZ you know that an apple was consumed (and reversely, that apple consumption result in finding the XYZ marker). A possible integration would be: nutritional survey is reporting that an apple was consumed (ONE), but can I find the XYZ marker (FOBI)?

Hope to have expressed some thoughts in a right and complete way. Got a bit carried away in the flow of thoughts, and not always it is easy to write them down while they are forming.

All of that said, as @ddooley suggested, I think that ONS could host a "Food biomarker analysis" class, which would inevitably have a rather generic definition. Then the "problem" would eventually move to ONE and to FOBI, and how they would be relating to this class.

[cyang0128]

@FrancescoVit thanks a lot! I also took some time to read the publication introducing FOBI. "Food biomarker analysis methods" were mentioned in the article, although they are not included as ontology terms in FOBI. In another word, this term, as well as its sub-classes (i.e. more specific methods), are specific terms regarding food biomarker data descriptions, rather than general (upper-level) terms for nutrition studies. If I am a user and I am going to annotate my food biomarker data, I would prefer to find the term from a specific ontology about food biomarkers. Just my personal comment, it could be wrong :)

But I also think that ONS could host the term, as @ddooley suggested.

[pcastellanoescuder]

Hi @FrancescoVit, @cyang0128, thanks for your detailed comments.

You got it perfectly. Regarding your comments, I think it's not a good idea to put FOBI classes under any methodological class such as HPLC, GC-MS or NMR, as the associations between metabolites and foods are not affected by the methodology used. Another point would be to put FOBI classes under the type of biofluid analyzed (urine, blood, serum, feces, etc.), since relationships between foods and metabolites can be altered depending on the biofluid analyzed. Currently the links defined in FOBI are generic but refer mainly to urine, and secondly to blood. The creation of a “biofluid” class could be explored to designate different markers. But I cannot yet see clearly where it should go and how it can be used to integrate FOBI, ONS and ONE.

I think the conceptual integration: "nutritional survey is reporting that an apple was consumed (ONE), but can I find the XYZ marker (FOBI)?" is clear and correct. However, I still think that FOBI should not incorporate any "assay" class.

Finally, although ONS incorporates this new class, I still fail to see which is the best way to integrate the three ontologies.

[cyang0128]

thanks @pcastellanoescuder, I understand your decision. @FrancescoVit would you mind holding the class (i.e. food biomarker anaysis method) in ONS?

[FrancescoVit]

Double posting from the ONS issue tracker:

So:

label: Nutritional research assay (ONS_XXXXXX) superclass: OBI:OBI_0000070 ("Assay") definition: Any type of assay that can be performed in the context of a nutritional study

label: Food biomarker analysis assay (ONS_YYYYYY) superclass: ONS_XXXXXX definition: An assay that is aimed at measuring the quantity of known (or unknown) food consumption biomarker in biological fluids (e.g. urine, blood, serum).

Problem I see, the ambiguity between "biomarker" and "metabolite". I hope that @pcastellanoescuder could contribute on this aspect. Thinking out loud, and have not checked if something like this is in FOBI or not, it seems almost unavoidable to also create:

label: food consumption biomarker (ONS_ZZZZZ or FOBI) superclass: unsure, some continuant I would say definition: A chemical analyte measured in a biological fluid (e.g. urine, blood, serum) originated as a result of the consumption and metabolism of a specific food by human.

how does that sound?

[pcastellanoescuder]

Hi @FrancescoVit,

I agree with the label "Nutritional research assay". However I have some comments on the labels "Food biomarker analysis assay" and "food consumption biomarker".

Regarding label "Food biomarker analysis assay (ONS_YYYYYY)", I think the structure is great but I there are two possible scenarios that can lead to different issues. The problem is how to define what is a "Food biomarker analysis assay".

• The first scenario occurs when the aim of the assay is to discover new biomarkers. Then, the definition can be something like: "An assay that is aimed at measuring the quantity (or presence) of food related metabolites in a biological sample (e.g. urine, blood, serum, feces) in order to identify/discover new food intake biomarkers". Note that food related metabolites are potential FIBs (Food Intake Biomarkers), as metabolites only become biomarkers when they meet different conditions like specificity, reproducibility, robustness, etc. Thus, when you are looking for new biomarkers I find reasonable to say that you are quantifying food related metabolites, not biomarkers (that comes later).

• Another scenario could be the measurement of compounds that have already been labeled as "biomarkers" (I think this is the scenario you had in mind). Then, the definition can be something like: "An assay that is aimed at measuring the quantity (or presence) of food intake biomarkers in a biological sample (e.g. urine, blood, serum, feces) in order to evaluate the consumption of defined foods".

This leads me to wonder if it is feasible to create two different classes (for discover and validation) under "Food biomarker analysis assay" class called "Food biomarker discovery analysis" and "Food biomarker validation analysis", or so.

On the other hand, I think that FOBI can consider to include the label "Food Intake Biomarker" (or "potential Food Intake Biomarker"), not as a new class but as a substitution of the existing class "Biomarkers (FOBI:01501)". I think "Food Intake Biomarker" term is more appropriated and specific, since "biomarker" can also be a gene, protein, etc.

What do you think?

[FrancescoVit]

Hi @pcastellanoescuder ,

it is not my field of experience, so your view on those term is central to take the most out of them. I am happy that you more or less agree, and I understand your integration on the classes. I think we can easily solve with a more general super-class and two sub-classes for discovery and validation.

So, to create a more general superclass

label: Nutritional research assay (ONS_XXXXXX) superclass: OBI:OBI_0000070 ("Assay") definition: Any type of assay that can be performed in the context of a nutritional study

label: Food biomarker analysis assay (ONS_YYYYYY) superclass: ONS_XXXXXX definition: An assay that is aimed at discovering new potential metabolites as food consumption biomarker, or to measure/validate known food consumption biomarker in biological fluids (e.g. urine, blood, serum).

label: Food biomarker validation assay (ONS_ZZZZZZ) superclass: ONS_YYYYYY definition: An assay that is aimed at measuring the quantity (or presence) of food intake biomarkers in a biological sample (e.g. urine, blood, serum, feces) in order to evaluate the consumption of defined foods

label: Food biomarker discovery assay (ONS_WWWWWW) superclass: ONS_YYYYYY definition: An assay that is aimed at measuring the quantity (or presence) of food related metabolites in a biological sample (e.g. urine, blood, serum, feces) in order to identify/discover new food intake biomarkers

Perfectly fine with me if you want to take the other class in FOBI! It appear to be quite centered with your scope. It could be the key class for the integration, something like:

"Food biomarker validation assay"[ONS] is_about "Food Intake Biomarker" [FOBI]
(this is a simple connection, maybe the real integration would be at the level of measured datum. We should see a little more in detail the model of OBI, but the "Food Intake Biomarker" could have the "analyte" role, and the connected axiom should be C (has_specified_input some %) and (realizes some ('analyte role' and ('inheres in' some %))) )

For the discovery, following what you wrote about the (perfectly right) distinction between metabolite and biomarker, I think that another class could be needed, a process maybe similar to something like "Food intake biomarkers annotation", in which the metabolites discovered by the assay are evaluated and eventually annotated as real biomarkers.

[pcastellanoescuder]

Hi @FrancescoVit,

Sounds great! Thanks for your detailed proposal. Some comments on that below:

label: Food biomarker analysis assay (ONS_YYYYYY) superclass: ONS_XXXXXX definition: An assay that is aimed at discovering new potential metabolites as food consumption biomarker, or to measure/validate known food consumption biomarker in biological fluids (e.g. urine, blood, serum).

I would say: "An assay that is aimed at discovering new potential metabolites as food consumption biomarkers, or to measure/validate known food consumption biomarkers in a biological sample (e.g. urine, blood, serum, feces)."

label: Food biomarker validation assay (ONS_ZZZZZZ) superclass: ONS_YYYYYY definition: An assay that is aimed at measuring the quantity (or presence) of food intake biomarkers in a biological sample (e.g. urine, blood, serum, feces) in order to evaluate the consumption of defined foods

Completely agree:)

label: Food biomarker discovery assay (ONS_WWWWWW) superclass: ONS_YYYYYY definition: An assay that is aimed at measuring the quantity (or presence) of food related metabolites in a biological sample (e.g. urine, blood, serum, feces) in order to identify/discover new food intake biomarkers

Also completely agree:)

"Food biomarker validation assay"[ONS] is_about "Food Intake Biomarker" [FOBI] (this is a simple connection, maybe the real integration would be at the level of measured datum. We should see a little more in detail the model of OBI, but the "Food Intake Biomarker" could have the "analyte" role, and the connected axiom should be _C (has_specifiedinput some %) and (realizes some ('analyte role' and ('inheres in' some %))) )

I think it's great! This is the point that was hardest for me to see. I see integration much clearer now with your proposal! Thanks:)

For the discovery, following what you wrote about the (perfectly right) distinction between metabolite and biomarker, I think that another class could be needed, a process maybe similar to something like "Food intake biomarkers annotation", in which the metabolites discovered by the assay are evaluated and eventually annotated as real biomarkers.

OK with that, however, what would be the position of this class within the ontology hierarchy?

[FrancescoVit]

In OBI I could find two "equivalent" concept in the DNA world,

http://purl.obolibrary.org/obo/OBI_0001944 http://purl.obolibrary.org/obo/OBI_0000905

don't yet know which could be the best, is not that easy as a class/concept as it is somewhere in between an assay (the process of performing the annotation) and an information content entity (the annotation itself). Moreover, we would decide between defining how a metabolite can be flagged as biomarkers (if possible) or leaving the class more general and vague

[ddooley]

One can create classes of chemical (or protein) by the role they play in some context. So a metabolite (chemical) is a chemical having a "metabolite role", and then you go on to define what that role is in a separate entity. If that helps?

Another trend is to recognize that roles are shorthand for a set of (functional) capabilities, and to work on defining what those capabilities are. Then anything that has those capabilities can then perform the role. E.g. both a hammer and a rock can perform a hammering role, but that's because they have capabilities: hand tool, don't shatter on impact, and can create a lot of momentum in a short distance.

[pcastellanoescuder]

Personally I think this class can be quite confusing. Unlike genomics with DNA sequences, this would not be a matter of "annotation" since the cataloging of a metabolite as a biomarker can be hard and extremely variable. Given that the process of assigning/"annotating" metabolites as biomarkers is still under discussion and unclear (the field of nutrimetabolomics is much younger than other omics), it might be worthwhile to wait to create this class.

[FrancescoVit]

Another trend is to recognize that roles are shorthand for a set of (functional) capabilities, and to work on defining what those capabilities are. Then anything that has those capabilities can then perform the role. E.g. both a hammer and a rock can perform a hammering role, but that's because they have capabilities: hand tool, don't shatter on impact, and can create a lot of momentum in a short distance.

Personally that's the solution I like the most, seems more "versatile" owing to the caution Paul is advising for the field.

Given that the process of assigning/"annotating" metabolites as biomarkers is still under discussion and unclear (the field of nutrimetabolomics is much younger than other omics), it might be worthwhile to wait to create this class.

I see now.

Following @ddooley suggestion, we could create a double of the biomarker class, as a role.