Infer trophic behavior from consumed resource

[nleguillarme]

Hi, I am considering using classes from ECOCORE to infer trophic behavior from trophic interaction observations.

To give you an example, suppose I have an individual which is an instance of 'organism' (http://purl.obolibrary.org/obo/OBI_0100026) and I know that this individual is involved in a 'eats' relationship with a second individual which is an instance of 'Bacteria' (http://purl.obolibrary.org/obo/NCBITaxon_2).

Currently, I cannot infer that the former is an instance of the 'bacterivore' class (http://purl.obolibrary.org/obo/ECOCORE_00000098) because the equivalence axiom states that 'bacterivore' is equivalent to 'heterotroph and (eats some Bacteria)'. However, we generally don't have the information that an organism is an heterotroph in our datasets (at least explicitly), just some observations of interactions between two taxa.

What would be the best way to automatically infer that an organism is a bacterivore, just knowing it is an organism and it is involved in a 'eats' relationship with a bacteria ?
One simple way would be to set 'bacterivore' as equivalent to '(eats some Bacteria)', without explicit reference to the organism being a heterotroph.

[diatomsRcool]

Interesting. I think that is a reasonable change that would help in many ways. I can make the change.

[diatomsRcool]

@cmungall I could change the axiom to remove the "genus", but then wouldn't that imply that bacterivore was no longer a subclass of heterotroph?

[diatomsRcool]

@nleguillarme see the 11 Dec release. Did this address the issue?

[nleguillarme]

Hi @diatomsRcool, sorry, I missed your last message.

I see that in the last release of ECOCORE, Bacterivore is equivalent to

('capable of' some bactivory)
and (eats some
(Bacteria
and Archaea)

IMO, the real equivalence here is rather Bacterivore = ('capable of' some bactivory) = (eats some (Bacteria or Archaea))
Still, there is these multiple equivalence problem. @cmungall proposed something interesting here but I don't know how to do that in practice.

Do you think it's worth creating a dedicated channel on the obo-community Slack to discuss this point ?

[nleguillarme]

This made me think of another related question concerning the representation of taxonomic entities: how to represent the fact that an organism belong to a specific taxonomic class, e.g. that an organism is a Bacteria.

For instance using GloBi's representation of interactions, this would be represented as follows:

_:bnode1 rdf:type CARO_0010004 (cellular organism)
_:bnode1 RO_0002350 (member of) http://purl.obolibrary.org/obo/NCBITaxon_2

With this representation, it's not possible to infer that an organism involved in a "eats" relation with _:bnode1 is a Bacterivore, as this would require that _:bnode1 rdf:type http://purl.obolibrary.org/obo/NCBITaxon_2.

So for inference to work with GloBi's representation of interactions, we would need Bacterivore to be defined as follows :

Bacterivore = eats some ('cellular organism' and 'member of' (Bacteria or Archea))

Or alternatively, we could state that Bacteria or Archea (from the NCBITaxon ontology) are subclasses of CARO's 'cellular organism' (or better, subclasses of CARO's 'organism or virus or viroid'), and that Bacterivore = eats some ('member of' (Bacteria or Archea))

@ramonawalls first raised the question of taxonomic concepts modelling in this thread. It woud be great if we could agree on a common representation for taxonomic entities.