Open cmungall opened 5 years ago
I haven't looked at ECTO yet, but administered exposures are fundamentally different from 'naturally occuring ones', where quite often the process described like 'living in an endemic area' gives a probabilistic view that an exposure is likely to have occurred. Similar with diet: A lab animal will not have a choice and eat what it is given. For an animal in the wild, you can list what is available for diet etc.
For OBI we care a lot about administered exposures, where there is route / dose / schedule / formulation / adjuvantation etc. Yes, these are all processes (planned processes in this case).
So where should the unplanned processes live?
We have been looking for a home for unplanned exposures, a few of which are currently captured in OBI (for lack of a better space). If ECTO can capture the generic exposure in a format 'organism being exposed' 'route of exposure' 'material it is being exposed to', and maybe also deals with different natural exposures (like living in an area where exposure is likely, but warning, those are hard), and leaves the 'planned exposures' to OBI (administered how, formulated how, dosing, schedule), that would be a great separation of work.
Another way to look at it - maybe compatible - is to take
"administering substance in vivo" http://www.ontobee.org/ontology/OBI?iri=http://purl.obolibrary.org/obo/OBI_0600007
The text def is "A process by which [...] resulting in exposure of the organism to that substance"
I like this, this suggests a model in which there are two processes - the administration E1 and an immediately causally downstream process of the exposure E2, where E1 causally-upstream-of E2
This also works for upstream events that are not experimental. E.g. E1 may be an accident such a a fire or chemical spillage or improper pesticide administration, which directly leads to the actual exposure event E2.
I don't think this is so different from what you are suggesting, but rather than thinking of generic and specific instances (experiment, accident, neutral etc) as I was originally suggesting at the top of the ticket, it's more of a compositional model. this has some disadvantages - e.g. multiplying entities.
We'd have to work out the details, but I think that things could be coordinated such that OBI and ECTO could hook up via inference. E.g. An administration of X leads to an X exposure.
Re: your example living in an area where an exposure is likely. My naive approach would be to first have a standard way of representing something like "living an a region with properties X" (endemic malaria, crowded living conditions, etc); then to make a probabilistic link between these occurrents and an exposure (to bite from plasmodium carrying mosquito, stress, etc) we're in the same boat as any other probabilistic association, e.g. between disease class and phenotype class, with a variety of approaches, but generally avoiding all-some ontology axioms.
Where the named classes / materialized expressions of "living in Malaria" live, I'm not sure. There is a risk of overmodularizing but it feels like this is also worth splitting off (but closely coordinating DPs)....?
[got to go more later]
As outlined in Randi's paper, we just want to define <20 terms, which when combined with other ontologies (e.g. NCBI taxonomy, DO) cover all the immune exposures we care about. We could include in their definition the 'has exposure stimulus' relationship in ECTO, which pretty much means following what you also suggested above and decomposing the exposure part from the rest.
Looking at ECTO, there were a lot of terms that are elsewhere (e.g. injection, exposure to viruses, etc), and some of the language is weird, such as 'the word 'stimulus' and receptor' which seem to be used throughout, lead to awkward definitions (e.g. for gamma radiation), but I guess you have a different definition what receptor + stimulus are supposed to be? I guess you are trying to use language of what entity is being exposed (receptor) and what the entity is being exposed to (stimulus), but those words have very clear meanings in other context... No better suggestion immediately comes to mind though.
from Bjoern:
This is great, I think many of these are complementary to what we may have.
Our plan in ECTO was to be neutral about intentionality. Biologically an exposure to X is an exposure to X. ECTO is already somewhat ridiculously compositional, and to have
{admininstered,natural} exposure to {C} via {Route}
could get too unwieldy.Was this what you had in mind?