ethylparabens & classification by role

[cmungall]

inferred superclasses for exposure to ethylparaben:

are these correct?

they all come from CHEBI:

Our role pattern is 'has exposure stimulus' some ('chemical entity' and 'has role' some %s), so if CHEBI says that C has role R, then an exposure to C will be classified as an exposure "to" R. Is this right in all cases? Let's look.

• phytoestrogen - explanation comes from CHEBI, seems to be correct (Note MRE did not have this link).
• antimicrobial food preservative - seems dubious. Even though ethylparaben is used for this purpose sometimes, if the exposure is in the context of cosmetics use seems odd to classify this way.

I think we need to distinguish between

1. classification by potentially realized roles, or rigid roles: 'has exposure stimulus' some ('chemical entity' and 'has role' some %s)
2. classification by roles that actually realized: 'has exposure stimulus' some ('chemical entity' and 'has role' some (%s and realized-by some process))

E.g.

• any exposure to ethylparaben is an exposure to a potential antimicrobial food preservative
• an exposure to ethylparaben used as a cosmetic is an exposure to a potential antimicrobial food preservative, but not an exposure to a chemical with a antimicrobial food preservative role that is actually realized

Should we always name these 2 classes, to avoid confusion, and for maximum precision? Also a 3rd class for when we know the role is not realized?

If we follow this pattern for 'hormone', we would have exposures to chemicals with potential hormone roles, and exposures to chemicals with realized hormone roles. The former is what most people would query by. For example, we would expect BPA to come back in a query for xenestrogen, whether or not it actually realized this role in any given exposure. Of course it's hard to determine whether in any given exposure the role is realized, so the existence of the subclass may seem odd.

It may be simplest in the short term to continue to use the uncommitted version (1), but for some cases, specifically those roles involving some kind of human intention, we make more specific subclasses.

[cmungall]

inference for MRE: https://github.com/cmungall/environmental-conditions/blob/eb59288ceb2421521fbbc22c85c8c1f8c0365d48/src/mapping/new-inf-mre.tsv#L48

Same kind of idea, but with drug: https://github.com/cmungall/environmental-conditions/blob/eb59288ceb2421521fbbc22c85c8c1f8c0365d48/src/mapping/new-inf-mre.tsv#L1