OBOFoundry
Proposal for CORE terms relating to proteins, macromolecules, aggregates, complexes
Please see attached OWL file. Reasoning is needed. There's also a list of considerations used when formulating this and of things to think about, also in an attached document (it's a wordy document).
I did not attempt to fit all this into the current CORE because I don't want to mess anything up or make non-discussed changes. Instead, I took a slice of various ontologies--using the original identifiers--and put them together into a (hopefully) cohesive unit. I did my best to label new or changed terms.
Thanks for this detailed work. Haven't fully digested some brief comments for now:
- I avoided use of the term ‘polymer’. I initially thought that the definition of ‘polymer’ would be uncontroversial—that it would be simply a covalently linked chain of monomers (often termed ‘residues’) of any length (though the obvious minimum would be two). However, here I see two problems occurring: (1) ‘polymer’ is often conflated with ‘macromolecule’; and (2) sometimes an alternative minimum is given. For example, by CHEBI definition (and IUPAC), an RNA chain less than 13 residues long would not be a polymer. This goes against all instinct. Accordingly, wherever you might normally see ‘polymer’ I used the more neutral ‘chain’.
I think your biomolecular chain class makes sense.
I think it's a bit confusing having chain constituent and chain component.
- Much of the work here attempts to make sense of ‘protein-containing complex’
Maybe let's take this to its own ticket and loop in GO people?
- “Group” (from CHEBI): explicit in the definition is that these are proper parts. At issue is whether or not the CORE definition of ‘molecular entity’ would allow group as a subclass. I think it does, but others might disagree.
Seems we want to include chemical group in core
- CHEBI’s ‘protein’: This is basically a union of PRO’s protein and GO’s protein-containing complex
Really? That is super-weird. Another case of has-part and is-a confusion in CHEBI. We need to mark this logical relationship in core.
I think it's a bit confusing having chain constituent and chain component.
I presume you refer to chain constituent and complex component? I actually originally had both labeled as component, but I found that to be mildly confusing too. It might not be necessary to have separate relations; really all that's needed is a non-transitive parthood. I divided into separate relations because the objects really are of two types.
Maybe let's take this to its own ticket and loop in GO people?
No objections. I expect this ticket to spawn many other tickets, but at this stage it didn't make sense to present the issues piecemeal without first presenting the whole package.
- CHEBI’s ‘protein’: This is basically a union of PRO’s protein and GO’s protein-containing complex
Really? That is super-weird. Another case of has-part and is-a confusion in CHEBI. We need to mark this logical relationship in core.
In my original treatment, I did have this marked in CORE. I can easily put it back, but opted to leave it out because of some other conversation about keeping defined classes to a minimum in CORE. I basically kept only those classes and relations needed to make this part of the CORE fully self-contained. I also added a comment to PRO's protein term making the relationship between the two clear.
If having the logical relationship is desired, here is how I would define CHEBI's version of protein (using OBO format for convenience)
intersection_of: CORE:0000013 ! molecular entity intersection_of: has_component PR:000000001 {minCardinality=1} ! protein (PRO)
Upon reasoning, both complexes and proteins (in the PRO sense) will class under this term.
I do understand the need to get all the current GO, PRO, ChEBI terms into alignment, but I am afraid that working on this solely with that in mind leads to weird design choices. Can we at least try to come up with definitions that normal scientists would have at first, focus on getting them right, and then trying to see where the existing terms fall? This is the time where we can do that. If we get a large number of people involved that are vested in the existing terms, we end up with power plays and trench digging. If we can find a set of definitions that appeal to 'conventional scientific wisdom', the chances are higher that we get everyone to align with them (I hope) - and that might mean they define their terms 'relative' to the core terms.
- Bjoern
On Mon, Aug 19, 2019 at 9:09 AM Darren A. Natale [email protected] wrote:
I think it's a bit confusing having chain constituent and chain component.
I presume you refer to chain constituent and complex component? I actually originally had both labeled as component, but I found that to be mildly confusing too. It might not be necessary to have separate relations; really all that's needed is a non-transitive parthood. I divided into separate relations because the objects really are of two types.
Maybe let's take this to its own ticket and loop in GO people?
No objections. I expect this ticket to spawn many other tickets, but at this stage it didn't make sense to present the issues piecemeal without first presenting the whole package.
- CHEBI’s ‘protein’: This is basically a union of PRO’s protein and GO’s protein-containing complex
Really? That is super-weird. Another case of has-part and is-a confusion in CHEBI. We need to mark this logical relationship in core.
In my original treatment, I did have this marked in CORE. I can easily put it back, but opted to leave it out because of some other conversation about keeping defined classes to a minimum in CORE. I basically kept only those classes and relations needed to make this part of the CORE fully self-contained. I also added a comment to PRO's protein term making the relationship between the two clear.
If having the logical relationship is desired, here is how I would define CHEBI's version of protein (using OBO format for convenience)
intersection_of: CORE:0000013 ! molecular entity intersection_of: has_component PR:000000001 {minCardinality=1} ! protein (PRO)
Upon reasoning, both complexes and proteins (in the PRO sense) will class under this term.
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/OBOFoundry/Experimental-OBO-Core/issues/46?email_source=notifications&email_token=ADJX2IQSQZLHW2ZN4FAS4KDQFLARXA5CNFSM4IMLWT6KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4TO7YQ#issuecomment-522645474, or mute the thread https://github.com/notifications/unsubscribe-auth/ADJX2ITFB2I2YAUVXUIKY73QFLARXANCNFSM4IMLWT6A .
-- Bjoern Peters Professor La Jolla Institute for Allergy and Immunology 9420 Athena Circle La Jolla, CA 92037, USA Tel: 858/752-6914 Fax: 858/752-6987 http://www.liai.org/pages/faculty-peters
I agree that there might be some weird design choices. But, I actually started with term labels and definitions that normal scientists would have. Those labels and definitions failed because there is clearly a lack of convention in 'conventional scientific wisdom' (as you yourself have seen when dealing with 'macromolecular entity').