Wikidata:Property proposal/role in the ChEBI ontology

has role (in ChEBI Ontology)

Not done

Description	the role of a certain compound in the context of the ChEBI (Chemical Entities of Biological Interest) ontology, e.g. anti-neoplastic agent
Data type	Item
Domain	chemistry
Allowed values	Wikidata items which are part of the ChEBI ontology and therefore carry an ChEBI ontology ID ChEBI ID (P683)
Example	vemurafenib (Q423111) → antineoplastic (Q23987513)
Source	https://www.ebi.ac.uk/chebi/init.do
Robot and gadget jobs	ProteinBoxBot will add this relation to drug items and other chemical entities.

Motivation

ChEBI is a central ontology for biology and chemistry. In order to represent the ChEBI ontology in Wikidata, a central requirement is the 'has role' property. The P794 (P794) is not suitable as it can only be used as a qualifier and the role property for actors is also restricted. Furthermore, biological process (P682), the one closes to the proposed property is not precise enough and also restricted to the Gene Ontology. subclass of (P279) does definitely not apply here (because it maps to 'is a') and instance of (P31) does not seem to fit as well for this purpose.

In principle, a 'has role' property could be used as a generalized property (for natural sciences), although redundancies with Gene Ontology need to be avoided. In addition, could we just use this property proposal to decide on all required properties to represent ChEBI in Wikidata at once [1]? Some of those already exist anyway. This would enormously speed up the process of getting complete ChEBI into Wikidata. thx! Sebotic (talk) 06:35, 3 May 2016 (UTC)[reply]

WikiProject Ontology has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. WikiProject Molecular_biology has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead.

Notified participants of WikiProject Chemistry

Notified participants of WikiProject Medicine

Discussion

Oppose @Sebotic: I don't think that the goal of WD is to integrate ChEBI. Doing that is the begin of a nightmare to be up-to-date with ChEBI modifications. WD is more a general ontology (top domain ontology) that a domain ontology. Snipre (talk) 13:18, 4 May 2016 (UTC)[reply]
Just a detail: I am not opposed to a similar property but this property should be independent from ChEBI ontology. But again there is no sense to copy ChEBI: better map the WD items with ChEBI elements and people can then easily extracted ChEBI relations from WD items. Snipre (talk) 14:01, 4 May 2016 (UTC)[reply]
- @Snipre: The reason that I want to import ChEBI is simply to give chemical compounds in Wikidata a structure. Currently, there is essentially no classification and structure in the chemical compound data in Wikidata. Moreover, it is not realistic to expect that such a structure will be created by the community nor can it be expected that a community-created structure will be of higher quality than ChEBI, which has been developed and maintained for many years now (since 2004). Nevertheless, I think for the usefulness of the chemical compound data in Wikidata, such a structure is a definite requirement. I agree that Wikidata itself is a high-level ontology, but it is not a help here but also not a hindrance. Importing and keeping Wikidata in sync with ChEBI is not a great challenge. I have done it before by importing the complete Gene Ontology structure. I have the code in place which keeps Gene Ontology in sync with Wikidata [2]. For ChEBI, this would work exactly the same way. ChEBI and Gene Ontology are approximately the same size (48,000 and 44,000 terms, respectively).

I would start with the backbone, which basically consists of 'is a' (subclass of (P279) in Wikidata) and 'has role', the rest of ChEBI can be imported if the properties are ready.

One very important point to make is that the ChEBI structure in Wikidata should serve as a backbone for the community to build on. So the import would essentially allow the community to expand ChEBI in the Wikidata space while still keeping compatibility with ChEBI (discrimination of contributions based on references and/or qualifiers). Sebotic (talk) 18:59, 4 May 2016 (UTC)[reply]

Sebotic You need to get approval by a several contibutors working in the field of chemistry/medecin/microbiology before doing that kind of importation: if you are the only doing that kind of importation and nobody knows what is the backbone of the classification, all your work will be destroy after some monthes. People can only respect the work of other contributors when they are aware of what is in front of them. Again I think that in term of maintenance this kind of initiatives are useless without a strong support of a large part of the community. I undo the work of several bots during the last monthes because they were not working properly: importing once is only the easy part of the job, maitain the structure over the time is the difficult one. Snipre (talk) 19:20, 5 May 2016 (UTC)[reply]

Snipre The goal of our effort is certainly to also do long term maintenance/sync of ChEBI in Wikidata, otherwise the value for Wikidata would be zero. As stated above, there is no technical challenge to keep ChEBI in sync with Wikidata. Shall I open a new thread on Project chemistry, so we can start a discussion? Sebotic (talk) 20:09, 16 May 2016 (UTC)[reply]

Oppose I agree with Snipre - also, unless I'm misunderstanding, the meaning of this seems clearly along the lines of what instance of (P31) is intended for and is used for in other areas. Can you argue more clearly why it "does not seem to fit as well" here? ArthurPSmith (talk) 13:55, 4 May 2016 (UTC)[reply]
ArthurPSmith Here instance of is a bad way to describe use of some items: we can have dozen of use/role for some items and this becomes difficult after to extract the correct values among 10 or 15 possibilities. A general property use/role is needed but not only for chemicals. Snipre (talk) 14:04, 4 May 2016 (UTC)[reply]
- @Snipre:@ArthurPSmith: I agree with Snipre that this property should be created as a generalized 'has role' property. I can adapt and maybe move the proposal. The instance of (P31) is mapped to W3C rdf:type and therefore represents an instance of a class, not its role in a certain context. Sebotic (talk) 18:59, 4 May 2016 (UTC)[reply]

'has role' maps to http://purl.obolibrary.org/obo/RO_0000087 in the Relations Ontology and is being used by several other ontologies as well (see list at the bottom of the page the purl resolves to) Sebotic (talk) 20:17, 4 May 2016 (UTC)[reply]

Sebotic The definition should be more clearly defined in order to avoid any use of this property as another "instance of". The quality of the item using the property and the one of the property value should be given. Ex. Is D-penicillamine (Q421239) has role medication (Q12140) correct ? Or should we have D-penicillamine (Q421239) has role chelating ? Snipre (talk) 12:22, 11 May 2016 (UTC)[reply]

Snipre Sorry for the delayed reply! The properties instance of (P31), subclass of (P279) and a potential 'has role' property are all be mapped OBO/OWL terms in order to connect them to knowledge bases or knowledge organization systems which are already out there. I think that is very important. Regarding your example: instance of (P31) is mapped to [3] so it might be better to replace instance of (P31) with subclass of (P279) (is a) for chemical compounds in Wikidata, because this better describes the relationship (many other ontologies make heavy use of 'is a'). For me, there is a clear semantic distinction of 'is a' and 'has role' and I think this should also be modeled in Wikidata, as it makes the data more useful for others. For chemical compounds, 'is a' can be used to organize stuff in Wikidata, but also to fully represent ChEBI. Differentiation where a certain relation comes from, can just be organized via the references. So if I want to import/update ChEBI, only touch statements with ChEBI references or equal values, leave the other statements intact. So in my opinion, several structures could exist together. Sebotic (talk) 20:09, 16 May 2016 (UTC)[reply]

@Sebotic: Not done, no consensus, but Wikidata:Property proposal/has role seems to have good chances. --Srittau (talk) 16:27, 19 May 2016 (UTC)[reply]