Re: New property added to BioChemEntity

On Wed, Oct 25, 2017 at 2:32 PM, Leyla Garcia <ljgarcia@ebi.ac.uk> wrote:

> Hello,
>
> There was a discussion last week about this new property, not many
> replies, no one objecting so I added it (and by doing that was hopping to
> get some more attention and more feedback). I guess at some point the
> reviewing committee will accept or reject, maybe based on what the
> community has said, still not sure about the process.
>

Ah okay, this was [1]?  I didn't appreciate the significance originally.

>
> Why on top of AdditionalType? There are multiple ontologies describing
> what a protein/chemical/gene/etc. is. Usually groups will prefer to point
> to their own ontology.  Yes, we could have a list of all ontology terms
> allowed for, let's say, protein. That list will have to grow whenever a new
> ontology term gets in use. The one using it will have to inform some how,
> the profile responsible will have to double check and then add it, the
> validators and tools will need to know that that additionalType also refers
> to proteins. So having a simple label seems easier, I do not expect that
> many new profiles as ontology terms describing the same thing in slightly
> different ways. In a way, that label is the name of the profile if it were
> a proper type.
>

Having thought about this, perhaps the critical question is whether we
really want to allow genericity.
You're saying that data sources may want to embed properties over and above
what's in a bioschemas profile. I was thinking that if data sources can
specify any BioChemEntity it would improve findabilty (I hadn't given much
consideration as to whether they would also embed arbitrary
AdditionalPropertys).

However, whilst the scruffy part of me really likes the idea of data
sources embedding arbitrary terms and values within the BioChemEntity
structure, applications doing the work to make sense of it, and new
profiles emerging once there is concensus, perhaps it's too much to ask.
Most data sources may be fairly conservative and not want to use
types/AdditionalProperty that haven't been blessed by bioschemas and so not
understood by some (most?) applications.  And on the application side,
afaik comprehensive (?) mapping solutions are only just emerging (e.g. the
work being done by Simon Jupp's team and the Pistoia Alliance).

Eliminating genericity would reduce complexity and make preferredLabel
redundant, I think?  Bioschemas could mandate 1-to-1 URL <-> profile
mappings for AdditionalType (whether they're from SIO or neutral bioschemas
URLs) and only bless BioChemEntitys with profiles. It would still be more
flexible than first class schema.org types as profiles could hopefully be
more quickly developed by Bioschemas. A big con would be that data sources
couldn't rock up and start describing arbitrary types. But maybe here a URL
for a new profile could be fast-track blessed by bioschemas.org, before a
more involved process to define what AdditionalPropertys would actually be
in it.

Thoughts?

(there's also the more schema.org-like way of doing AdditionalProperty that
Alasdair is going to talk about, but I think that's orthogonal to the idea
of whether we allow profile-less BioChemEntity).

[1] https://lists.w3.org/Archives/Public/public-bioschemas/2017Oct/0011.html

<https://lists.w3.org/Archives/Public/public-bioschemas/2017Oct/0011.html>

>
> Mandatory or not, one or many, it all belongs to Bioschemas, so I would
> say that any BioChemEntity that does not have a profile, still should
> follow the minimums defined by Bioschemas. In that sense, if no specific
> profile available, "BioChemEntity" would be the profile to follow.
>
> Any thoughts?
>
> Regards,
>
>
> On 25/10/2017 11:58, Justin Clark-Casey wrote:
>
> Hello Leyla,
>
> What about BioChemEntitys that do not yet have or may never have profiles?
> This is the generic case.
>
> Why is this required on top of AdditionalType?
>
> Regards,
>
> Justin
>
> On 25 Oct 2017 11:12, "Leyla Garcia" <ljgarcia@ebi.ac.uk> wrote:
>
>> Dear all,
>>
>> A new property has been added to BioChemEntity. This property MUST
>> contain the name of the profile so it will be easier for validators and
>> tools to distinguished one profile from another.
>>
>> property: preferredLabel
>> expectedType: Text
>> Description: Indicates the preferred label to refer to a specific
>> (sub)type of BioChemEntity Bioschemas description: Profile name
>> Marginality: Minimum Cardinality: ONE Bioschemas description: Bioschemas
>> profiles Regards,
>>
>
>