Direct link between atlases and tools/services that use them?

[UlrikeS91]

This is just an idea, but it might be worthwhile thinking about it. Atlases are obviously research outputs themselves, but (one of) their main purpose is to be used somehow, e.g. as a visual reference or within tools/services.

So, wouldn't it make sense to capture this somehow?

Example: QuickNII It's a tool used to register data to a reference atlas. Right now it supports the Allen mouse brain atlas and the Waxholm rat brain atlas (and many more to come). QuickNII is/will be registered as a Software(Version) and both atlases will be/are registered as BrainAtlas(Versions), but there is no connection between the two...

[lzehl]

@bweyers and @jagru20 any idea how this linkage could be established in a Software? (it is a rather specific case, but maybe you know about similar cases for e.g., software that only supports certain models?)

[bweyers]

Isn't it the case that atlases are (from a software point of view) just a specific type of data to be read and written on? Thus, a linkage similar to datasets could be thought...

[lzehl]

@bweyers generally speaking atlases can match a single dataset, but could also be composed of multiple datasets. Datasets and software are currently linked through contentTypes. I think contentTypes are not sufficient to state the support of a particular atlas or model within a software, because their original contentType is often irrelevant in these cases. They are neither input nor output format, but integrated (typically hard-coded) within the software directly, no? For example, QuickNII supports two atlases and allows datasets with certain contentTypes to be registered to these software-integrated atlases. The contentTypes of the atlases do not play a role in this context, and other atlases with the same contentTypes are not supported by the software. I think the atlases are more to be seen as implementation aspect of the software itself. What do you think?

[lzehl]

@bweyers and @jagru20 did you have the time to think about this a bit?

[lzehl]

@bweyers and @jagru20 Could you comment on this issue? I could imagine the following:

Adding an additional optional property to SoftwareVersion schema "supported construct" (feel free suggest a different name) which can get as input (type array (0-N)) a BrainAtlasVersion, a ModelVersion, a MetaDataModelVersion, a CommonCoordinateSpace (for now).

[jagru20]

Hey Lyuba, please apologize the delay from our side. Introducing a property like the one you proposed sounds like a good idea for this purpose. If it is an Array I would rather suggest naming it "supportedConstructs" but that is nitpicking.

[UlrikeS91]

I moved this to core, since the implementation proposed will require a change to a core schema (and not a SANDS schema).