Cell and Repertoire definition

[sharikrish]

In our previous discussion on cell object specification (issue : #320 ) and repertoire definition described here represents "actual" Repertoire and "virtual" Repertoire

Actual repertoires being the ones that contain each cell object only once and virtual repertoire can contain them multiple times (either copied or linked).

From the previous call on MiniStd, there is a consensus that overlapping populations/repertoires need to be addressed and that two tiers of Repertoire objects could be a possible solution.

Any thoughts how we represent them in the schema?

[scharch]

Can we just add a boolean virtual field? Or maybe better would be list of linked_repertoires which would by its presence define a Repertoire as virtual

[javh]

I'm having trouble understanding the distinction. A Repertoire is an arbitrary collection of records, with the scope defined by whoever decided to label the collection as a Repertoire.

What's the difference between an "actual" and "virtual" arbitrary collection? And what is the problem we want to solve? If it's data compression, by collapsing multiple Cell records into a single consensus Cell record, then it may just be a matter of allowing repertoire_id and/or sample_id to be an array.

[lgcowell]

I haven't been carefully following the thread, but it seems a virtual repertoire is what Jason is defining (arbitrary collection of records) while an actual repertoire is a biological entity and so is not composed of records and membership in the repertoire is not arbitrary but is based on a combination of cell type and anatomical location.

[sharikrish]

Can we just add a boolean virtual field? Or maybe better would be list of linked_repertoires which would by its presence define a Repertoire as virtual

Adding the 'virtual' flag makes sense and should be possible to implement. How to handle the meta-data associated with the linked repertoires?

[scharch]

How to handle the meta-data associated with the linked repertoires?

The point of having a linked_repertoires field would be to get the metadata from the originals without needing to copy it. I haven't completely thought it through yet, though.

[bcorrie]

Not sure whether this is still a v2.0 task, @scharch ? This kind of feels like a RepertoireGroup type of issue - so maybe have this land wherever RepertoireGroup lands? See #578

[scharch]

Yes, I think this would ideally be solved by RepertoireGroup...

[bussec]

Re-reading this thread and parts of #320, I also think that RepertoireGroup is likely the second level we are looking for. We should save the previously mentioned use case, but otherwise we can IMO close this.

[bcorrie]

I have create a Repertoire Group pull request (#748) and associated this issue with the pull request. So when we are comfortable the Repertoire Group addresses this we can close it.