Update cap_anndata_schema.md

[mfutey]

For MVP-5404

Made the following two updates:

• added a correct example value for [cellannotation_setname]_metadata
• changed [cellannotation_setname]--metadata to [cellannotation_setname]_metadata

[rm1113]

we don't have the [cellannotation_setname]_metadata field in the schema.
uns consists of constant name cellannotation_metadata which will maps to dict whith [cellannotation_setname] keys.
So the right example is:

uns = {
    "title": blabla,
    "cellannotation_metadata": {
        "cell_type": {"description": blabla,  "annotation_method": blabla...},
        "author_cell_type":{"description": blabla,  "annotation_method": blabla...},
    }
}

[mfutey]

Excellent point, I updated [cellannotation_setname]_metadata to cellannotation_metadata but used Ugur's suggestion to have the example value be the value only and not include the key

[ubyndr]

this should change to cellannotation_metadata right?

Suggested change

	#### [cellannotation_setname]_metadata
	#### [cellannotation_metadata]_metadata

[mfutey]

@ubyndr it should be [cellannotation_metadata]

[evanbiederstedt]

@ubyndr @mfutey @rm1113

Might be confusion above

[cellannotation_setname] is the name of the "labelset" or cell annotation set. e.g. cell_type

cf #118 (comment)

I am guessing the misalignment is that Roman + Evan simplified this earlier. That example is correct.

• We tag all of this uns stuff as cellannotation_metadata.
• Keys within that dictionary are the set names themselves i.e. [cellannotation_setname]

Are we aligned now?

[ubyndr]

I've suggested two possible changes for your review.

[evanbiederstedt]

LGTM

[dosumis]

@evanbiederstedt - Did this just change the schema so that it now differs from the one you are actually using on CAP?

I want something for now that will conform to what CAP currently does so that we can code to it it load data. If you are changing the key name, then (a) please delay until implemented in CAP (b) please, please, please change this to labelsets so that this conforms to the nomenclature in CAS. Absence of alignment is going to cause endless confusion.

(Also worth noting that the problems here are a good illustration of why programatically validateable schemas are good)