Should a Reference{,Set} support multiple accessions? about ga4gh-schemas HOT 3 CLOSED

It is common for the same sequence to have different accessions in different sub-databases from NCBI, but in one database, such as RefSeq, there is no redundancy, I think. That said, I am supportive of having multiple sourceAccessions associated with one GAReference.

from ga4gh-schemas.

OK, so the remaining issue is around the relationship of GAReferenceSet::sourceAccession and GAReference::sourceAccession. I don't think I understand the current semantics, as I thought that an assembly (ReferenceSet) had one accession number, while the individual contigs in the assembly (References) had entirely different accession numbers.

To give a concrete example:
GRCh38 as the assembly has accession GCF_000001405.26 in RefSeq. I would expect this to be the value of GAReferenceSet::sourceAccession.
chr1 of GRCh38 has an accession of NC_000001.11 in RefSeq. I would expect this to be the value of GAReference::sourceAccession.

Am I misunderstanding something? If not, I don't understand the semantics on GAReference::sourceAccession:

The accession from which the sequence was obtained, if not given in the
containing GAReferenceSet.
In INSDC (GenBank/ENA/DDBJ) with a version number. (e.g. GCF_000001405.26)

from ga4gh-schemas.

@calbach, you are right, less is more by having one assembly identified in GAReferenceSet::sourceAccession, but the key here is that with GAReferenceSet you specify multiple GAReference via array<string> referenceIds = [];. If you have multiple GAReference, and you want sourceAccession granularity for specific GAReference that customize your GAReferenceSet, you can overide it with GAReference::sourceAccession as the comment mentions: "Can be overriden by the sourceAccession field in a specific GAReference". This way you have flexibility in how you build your GAReferenceSet.

from ga4gh-schemas.