[Joe-Wdz]

Split gene counts into human gene counts and model organism gene counts

Test:
Human genes (Alzheimer's disease):
https://stage.alliancegenome.org/api/disease/DOID:10652/human_genes_counts

Model organism genes (Alzheimer's disease):
https://stage.alliancegenome.org/api/disease/DOID:10652/model_genes_counts

Total genes (Alzheimer's disease):
https://stage.alliancegenome.org/api/disease/DOID:10652/genes_counts.

[claude]

Reviewed the changes: the two new endpoints (/{id}/human_genes_counts, /{id}/model_genes_counts) and countDistinctPositiveGenesBySpeciesGroup. I found no blocking issues.

• The query is the same as the existing countDistinctPositiveSubjects (category filter, parentSlimIDs.keyword, excluding negated annotations, cardinality with an exact precision threshold). The only addition is a human-taxon filter or mustNot, so it runs as one size-0 aggregation query per call and doesn't add N+1 or extra load.
• subject.taxon.curie.keyword isn't listed explicitly in Mapping.java, but the main search index relies on dynamic mapping (parentSlimIDs.keyword isn't listed either). The file generators already read *.taxon.curie from these docs, so the field should exist.
• The changes only add endpoints and don't change any existing API.
• Minor point, not blocking: the model count uses mustNot(human), so a gene annotation with no subject.taxon.curie counts as a model-organism gene. As a result, human + model always equals the existing genes_counts, which is probably what you want.

LGTM.