grassp.ds.gene_sets_compartments#
- gene_sets_compartments(species='hsap', *, min_genes=None)[source]#
COMPARTMENTS curated subcellular localisation, 874 terms for human.
COMPARTMENTS (Binder et al. 2014, https://compartments.jensenlab.org) integrates localisation evidence from several channels; this is the knowledge channel, the curated one – manually assigned GO cellular-component terms carrying a 1-5 star confidence, already propagated over the ontology. Kept at confidence >= 3 and >= 5 genes, with generic and non-locational terms removed (whole-cell aggregators, topology qualifiers, “region of” groupings, the extracellular-exosome dump) and near-duplicate gene sets collapsed at Jaccard >= 0.8, keeping the canonical organelle name.
Sits between
gene_sets_curated()andgene_sets_go_cc()in granularity. Term labels are upper-case, as in both published grassp analyses; the GO id is in the GMT’s description column.Vendored build: 2026-09-24. COMPARTMENTS publishes no versioned archive, only a rolling download, so this build date is the only version this vocabulary has – quote it when reporting results.
grassp.datasets.gene_sets.provenance()returns the full record, andmarker_curation/fetch_compartments.pyrebuilds it.- Parameters:
- species
Literal['hsap','mmus','scer'] (default:'hsap') Only
"hsap"is bundled; the knowledge channel for other species is not curated to a comparable standard.- min_genes
int|None(default:None) Drop terms with fewer than this many genes. The build already applies a floor of 5; raising it further is worthwhile for enrichment testing.
- species
- Return type:
- Returns:
dict[str, list[str]] Mapping of upper-case compartment name to gene symbols.
See also
gene_sets_curatedthe coarse, experiment-resolvable compartments to start from.
gene_sets_go_ccthe finer GO cellular-component vocabulary.
Examples
>>> import grassp as gr >>> sets = gr.ds.gene_sets_compartments() >>> len(sets) 874 >>> len(gr.ds.gene_sets_compartments(min_genes=20)) 404
Notes
13 term names contain
/, which h5ad rejects as an obsm column name; annotators fall back touns[f"{key}_probabilities_categories"for those.