Amalgam

Amalgam is a biomedical metaontology: a cross-resource integration layer of entity identifiers, names, and typed cross-references organized by Biolink class. It integrates source ontologies rather than replacing their semantics; typed associations are not equivalence assertions.

Biomedical entity names for Biolink classes, built from BABEL snapshot local:/home/sgoetz/babel-snapshots/2026jul22. Cliques that share a preferred name within a class and taxon scope may merge unless identifier namespaces conflict. Every synonym is assigned to one CURIE within its class and taxon scope. Strings are kept as written; normalization is left to consumers.

Additional sources:

• GTDB: GTDB: taxa (organism_taxon); species carry their NCBI species taxid crossref as a GTDB: synonym of the NCBITaxon: row (at most one GTDB species per taxid).
• ClinVar variant_summary: one CLINVAR: row per variation and one DBSNP:rs... row per referenced rsID (sequence_variant), never merged with each other.
• OBO ontologies not in BABEL (config/obo_sources.yaml): active own-prefix terms with names, EXACT synonyms and alt_ids.

Subsets

cliques (default)

One row per clique, one split per Biolink class.

column	type	meaning
CURIE	string	clique leader identifier
PREFERRED_NAME	string	preferred name, as written
CLASS	string	Biolink class, e.g. biolink:Gene (split gene)
SYNONYMS	list	sorted names; may be empty when every name lost arbitration
TAXA	list	sorted NCBITaxon: CURIEs; empty when taxonless

split	class	rows
activity	biolink:Activity	9092
anatomical_entity	biolink:AnatomicalEntity	149423
behavior	biolink:Behavior	4202
biological_entity	biolink:BiologicalEntity	110
biological_process	biolink:BiologicalProcess	43888
cell	biolink:Cell	9096
cell_line	biolink:CellLine	38896
cellular_component	biolink:CellularComponent	8779
chemical_entity	biolink:ChemicalEntity	372988
chemical_mixture	biolink:ChemicalMixture	489
clinical_attribute	biolink:ClinicalAttribute	112982
clinical_finding	biolink:ClinicalFinding	11065
clinical_intervention	biolink:ClinicalIntervention	36772
cohort	biolink:Cohort	950
complex_molecular_mixture	biolink:ComplexMolecularMixture	572
device	biolink:Device	73167
disease	biolink:Disease	365181
drug	biolink:Drug	293866
event	biolink:Event	288
food	biolink:Food	9108
gene	biolink:Gene	2408081
gene_family	biolink:GeneFamily	28437
genomic_entity	biolink:GenomicEntity	42421
geographic_location	biolink:GeographicLocation	4698
gross_anatomical_structure	biolink:GrossAnatomicalStructure	12610
human	biolink:Human	46
information_content_entity	biolink:InformationContentEntity	54550
life_stage	biolink:LifeStage	1357
macromolecular_complex	biolink:MacromolecularComplex	19843
molecular_activity	biolink:MolecularActivity	198929
molecular_entity	biolink:MolecularEntity	14
molecular_mixture	biolink:MolecularMixture	11625271
named_thing	biolink:NamedThing	32
nucleic_acid_entity	biolink:NucleicAcidEntity	551
organism_taxon	biolink:OrganismTaxon	3141304
pathway	biolink:Pathway	53769
phenomenon	biolink:Phenomenon	332529
phenotypic_feature	biolink:PhenotypicFeature	88485
physical_entity	biolink:PhysicalEntity	6624
physiological_process	biolink:PhysiologicalProcess	108
polypeptide	biolink:Polypeptide	219
population_of_individual_organisms	biolink:PopulationOfIndividualOrganisms	13383
posttranslational_modification	biolink:PosttranslationalModification	1971
procedure	biolink:Procedure	449455
protein	biolink:Protein	2344320
sequence_variant	biolink:SequenceVariant	7336430
small_molecule	biolink:SmallMolecule	111621082
taxonomic_rank	biolink:TaxonomicRank	77

crossrefs

Typed associations, not equivalence. One split per source; source lives in the split name, not a column. Edges retain pre-merge IDs and may refer to rows absent from cliques. A zero-row empty split is emitted when no source contributes edges. Read that empty split with streaming=True; the datasets library rejects zero-row splits in materialized mode.

column	type	meaning
SUBJECT	string	source CURIE (pre-merge identifier, may be obsolete)
PREDICATE	string	edge type, below
OBJECT	string	target CURIE

split	source	rows
clinvar	clinvar	7530759
obo_apo	obo:APO	9
obo_fao	obo:FAO	33
obo_fbdv	obo:FBdv	8
obo_fypo	obo:FYPO	100
obo_hsapdv	obo:HsapDv	225
obo_maxo	obo:MAXO	654
obo_mmusdv	obo:MmusDv	77
obo_mp	obo:MP	749
obo_taxrank	obo:TAXRANK	96
obo_to	obo:TO	897
obo_wbls	obo:WBls	2
obo_wbphenotype	obo:WBPhenotype	42
obo_zfs	obo:ZFS	90
obo_zp	obo:ZP	52

• dbsnp_reference: ClinVar variation to the dbSNP rsID it references (one rsID can cover several alleles)
• gene_association: variant to the NCBIGene it lies in (location, not identity)
• replaced_by: obsolete OBO term to its successor
• xref: OBO term cross-reference (relevance only, per OBO 1.4)

Usage

from datasets import load_dataset

genes = load_dataset("SkyeAv/amalgam", "cliques", split="gene")
cliques = load_dataset("SkyeAv/amalgam", "cliques")  # DatasetDict, one entry per class
edges = load_dataset("SkyeAv/amalgam", "crossrefs", split="clinvar")

Construction

• Synonyms are unique per (class, taxon scope); the same string may recur across classes or taxa. Ties go to: preferred-name match, most synonyms, most identifiers, curated sources, smallest CURIE.

• Conflated BABEL files are never read; drugs and chemicals, genes and proteins stay distinct.

• A CURIE-like preferred name uses the first sorted, admitted non-CURIE synonym when available. Existing human-readable preferred names are preserved.

• biolink:Publication and biolink:Agent are dropped.

• Except biolink:OrganismTaxon, rows with taxa keep at least one of the 100 allowlisted taxa (config/taxa.yaml). Taxonless rows first inherit the taxon of single-species identifiers (MGI, RGD, ZFIN, FlyBase, WormBase, SGD, HGNC, Ensembl).

• Junk names are removed: numbers, uncharacterized, hypothetical, language, and stopwords (config/filters.yaml). Leader CURIEs and equivalent identifiers are indexed as synonyms (fullmap parity).

Generated 2026-10-28 by Amalgam.

License

No single license covers this dataset: each name keeps the terms of the resource it came from, and several (UMLS, OMIM, DrugBank, KEGG, Orphanet, HMDB) restrict redistribution or commercial use. See LICENSE.md before use.