frequencies.csv

Identity

Row model	FrequencyRow (just_dna_format.frequency)
Becomes	frequencies.parquet
Authored or derived	derived — an enricher pass writes it; an author corrects it via overrides.csv
Draftable	no
Written and checked by	just-dna-enricher frequencies (frequencies) from gnomad
Natural key	variant_key, population
Fact signature	yes — its sidecar carries one
In the attestation binding	no

Columns

Column	Type	Required	Values	Meaning
variant_key	str	required		Coordinate-derived identity of the allele (matches the post-expansion weights key)
rsid	str | None	optional		dbSNP identifier, when known
chrom	str | None	optional		Chromosome without 'chr' prefix
start	int | None	optional		1-based genomic position (VCF POS convention)
ref	str | None	optional		Reference allele
alt	str | None	optional		The ONE alt allele this frequency is for — deliberately singular, unlike ResolutionRow.alts: an allele frequency is per-allele, so a multi-allelic site is several rows, never one comma-joined cell.
genome_build	str	defaulted		Assembly the coordinate is in (the RM15 forward hook)
population	str	required		Ancestry group the counts are for, or 'global' for the whole dataset. Open, seeded vocabulary (vocab.RECOMMENDED_ANCESTRY_GROUPS) — a label is interpretable only together with dataset.
allele_count	int | None	optional		AC — observed copies of alt in this group
allele_number	int | None	optional		AN — total called alleles in this group (the denominator)
homozygote_count	int | None	optional		Individuals homozygous for alt in this group
hemizygote_count	int | None	optional		Hemizygous calls (X/Y outside the PAR, and MT)
faf95	float | None	optional		Filtering allele frequency, 95% CI lower bound — the number an ACMG BA1/BS1 filter actually uses. The source reports ONE faf95 with a named owning group, so it is set only on that group's row and null everywhere else (no extra column, no overloaded field).
dataset	str	required		Which release these counts are from, e.g. 'gnomad_v4.1_joint'. A FACT, not provenance: the same allele has different (equally correct) counts in different releases.
vrs_id	str | None	optional		GA4GH VRS allele id (ga4gh:VA.…)
caid	str | None	optional		ClinGen Allele Registry canonical allele id (CA<digits>)
source	str | None	optional		Which link filled this: gnomad|manual|reversed (open)
status	str | None	optional	one of: not_covered, not_found, resolved	Outcome: resolved|not_found|not_covered (VALID_FREQUENCY_STATUS)
fetched_at	str | None	optional		ISO-8601 UTC timestamp, second resolution (e.g. '2026-08-03T02:03:23Z'). Canonicalized on load; records when this row was last written by a pass, not when the source published anything

Generated from the row model at build time — reference.authoring_reference(), the same answer describe_table gives an authoring tool. Nothing on this page is hand-kept.