gnomAD allele frequency lookup by rsID on the CLI or the open API.

What a gnomAD allele frequency lookup actually measures

gnomAD aggregates sequencing from many studies. For a variant it reports how often the alternate allele appears among confidently sequenced haplotypes, globally and by ancestry group.

That number is an observation about sequenced people after quality filters. It is not a patient registry. It is not a ClinVar interpretation. It is not your genotype.

ACMG/AMP uses frequency as evidence. A high frequency in a continental population of enough alleles can support a benign interpretation for a severe Mendelian condition. A frequency of zero does not prove pathogenicity.

• gnomAD browser
• ClinVar vs gnomAD

AC, AN, AF, and ancestry groups

Allele count (AC) is how many times the alternate allele was observed. Allele number (AN) is how many alleles were confidently sequenced at the site. Allele frequency (AF) is AC divided by AN when AN is greater than zero.

Global AF can hide a common allele in one group. When the JSON includes group frequencies, read the maximum relevant group, not only the pooled number. A 0.1% global AF with a 4% frequency in one ancestry group is a different object.

Coverage matters. A site with tiny AN is a weak frequency statement. Homozygous counts and hemizygous counts, when present, are extra context, not a substitute for a local file.

• AC: alternate allele count
• AN: alleles sequenced with confidence
• AF: AC / AN, global and by group when present
• FAF: filtering allele frequency in the gnomAD browser

Query frequency with genome.sh

The general query and the dedicated gnomAD endpoint both expose frequency when the database includes gnomAD. lite may omit it. standard and full are the tiers that carry population frequencies.

Install the crate genome-sh; the binary is genome. Then run genome db install standard and query an rsID the same way you would for ClinVar. JSON is the form that keeps AC, AN, and group fields intact.

Prefer the rsID. rs429358, rs1800562, rs1799945, and rs334 are the sites people actually type. Coordinates require the assembly that matches the gnomAD version you are reading.

cargo install genome-sh
genome db install standard
genome query rs429358 --format json
genome query rs1800562 --format json | jq .

• How to look up an rsID
• Install genome.sh

The dedicated /v1/gnomad endpoint

GET https://api.genome.sh/v1/gnomad/{id} returns population frequencies for an rsID. No API key. The server still refuses VCF, BAM, and consumer DNA files.

GET /v1/query/{id} joins ClinVar, dbSNP, and gnomAD when those sources are loaded. GET /v1/sources lists versions. Use /v1/gnomad when frequency is the only field you want to cache.

For 100,000 rsIDs, install a local database. Remote frequency lookup is for interactive work and small scripts, not for annotating a whole VCF over HTTPS.

curl -s https://api.genome.sh/v1/gnomad/rs429358 | jq .
curl -s https://api.genome.sh/v1/query/rs334 | jq .
curl -s https://api.genome.sh/v1/sources | jq .

• Genetic variant REST API
• API documentation

Frequency next to ClinVar, not instead of it

A variant can be common and still listed in ClinVar, including founder alleles and low-penetrance risk alleles. Do not discard a ClinVar record only because gnomAD is non-zero.

Frequency answers how rare the allele is among sequenced haplotypes. ClinVar answers what labs have submitted. You need both, then a local file if you care about a person.

rs1800562 (HFE C282Y) is the usual example: not ultra-rare in European ancestry gnomAD, and still a ClinVar staple. Read both records. Do not turn either into a care plan from a CLI row.

• HFE rs1800562 and rs1799945
• ClinVar clinical significance

GRCh37 v2 versus GRCh38 v4

gnomAD v2 is GRCh37. gnomAD v4 is GRCh38. Mixing a GRCh37 VCF coordinate with a GRCh38 frequency table maps the wrong site.

An rsID is the stable key across those builds. If your file has only chr:pos, lift over or query with the assembly that matches the file header.

genome.sh reports the fields present in its snapshot. Check genome db stats or GET /v1/sources for the bundled gnomAD version. Open the gnomAD variant page when you need the full frequency table, FAF, and coverage plots.

• gnomAD v2: GRCh37 coordinates
• gnomAD v4: GRCh38 coordinates
• rsID: the key that survives a lift-over
• genome db stats or GET /v1/sources: which snapshot you have

Filtering allele frequency and founder alleles

Filtering allele frequency (FAF) in the gnomAD browser is a conservative estimate used in rare-disease work. It is not always copied into every downstream JSON field. If your pipeline needs FAF, confirm it on the gnomAD page.

Founder alleles break naive rare-equals-pathogenic logic. A relatively common allele in one population can still have a submitted ClinVar interpretation. Read ancestry-group AF before you apply a global cutoff.

Chip files make this worse. A consumer array samples a biased subset of sites, often including common GWAS SNPs. Frequency on the chip is not frequency in gnomAD, and missing chip sites are not called.

When zero in gnomAD is not evidence

Absence can mean rare. It can also mean poorly covered, quality-filtered, or not in this gnomAD version. Do not promote a missing frequency to pathogenic.

Indels and structural variants are under-counted relative to SNVs. A novel VCF allele with no rsID may have no gnomAD row. Query chr:pos:ref:alt, then still treat a miss as unknown, not as proof.

genome.sh will not invent a frequency. If the local row has no AF, say so. Match the identifier to a local genotype only after you have the public numbers you actually retrieved.

Questions