gnomad-genetics-guide

Scope

Use this server for human population frequency and gene constraint evidence, or for ClinVar gene-level curation followed by population lookups. Core lookups and ClinVar search are keyless. It does not translate coordinates between builds: an absent coordinate ID whose string also lifts over from the other build fails as variant_not_found_other_build with the lifted IDs in candidates, a hint the server never acts on. It does not annotate structural or mitochondrial variants or submit clinical classifications.

For a single candidate, the user-invoked gnomad_variant_triage prompt already supplies the frequency, constraint, and callability workflow. The workflows below address candidate discovery and cross-source follow-up.

Identifiers

• Resolve an Ensembl gene ID to its HGNC symbol before gnomad_search_clinvar; other gene tools accept the Ensembl ID directly (versioned or lowercase forms included). An Ensembl lookup can supply the symbol.
• ClinVar search returns its own VariationID and accession for provenance. Neither is a gnomAD variant identifier; the HGVS title is not one either.
• Prefer a returned grch38_variant_id for gnomad_get_variant on a GRCh38 dataset. It is derived from the SPDI for SNVs, MNVs, and delins, and for deletions, insertions, and duplications it is filled from gnomAD's own ClinVar mapping; most of those indel IDs are absent from gnomAD, so variant_not_found there is absence evidence to pair with gnomad_get_coverage, not a failed conversion. When it is null (mitochondrial, multi-allele, copy-number, or structural records, records newer than gnomAD's ClinVar release, or a mapping notice reporting the gnomAD request failed), the row may still supply dbSNP rsIDs; split the semicolon-joined rsids into individual identifiers before lookup.
• A canonical SPDI is not a VCF-style gnomAD ID. SPDI positions are zero-based; canonical insertions and deletions can be repeat-expanded and lack the VCF anchor base. Do not construct a gnomAD ID by replacing separators or merely adding one to the position. Obtain a normalized allele on the required build from a reference-aware source instead.