clinvar-gkm 1.1.0

clinvar-gkm 1.1.0

The general-availability release. It reorganizes propositions by meaning, carries a consolidated set of data
corrections, and moves distribution to a new layout modeled on NCBI's ClinVar FTP site. Every published file was
rebuilt and re-published.

• Full details and migration notes: Changelog → 1.1.0
• Validation results: Release Validation

Breaking changes

• Every download URL changed. Data now lives under the line folder v1/1.1/ on data.clinvar-gkm.org:

  • monthly full bundles at the top level;
  • this month's weekly full bundles in weekly/;
  • weekly deltas in deltas/;
  • one current Parquet set in parquet/.

  1.0.1 is frozen at v1/1.0/, and the old v1/datasets/ and v1/deltas/ paths are gone. The root index.json
  lists both lines. See Downloads.

• Propositions are grouped by meaning. The four sections and Parquet tables are now propositionPathogenicity,
  propositionOncogenicity, propositionSomaticClinicalImpact and propositionClinvarCustom. They replace the
  shape-based varcond-, varcustom-, vartumor- and vartherapy-proposition. Proposition ids are unchanged;
  pointers name the new sections. The four Parquet tables share one column schema.

• The schema $id version and the bundle schema_version are 1.1.0.

• Do not apply 1.1.0 deltas to a 1.0.1 monthly. Start from a 1.1.0 monthly or weekly full.

Data corrections

• Gene context comes from the upstream single-gene method (geneContextSource records where it came from).
• SCV citations: one Document per ClinVar citation, keeping URL-only citations and adding documentType.
• Condition references are uniform #/ pointers.
• Conditions without a Preferred name are named from ClinVar's own data.
• hgvs.? expressions are kept as extensions, never altered.
• Duplicate allele keys are merged.
• RCV propositions now always carry an object.
• Every array has a total, deterministic order.

What's new

• Weekly full bundles for the current month, alongside the official monthlies and weekly deltas. Every bundle
  has a .sha256 sidecar.
• One Parquet set, rebuilt every week from the newest release, with a manifest.json.
• General availability. Published files are immutable. A correction that rewrites bundles opens a new minor line
  (with a matching schema_version), and superseded lines are frozen and kept indefinitely.
• New docs:
  • How to Read These Docs;
  • Data Use and Citation, covering ClinVar attribution and disclaimer and CC0 for code and data;
  • Stay Informed, for joining the announcement list;
  • Release Validation;
  • a reorganized six-tab site.

Validation

All 17 datasets (2026-06-27 to 2026-10-18) were rebuilt and gated before publication:

• proposition reference integrity on every dataset;
• every change against the previous build attributed to a shipped fix;
• delta reconstruction exact (0,0,0) on all 16 pairs;
• strict GA4GH GKM Toolkit (ga4gh.gkm 0.4.0) validation of the published bundles, with every section passing;
• no null or empty values;
• the Parquet set matching the bundle exactly.

Data use

The underlying data come from NCBI's ClinVar. Please attribute ClinVar and follow its
data use and disclaimer. See
Data Use and Citation.

Release announcements: clinvar-gkm@broadinstitute.org.

clinvar-gkm 1.0.1

clinvar-gkm 1.0.1

Schema-conformance patch release. Every published bundle (full + delta) now validates against the
schema/clinvar-gkm/ schemas with the official GA4GH gkm-toolkit — the largest published delta went from
7,623 failing records to 0. All 16 datasets were rebuilt, oracle-gated (full-vs-incremental +
delta-reconstruction, 0,0,0 across every conformance section), and re-published to R2 /v1/.

Schema / data fixes (PR #137)

• Unmappable variants no longer emit constraints: [{}] — the catvar procedure emits no constraints
  attribute, so haplotypes/genotypes/unmapped variants validate as ClinvarNonConstrainedVariant.
• Extension values collapsed everywhere — typed value_<type> keys (incl. nested
  classification.extensions) are collapsed to the polymorphic value recursively across every section, in
  both the JSON bundle and the typed Parquet data columns.
• Integer copies — CopyCountConstraint.copies is emitted as a JSON integer (was a quoted string from
  bq extract INT64 serialization), satisfying cat-vrs oneOf[integer, Range].
• flagged submission added to the clinvarReviewStatus / clinvarScvReviewStatus enums (FLAG tier).
• Single-member conditionSets dropped — a one-concept ConceptSet violates va-spec minItems: 2 and is
  always an unreferenced orphan; excluded from the published conditionSet section.
• Empty delta sections omitted — a zero-record section is no longer emitted as {}.
• Schema $id version + bundle schema_version 1.0.0 → 1.0.1; /v1/ R2 layout unchanged (major-keyed, ADR 0004).

Distribution (PR #138)

• Public distribution served from the stable custom domain https://data.clinvar-gkm.org behind a
  Cloudflare CDN (replacing the dev-only *.r2.dev endpoint). See
  ADR 0005.
• Release tooling gains an opt-in --purge (Cloudflare cache purge) for re-publishing already-cached paths,
  plus R2 size-aware server-side copy fixes.

Data

Re-released 4 monthly fulls (2026-07 … 2026-10) + 15 weekly deltas (2026-07-06 … 2026-10-11, a newly added
weekly). 00-latest advances to 2026-10-11.

Known limitation

A small number of VRS allele records don't perfectly delta-reconstruct (cross-version SPDI shift at 2026-08-29

• a non-deterministic allele ordering) — keys resolve and full bundles are complete; tracked for a
  deterministic-ordering fix. See
  Known Issues.

Full diffs: PR #137 (schema/data) and PR #138 (distribution).

1.0.0

ClinVar-GKM 1.0.0

First stable release of ClinVar-GKM — ClinVar variant classifications transformed into the GA4GH
Genomic Knowledge Model (GKM) schema set (VRS, Cat-VRS, VA-Spec), curated by the GA4GH GKS workstream.

Distribution

• Monthly full bundles + weekly incremental deltas, each a single gzipped JSON bundle plus typed
  Parquet, distributed via Cloudflare R2.
• Versioned layout (ADR 0004): a per-major-version prefix (/v1/), a root version manifest
  (index.json), a per-line index (v1/index.json), and an Atom feed (v1/feed.xml).
• Current data window: monthly fulls 2026-07 … 2026-10; weekly deltas 2026-07-06 … 2026-10-04.

Highlights

• Self-identifying bundles — every bundle carries a top-level schema_version: "1.0.0".
• submittedCondition / submittedConditionSet extension fields emit bare CURIEs (e.g.
  clinvar.trait:3622) instead of JSON pointers, matching their STRING typing. Object-reference pointers
  used by propositions and condition sets are unchanged.
• Complete condition references — every condition and condition-set reference resolves, including the
  content-addressed ConceptSets for multi-condition VCV aggregates.
• Versioning & release change-management policy — semantic versioning per major line, parallel release
  lines, and a published end-of-life policy. See ADR 0004 and docs/reference/versioning.md.

Schema

GKM bundle schema version 1.0.0 (VRS, Cat-VRS, and VA-Spec). The bundle self-describes via
schema_version; the JSON Schema is published under schema/clinvar-gkm/.

1.0-rc3

What's Changed

• Scope GKS output conformance initiative (+ validator parse-error fix) by @larrybabb in #82
• Conformance fixes: ClinSig predicate + MappableConcept type (gap 1) + extension value collapse (gap 2) by @larrybabb in #83
• Incremental gks_catvar + delta datasets (Plan 1: catvar + foundations) by @larrybabb in #74
• Incremental gks_scv_condition + gks_scv_statement + deltas (Plan 2) by @larrybabb in #75
• Incremental gks_rcv + gks_vcv aggregation cascade + deltas (Plan 3) by @larrybabb in #76
• Plan 4: retire gks_json + R2 delta publishing + bundle JSON cleanup by @larrybabb in #77
• Date the monthly-full Parquet layout + Parquet delta-apply docs by @larrybabb in #85
• fix(r2): size-aware server-side copy (R2 lacks a universal tag-copy mode) by @larrybabb in #86
• fix(delta): manifest truncated to 100 rows by bq default max_rows by @larrybabb in #87
• chore(schema): clinvar-gks schema version 1.x -> 1.0-rc3 by @larrybabb in #88
• chore(rebrand): clinvar-gks -> clinvar-gkm (code, docs, schema, R2) by @larrybabb in #89
• chore(rename): gks_ -> gkm_ BigQuery procs/tables + files by @larrybabb in #90
• fix(rename): revert gkm_type -> gks_type (upstream column) by @larrybabb in #91
• docs: GKS->GKM terminology (workstream vs schema set) by @larrybabb in #92
• docs: fix Parquet examples for MappableConcept struct columns by @larrybabb in #93
• docs: fix condition.parquet column ref (primary_coding → primaryCoding) by @larrybabb in #94
• feat(release): anchor monthly full trigger to ClinVar's monthly index by @larrybabb in #95
• fix(vrs): deep-copy fallback when gkm_vrs clone chain hits BigQuery's depth-3 limit by @larrybabb in #96
• design: VCV/RCV clinsig + star aggregate extensions spec by @larrybabb in #98
• docs: add Roadmap page under Reference (community-voted backlog) by @larrybabb in #106
• chore(r2): check in CORS policy config + clearer CORS error in the release browser by @larrybabb in #97
• design: release browser reorg + weekly Parquet + archiving spec by @larrybabb in #111
• docs(roadmap): label-driven auto-generation (upvote-sorted), under Reference by @larrybabb in #108
• docs(reference): populate GA4GH GKS Standards with spec + repo + python links by @larrybabb in #113
• docs: Known Issues (RC status, coverage, outdated examples) + examples refresh plan by @larrybabb in #114
• docs(download): prerequisites notes for Python/DuckDB/pandas examples by @larrybabb in #116

Full Changelog: 1.0-rc2...1.0-rc3

1.0-rc2

What's Changed

Documentation Rework

Comprehensive documentation update reflecting all structural changes from PRs #42–#54 (Parquet release pipeline).

• Docs rework: post-Parquet pipeline changes by @larrybabb in #55
• Docs: Parquet download-first guide, JOIN examples, file browser by @larrybabb in #57

Key updates

• Parquet access guide — download-first workflow (R2 rate limits), DuckDB and pandas examples, JOIN examples with typed columns vs JSON extraction tradeoffs, column reference
• Parquet file browser — static Parquet section added to the JavaScript release browser
• Confidence updated from string to Concept struct across all statement docs
• Evidence lines updated from inlined objects to FK reference pattern (hasEvidenceLines → #/evidenceLine/)
• Gene updated to MappableConcept with NCBI Gene/HGNC codings; GeneListItem now FK reference
• Bundle sections — copyNumberCount, copyNumberChange, evidenceLine added to all JSON structure examples
• Export pipeline — corrected assembler args, upload params, archive behavior; documented release-gks.sh wrapper
• VCV layer naming — "Base Grouping" → "Classification Grouping", "Tier Grouping" → "Priority Grouping"
• Miscellaneous — statement type numbering (G.01–G.09, O.10, S.11–S.14), stale text, typos

Full Changelog: 1.0-rc1...1.0-rc2

1.0-rc1

What's Changed

• Add Parquet release pipeline implementation plan by @larrybabb in #42
• Standardise output table names to gks_dict_* convention by @larrybabb in #43
• Split allele dict by VRS type, rename Non-Constrained to Undefined by @larrybabb in #44
• Restructure gks_dict_gene as MappableConcept by @larrybabb in #45
• Fix variation members FK to route by VRS type by @larrybabb in #46
• Promote condition aliases from extension to top-level field by @larrybabb in #47
• Rename condition_refs to concepts in conditionSet by @larrybabb in #48
• Promote confidence from plain string to concept struct by @larrybabb in #49
• Extract SCV evidenceLines into dict table, add methodType/methodName by @larrybabb in #50
• Add export and assembly support for new dict tables by @larrybabb in #51
• Task 0b-9: Use system URLs in coding.system output by @larrybabb in #52
• Task 0b-8: Refactor VCV/RCV to use shared evidence line dicts by @larrybabb in #53
• Add Parquet output to release pipeline by @larrybabb in #54
• Docs rework: post-Parquet pipeline changes by @larrybabb in #55
• Fix incorrect GCS bucket name and CLI examples in pipeline docs by @larrybabb in #56

Full Changelog: 1.0-beta6...1.0-rc1

1.0-beta6

What's Changed

• Add downloads page, release pipeline, and docs updates by @larrybabb in #41

Full Changelog: 1.0-beta5...1.0-beta6

1.0-beta5

What's Changed

• Restructure docs nav and add extension reference tables by @larrybabb in #40

Full Changelog: 1.0-beta4...1.0-beta5

1.0-beta4

What's Changed

• Add JSON schema profiles for ClinVar-GKS entity types by @larrybabb in #39

Full Changelog: 1.0-beta3...1.0-beta4

1.0-beta3

What's Changed

• Restructure R2 distribution with datasets/weekly layout by @larrybabb in #38

Full Changelog: 1.0-beta2...1.0-beta3