biodatageeks/vepyr_116_GRCh38_plugin_clinvar

vepyr plugin cache — ClinVar (GRCh38, VEP 116)

A prebuilt, frequency-tiered Parquet cache of ClinVar clinical-significance annotations for use with vepyr, the Rust/DataFusion VEP-compatible variant annotation engine. It reproduces the CSQ output of Ensembl VEP 116 run with ClinVar as a --custom annotation, without requiring the upstream VCF at annotation time.

Source version

This is the fact you most likely came here for.

Source file	clinvar.vcf.gz (GRCh38)
Source URL	https://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/archive_2.0/2026/clinvar_20260706.vcf.gz (the dated archive copy of that week's clinvar.vcf.gz)
Source MD5	f78d25d49e17a070957a127e409f87b9 (upstream clinvar_20260706.vcf.gz.md5)
Upstream release	ClinVar weekly release, VCF header fileDate=2026-07-06
Source retrieved	2026-07-06
Genome build	GRCh38, 1-based, bare (non-chr-prefixed) contigs in source
Cache built	2026-09-05
Target VEP version	Ensembl VEP 116
Build manifest	plugins/clinvar/clinvar.source.toml
vepyr-plugins tag	v0.1.1 — recorded in manifest.json as cache_source_version: v0.1.1@3e1c039
Source verification	every source file and its .tbi were MD5-verified against the manifest before the build (verified_md5 in manifest.json)

ClinVar is republished weekly and its content changes materially between releases — significance calls are reclassified, and records are added and retired. Pin your analyses to the fileDate above; do not assume this cache tracks current ClinVar.

Provenance

Rebuilt on 2026-09-05 from sources verified against the v0.1.1 manifest; shard bytes are reproducible (a second build yields identical MD5s) since the tier stage orders rows totally. The sources block in manifest.json records url, declared and verified MD5, size and index digest for each input.

Contents

chr1.parquet … chr22.parquet, chrX/chrY/chrMT.parquet  25 per-contig shards
manifest.json                   schema, CSQ field mapping, per-shard row/tier counts, source provenance

Covers chr1–chr22, chrX, chrY and chrMT. Total ≈ 85 MB, 4,439,569 rows (113,018 warm / 4,326,551 cold).

Schema

column	type	CSQ field
chrom	string	— (contig)
start	uint32	— (1-based position)
end	uint32	—
allele_string	string	— (REF/ALT, VEP-minimised)
id	string	ClinVar (ClinVar Variation ID, from the VCF ID column)
clnsig	string	ClinVar_CLNSIG
clnrevstat	string	ClinVar_CLNREVSTAT
clndn	string	ClinVar_CLNDN
clnvc	string	ClinVar_CLNVC
clnvi	string	ClinVar_CLNVI
tier	int8	— (frequency tier: 0 = warm, 1 = cold)

Six CSQ fields, in declared order — matching the VEP invocation

--custom clinvar.vcf.gz,ClinVar,vcf,exact,0,CLNSIG,CLNREVSTAT,CLNDN,CLNVC,CLNVI

which emits a bare ClinVar field (the short_name itself, carrying the source VCF ID) alongside the five named subfields.

Matching semantics

ClinVar is a per-variant annotation: it describes the variant, not a per-transcript amino-acid change, so the same values apply to every transcript line of a variant. The lookup key is (chrom, start, end, allele_string) with no additional discriminator.

Alleles are stored minimised. VEP's --custom VCF path minimises alleles before comparing (the exact in --custom …,exact is the overlap mode, not the allele rule), so indel rows here carry the anchor-base-trimmed, start-shifted form that vepyr's runtime probe produces — without which no indel would ever match.

Multi-allelic source records are split to one row per ALT allele at build time. The 2026-07-06 release contains none, but the split is applied unconditionally so a future release cannot silently produce an unmatchable G/A|C allele string.

An explicit . in a source INFO field is preserved as .; an absent INFO key yields an empty CSQ field. This reproduces VEP --custom behaviour, which typed-column parsing alone cannot distinguish.

Frequency tiering

Each shard is sorted by (tier, start). A row's tier is inherited row-for-row from the release-116 GRCh38 variation cache this plugin cache was built against: a plugin row takes the tier of its matching variation row, and a plugin row with no match there is cold. tier = 0 is warm — 109,838 rows, 2.6% of the cache; tier = 1 is cold. Because warm rows are physically contiguous at the front of the file, a warm-only probe touches a handful of row groups instead of scanning the shard. Per-shard warm/cold counts are in manifest.json.

Quality profile

Generated 2026-09-05 by profile_plugin_cache.py (vepyr 0.4.0, Polars 1.39.3) from the shards in this commit; machine-readable copy in qa_profile.json.

Invariants

check	status	detail
schema	✅ pass	25 shards match the manifest
contig	✅ pass	0 foreign-contig rows in 25 shards
order	✅ pass	0 descending steps in 25 shards
tier_domain	✅ pass	0 rows with tier outside {0,1} in 25 shards
manifest_counts	✅ pass	rows/warm/cold match in 25 shards
manifest_files	✅ pass	25 manifest contigs, no stray shards
positions	✅ pass	0 rows with start < 1 or end < start - 1 in 25 shards
allele_form	⚠️ warn	1102 rows with an empty ALT (source ALT '.') in 25 shards
duplicates	✅ pass	0 duplicate probe keys in 25 shards (manifest assume_unique=false)

Contigs

contig	rows	warm	cold	warm %	size
chr1	401,099	10,445	390,654	2.6%	7.6 MB
chr10	164,036	5,005	159,031	3.1%	3.1 MB
chr11	259,912	6,038	253,874	2.3%	5.0 MB
chr12	204,104	5,778	198,326	2.8%	3.9 MB
chr13	87,747	2,146	85,601	2.4%	1.8 MB
chr14	140,942	3,575	137,367	2.5%	2.7 MB
chr15	153,974	4,202	149,772	2.7%	2.9 MB
chr16	228,695	4,809	223,886	2.1%	4.4 MB
chr17	267,893	5,980	261,913	2.2%	5.2 MB
chr18	74,024	2,471	71,553	3.3%	1.4 MB
chr19	254,711	5,983	248,728	2.3%	4.6 MB
chr2	387,847	9,248	378,599	2.4%	7.7 MB
chr20	96,012	2,667	93,345	2.8%	1.8 MB
chr21	49,937	1,638	48,299	3.3%	943 KB
chr22	96,223	2,701	93,522	2.8%	1.8 MB
chr3	243,036	6,201	236,835	2.6%	4.6 MB
chr4	157,998	4,651	153,347	2.9%	3.0 MB
chr5	215,274	5,760	209,514	2.7%	4.1 MB
chr6	199,635	6,184	193,451	3.1%	3.8 MB
chr7	215,874	5,487	210,387	2.5%	4.2 MB
chr8	150,271	3,786	146,485	2.5%	2.9 MB
chr9	196,308	5,083	191,225	2.6%	3.7 MB
chrMT	3,125	0	3,125	0.0%	73 KB
chrX	190,214	3,173	187,041	1.7%	3.7 MB
chrY	678	7	671	1.0%	12 KB
total	4,439,569	113,018	4,326,551	2.5%	85 MB

Columns

column	role	type	null %	empty %	distinct	numeric (min / p50 / p95 / max)	top values
id	value	String	0.00	0.00	~4.2M	2.000 / 2554709.000 / 4630707.000 / 4857410.000	—
clnsig	value	String	5.49	0.00	105	—	—
clnrevstat	value	String	5.49	0.00	10	—	criteria_provided,_single_submitter (3.2M), criteria_provided,_multiple_submitters,_no_conflicts (663K), None (244K), criteria_provided,_conflicting_classifications (164K), no_assertion_criteria_provided (105K), reviewed_by_expert_panel (22K), no_classification_provided (6.8K), no_classification_for_the_single_variant (668), no_classifications_from_unflagged_records (146), practice_guideline (51)
clndn	value	String	5.50	0.00	~205K	—	—
clnvc	value	String	0.00	0.00	8	—	single_nucleotide_variant (4.1M), Deletion (162K), Duplication (69K), Microsatellite (39K), Indel (19K), Insertion (14K), Inversion (1.5K), Variation (459)
clnvi	value	String	35.04	0.00	~2.9M	—	—

Usage

hf download biodatageeks/vepyr_116_GRCh38_plugin_clinvar \
  --repo-type dataset --local-dir ~/vepyr_plugin_cache/plugin/clinvar

The files are plain Parquet — usable directly from DuckDB, Polars or DataFusion independently of vepyr:

SELECT start, allele_string, id, clnsig, clnrevstat, clndn
FROM 'chr21.parquet'
WHERE clnsig LIKE '%Pathogenic%';

Licence and intended use

ClinVar is produced by the NCBI and placed in the public domain; see NCBI's ClinVar maintenance and use policy. This cache is a format conversion of the selected INFO fields and adds no new assertions.

ClinVar submissions vary widely in evidence quality — always read CLNREVSTAT (review status) alongside CLNSIG. This dataset is for research and software development; it is not a clinical decision-making tool, and a significance call here is not a diagnosis.

Citation

Landrum, M. J., Lee, J. M., Benson, M., et al. ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Research 46, D1062–D1067 (2018). doi:10.1093/nar/gkx1153