Datasets:

License:
File size: 5,384 Bytes
87a81d8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e52b1da
 
 
 
87a81d8
ac7a666
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ecb09ea
 
 
 
 
 
 
 
 
 
 
 
 
 
bf4556b
 
 
 
ecb09ea
87a81d8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ac7a666
 
 
 
 
 
87a81d8
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
---
pretty_name: Elsie prepared workflow data
license: other
license_name: component-specific-public-data-terms
license_link: https://huggingface.co/datasets/dev-prescience/elsie-workflow-cache/blob/main/README.md#source-terms
---

# Elsie prepared workflow data

Versioned public reference data for Elsie workflow installations. No participant
reads, genotypes or analysis results are included.

## Available artifacts

The Talos GRCh38 October 2026 **public subset** contains prepared AlphaMissense,
MANE/gene regions and ClinVar/ClinvArbitration tables. See its `manifest.json`
for the preparation date, exact file list, measured sizes and SHA-256 hashes.
It excludes intermediate VCFs, APOGEE/nAPOGEE/MitoTIP archives and PanelApp data.
Talos installation obtains and prepares those remaining small components locally
from the original resources. Its separate large reference pack is still needed.

Rare Disease GRCh38 public population annotations are available under
`nf-core-raredisease/GRCh38/2026-10-27/`. Its manifest records measured sizes,
source versions, site counts and checksums. Reference sequences, indexes, VEP
and provider support files remain separately acquired resources.

The **SNP-array reference sites** (`snp-arrays/GRCh38/dbsnp-b157/`) are dbSNP
build 157 GRCh38 records (CHROM, POS, ID, REF, ALT only) for converting
consumer SNP-array exports to VCF. They come as three packs, each with its
own `manifest.json`:

- `consumer-chips` (15 MB, installed by default): the rsIDs carried by
  23andMe v2–v5, AncestryDNA, MyHeritage, FamilyTreeDNA, Dynamic DNA/Illumina
  GSA and Illumina HumanOmni2.5-8v1_A exports.
- `omni25-kgp` (16 MB, optional): records at the GRCh38 positions of
  HumanOmni2.5-8v1_A markers named without an rsID.
- `imputed` (64 MB, optional): the rsIDs of imputed exports such as Genes for
  Good.

Each optional pack is complete on top of `consumer-chips`. `rsid-merges.tsv.gz`
maps exported rsIDs that dbSNP has since merged to their current rsID. Chip
position maps are not included; installs download them from their source.

## RUO datasets

`ruo/` holds data for the RUO app's plugins. Each dataset has a
`catalog.json` listing every version, newest first, signed with the RUO
publisher key (`catalog.json.minisig`, minisign). A device trusts only a
catalog signed by the key its plugin names, then checks each version's
`dataset.json` and files against the SHA-256 hashes that chain from it.

- `ruo/clinvar-arrays/GRCh38/`: ClinVar's GRCh38 release restricted to the
  sites consumer SNP arrays measure (the `consumer-chips` pack above), with
  the INFO fields a report needs (`CLNSIG`, `CLNREVSTAT`, `GENEINFO`,
  `CLNDN`, `RS`). 2026-11-01: 144,198 of 4,616,247 records, 3.2 MB. Built by
  `plugins/clinvar/publish` in the RUO workspace, byte-for-byte from the
  release and the sites, so later versions can ship as patches.
- `ruo/snp-array-sites/GRCh38/`: the `consumer-chips` pack above
  (`sites.vcf.gz` and `rsid-merges.tsv.gz`, unchanged) as a signed RUO
  dataset, for converting SNP-array exports to VCF on the device. Published
  by `plugins/publish` in the RUO workspace.

## Source terms

- AlphaMissense predictions: Google DeepMind, Cheng et al., Science (2023),
  doi:10.1126/science.adg7492. CC BY 4.0. Predictions are reformatted and encoded
  into Talos's echtvar archive; original prediction values are represented using
  the upstream encoding configuration.
  https://github.com/google-deepmind/alphamissense#alphamissense-predictions-license
- ClinVar: NCBI/NLM/NIH and its submitters. Attribute ClinVar and its release;
  retain the original data-use and clinical-review notices. ClinVar classifications
  are transformed by Talos/ClinvArbitration into Hail tables; these are derived
  data, not an official NCBI release. NIH does not independently verify submitted
  information. The information is not intended for direct diagnostic use or
  medical decision-making without review by a genetics professional.
  https://www.ncbi.nlm.nih.gov/clinvar/docs/maintenance_use/
- MANE and gene-region mappings: NCBI/Ensembl reference data, transformed by Talos.
  Preserve source attribution. Ensembl-generated data have no use restrictions;
  third-party data retain their own terms.
  https://useast.ensembl.org/info/about/legal/disclaimer.html
- dbSNP: National Center for Biotechnology Information, National Library of
  Medicine. dbSNP build 157 (GRCh38), retrieved from
  https://ftp.ncbi.nih.gov/snp/archive/b157/ ; records are subset and reduced
  to CHROM, POS, ID, REF and ALT, and contigs renamed to `chr` style. NCBI
  molecular data are not subject to copyright restrictions:
  https://www.ncbi.nlm.nih.gov/home/about/policies/
- Talos's preparation workflow: Population Genomics, MIT-licensed source at
  https://github.com/populationgenomics/talos/tree/2085846814d4e92df9e9fcc14dcc6af466f27a9b

The cache does not grant additional rights over any upstream data and does not
imply endorsement by the providers. APOGEE and nAPOGEE scores are excluded from
the published Talos subset; their current provider terms specify CC BY-NC 4.0
and require a separate licence for commercial use:
https://mitimpact.mcb2lab.org/license

## Reproducibility

Installers use a fixed Hugging Face Git revision, exact archive length and
SHA-256. Fallback source-preparation recipes stay in the workflow packages.