---
pretty_name: GWASLab Reference
license: other
license_name: source-data-terms
task_categories:
  - other
tags:
  - gwas
  - genomics
  - 1000-genomes
  - hapmap
  - ld-reference
  - allele-frequency
---

# GWASLab reference datasets

Processed genomic reference files used by [GWASLab](https://cloufield.github.io/gwaslab/) (`download_ref` / `gwaslab download ref`). This dataset replaces the previous Dropbox hosting for GWASLab-processed panels. Official dbSNP VCFs, UCSC FASTA, Ensembl/RefSeq GTF, and liftOver chains stay at their original hosts.

Package catalog: [`reference.json`](https://github.com/Cloufield/gwaslab/blob/main/src/gwaslab/data/reference.json). Checksums for every file in this repo are in [`md5sum.txt`](md5sum.txt).

## Download with GWASLab

```python
import gwaslab as gl

gl.download_ref("1kg_eas_hg19")
print(gl.get_path("1kg_eas_hg19"))
```

```bash
gwaslab download ref 1kg_eas_hg19
gwaslab path 1kg_eas_hg19
```

Direct Hub URL (basename is the local filename `download_ref` writes):

`https://huggingface.co/datasets/Cloufield/gwaslab-reference/resolve/main/1kg/hg19/EAS.ALL.split_norm_af.1kgp3v5.hg19.vcf.gz`

## Layout

| Path | GWASLab keyword(s) | Use |
|------|--------------------|-----|
| `1kg/hg19/*.vcf.gz` (+ `.tbi`) | `1kg_{afr,amr,eas,eur,pan,sas}_hg19` | LD / strand / AF (1KGP3v5, hg19) |
| `1kg/hg38/*.vcf.gz` (+ `.tbi`) | `1kg_{afr,amr,eas,eur,pan,sas}_hg38` | LD / strand / AF (1KG 30x, hg38) |
| `rsid/1kg_dbsnp151_*_auto.txt.gz` | `1kg_dbsnp151_hg19_auto`, `1kg_dbsnp151_hg38_auto` | SNPID–rsID tables (autosomes) |
| `eaf/PAN.hapmap3.*.EAF.tsv.gz` | `1kg_hm3_hg19_eaf`, `1kg_hm3_hg38_eaf` | HapMap3 EAF for ancestry |
| `recombination/recombination_hg*.tar.gz` | `recombination_hg19`, `recombination_hg38` | Regional recombination tracks |
| `examples/t2d_bbj.txt.gz` | *(not a catalog keyword)* | Tutorial BBJ T2D sumstats |

Ancestries: AFR, AMR, EAS, EUR, SAS, PAN (all 1KG super-populations combined). Multi-allelic variants were decomposed and normalized; INFO includes population `AF`.

## Processing

1KG VCFs were processed by GWASLab for regional LD plots and strand inference. They are not a substitute for the official 1000 Genomes release files.

## Citations

- **GWASLab:** He Y, Koido M, Shimmori Y, Kamatani Y. GWASLab: a Python package for processing and visualizing GWAS summary statistics. Jxiv (2023). https://doi.org/10.51094/jxiv.305
- **1000 Genomes Project:** The 1000 Genomes Project Consortium. A global reference for human genetic variation. *Nature* (2015). 30x high-coverage data: Byrska-Bishop et al., *Cell* (2022).
- **HapMap recombination maps:** International HapMap Consortium.
- **BBJ T2D example:** Suzuki K et al. Identification of 28 new susceptibility loci for type 2 diabetes in the Japanese population. *Nat Genet* (2019). Source: http://jenger.riken.jp/

Redistribute and cite the original consortia terms for 1KG, HapMap, dbSNP-derived tables, and BBJ summary statistics.