GWAS Catalog Associations

Dataset Description

This dataset contains curated genetic association results from the NHGRI-EBI GWAS Catalog, a manually curated resource of published genome-wide association studies (GWAS).

The dataset captures SNP–trait associations reported in peer-reviewed studies. Each row represents an association between a genetic variant (typically a single nucleotide polymorphism, SNP) and a disease or trait reported in a publication.

The GWAS catalog aggregates information from thousands of GWAS publications and standardises metadata about studies, genomic loci, variants, genes, and statistical significance.

This Hugging Face dataset provides a tabular representation of the association records suitable for downstream analysis, machine learning, and genomics research workflows.

Dataset Summary

• Task categories: genomics, biomedical data mining
• Data type: tabular
• Primary domain: genome-wide association studies (GWAS)
• Unit of observation: SNP–trait association
• Source: curated literature database

Typical uses include:

• genomic risk analysis
• variant annotation pipelines
• phenotype–genotype relationship studies
• machine learning on genetic associations
• meta-analysis of GWAS findings

---

Dataset Structure

Each row corresponds to a reported association between a variant and a trait.

Columns

Column	Description
DATE ADDED TO CATALOG	Date the study was added to the GWAS Catalog.
PUBMEDID	PubMed identifier for the publication reporting the association.
FIRST AUTHOR	Last name and initials of the first author of the publication.
DATE	Publication date (online/epub date if available).
JOURNAL	Abbreviated journal name in which the study appeared.
LINK	URL linking to the publication record in PubMed.
STUDY	Title of the publication reporting the GWAS.
DISEASE/TRAIT	Disease or trait investigated in the study.
INITIAL SAMPLE DESCRIPTION	Sample size and ancestry description for Stage 1 GWAS discovery cohort.
REPLICATION SAMPLE DESCRIPTION	Sample size and ancestry description for replication cohorts used to validate associations.
REGION	Cytogenetic region associated with the SNP.
CHR_ID	Chromosome number containing the SNP.
CHR_POS	Chromosomal coordinate of the SNP.
REPORTED GENE(S)	Gene(s) reported by the study authors as associated with the SNP.
MAPPED GENE(S)	Gene(s) mapped to the SNP based on genomic position. If intergenic, the nearest upstream and downstream genes are reported.
UPSTREAM_GENE_ID	Entrez Gene ID of the closest upstream gene if the SNP lies outside a gene.
DOWNSTREAM_GENE_ID	Entrez Gene ID of the closest downstream gene if the SNP lies outside a gene.
SNP_GENE_IDS	Entrez Gene ID(s) if the SNP is located within a gene. Multiple IDs indicate overlapping transcripts.
UPSTREAM_GENE_DISTANCE	Distance in base pairs from the SNP to the nearest upstream gene if intergenic.
DOWNSTREAM_GENE_DISTANCE	Distance in base pairs from the SNP to the nearest downstream gene if intergenic.
STRONGEST SNP-RISK ALLELE	SNP most strongly associated with the trait and its risk allele (or haplotype if applicable).
SNPS	Identifier of the strongest SNP; may include multiple rsIDs for haplotypes.
MERGED	Indicates whether the SNP record has been merged with another rsID (0 = no, 1 = yes).
SNP_ID_CURRENT	Current rsID identifier when the original SNP has been merged.
CONTEXT	Predicted functional context of the variant (e.g., intronic, intergenic) based on Ensembl annotations.
INTERGENIC	Indicator for whether the SNP lies in an intergenic region (0 = no, 1 = yes).
RISK ALLELE FREQUENCY	Frequency of the risk allele among control individuals (or the largest control group if multiple are available).
P-VALUE	Reported p-value for the SNP association. Values are rounded to one significant digit.
PVALUE_MLOG	Negative log10 transformation of the p-value.
P-VALUE (TEXT)	Additional context about the p-value (e.g., subgroup analyses such as sex or smoking status).
OR or BETA	Reported odds ratio (OR) or beta coefficient associated with the risk allele.
95% CI (TEXT)	Reported 95% confidence interval for the effect estimate.
PLATFORM (SNPS PASSING QC)	Genotyping platform used for Stage 1 GWAS, including notes on imputation or pooled designs where applicable.
CNV	Indicates whether the study involves copy number variation analysis (yes/no).
MAPPED_TRAIT	Mapped Experimental Factor Ontology trait for this study
MAPPED_TRAIT_URI	URI of the EFO trait
STUDY ACCESSION	Accession ID allocated to a GWAS Catalog study
GENOTYPING TECHNOLOGY	Genotyping technology/ies used in this study, with additional array information (ex. Immunochip or Exome array) in brackets.

---

Curation Process

The GWAS Catalog is curated through a combination of automated and manual processes:

1. Literature identification

   • Publications describing genome-wide association studies are identified through literature searches and author submissions.

2. Manual curation

   • Expert curators review publications and extract key information including:

     • variant identifiers (e.g., rsIDs)
     • associated traits or diseases
     • statistical significance metrics
     • effect sizes
     • sample descriptions

3. Standardisation

   • Extracted data are normalized using standardized vocabularies and identifiers where possible, including:

     • controlled trait terms, including ontology terms from the Experimental Factor Ontology (EFO)
     • genomic coordinates
     • gene identifiers
     • standardised ancestry label framework

4. Annotation

   • Variants are annotated with additional genomic information such as:

     • mapped genes
     • variant context (e.g., intronic, intergenic)
     • genomic distances to nearby genes

5. Quality control

   • Curated records undergo internal quality checks to ensure consistency, correct variant identifiers, and valid genomic annotations.

For more information about the curation process, please see our documentation

The Hugging Face dataset mirrors the tabular association records published by the GWAS Catalog on 2026-03-17.

---

Bias, Limitations, and Population Representation

Genome-wide association studies have several well-known limitations that may affect analyses using this dataset.

Population Bias

A large proportion of GWAS studies have historically been conducted with individuals genetically similar to European reference populations. Please note:

• genetic associations may not generalise across populations
• allele frequencies may differ substantially between ancestries
• effect sizes may vary across populations

Users should exercise caution when applying results derived from GWAS to diverse populations.

Publication Bias

The catalog reflects published associations, which introduces potential bias:

• studies with statistically significant findings are more likely to be published
• null results are often underrepresented
• some loci may appear more frequently because they are studied more extensively

Study Heterogeneity

GWAS included in the catalog differ in:

• sample size
• cohort composition
• genotyping platform
• statistical methodology
• phenotype definitions

These differences can influence reported effect sizes and significance levels.

---

Summary statistics

This dataset includes only GWAS-significant associations.

Full summary statistics, including variants which fail to meet GWAS significance, are available directly from the GWAS Catalog.

Summary statistics files in the GWAS Catalog undergo extensive quality control steps to improve their reusability.

---

Credits

This dataset is derived from the NHGRI-EBI GWAS Catalog.

We would like to thank:

• Authors who submit their data to the catalog, including full summary statistics
• Authors of the original GWAS publications included in the catalog
• GWAS Catalog team members, past and present
• Research participants who contributed data to the underlying genetic studies

---

Citation

If you use this dataset in research, please cite the GWAS Catalog publication:

Maria Cerezo, Elliot Sollis, Yue Ji, Elizabeth Lewis, Ala Abid, Karatuğ Ozan Bircan, Peggy Hall, James Hayhurst, Sajo John, Abayomi Mosaku, Santhi Ramachandran, Amy Foreman, Arwa Ibrahim, James McLaughlin, Zoë Pendlington, Ray Stefancsik, Samuel A Lambert, Aoife McMahon, Joannella Morales, Thomas Keane, Michael Inouye, Helen Parkinson, Laura W Harris, The NHGRI-EBI GWAS Catalog: standards for reusability, sustainability and diversity, Nucleic Acids Research, Volume 53, Issue D1, 6 January 2025, Pages D998–D1005, https://doi.org/10.1093/nar/gkae1070

@article{cerezo2025nhgri,
  title={The NHGRI-EBI GWAS Catalog: standards for reusability, sustainability and diversity},
  author={Cerezo, Maria and Sollis, Elliot and Ji, Yue and Lewis, Elizabeth and Abid, Ala and Bircan, Karatu{\u{g}} Ozan and Hall, Peggy and Hayhurst, James and John, Sajo and Mosaku, Abayomi and others},
  journal={Nucleic acids research},
  volume={53},
  number={D1},
  pages={D998--D1005},
  year={2025},
  publisher={Oxford University Press}
}

---

License

The NHGRI-EBI GWAS Catalog and all its contents are available under the general Terms of Use for EMBL-EBI Services. Summary statistics are made available under CC0 unless otherwise stated. We advise consumers of data hosted by the GWAS Catalog to note the license terms of individual datasets, if applicable to their specific use case.