Gwas Database

GWAS Catalog Database — SNP-Trait Association Queries

Overview

The NHGRI-EBI GWAS Catalog is a curated collection of published genome-wide association studies, mapping SNP-trait associations with genomic context. The REST API provides programmatic access to studies, associations, variants, traits, genes, and summary statistics. All responses are HAL+JSON with embedded _links for pagination.

When to Use

• Finding genetic variants associated with a disease or trait (e.g., "which SNPs are linked to type 2 diabetes?")
• Retrieving genome-wide significant associations for a specific variant (rs ID)
• Exploring the genetic architecture of complex traits (number of loci, effect sizes)
• Checking variant pleiotropy (how many traits a single SNP affects)
• Downloading summary statistics for meta-analysis or polygenic risk score construction
• Identifying published GWAS studies by disease, gene, or PubMed ID
• Cross-referencing EFO trait ontology terms with GWAS evidence
• Building candidate gene lists from GWAS association regions
• Use omics-plotting SKILL to render Manhattan, QQ, and forest plots from the summary statistics / association results
• For drug target validation from GWAS hits, use opentargets-database instead
• For variant functional annotation (consequence prediction, regulatory impact), use Ensembl VEP via gget

Prerequisites

pip install requests matplotlib numpy

API access:

• No authentication required -- fully open access
• Rate limits: no official limit, but add time.sleep(0.2) between requests to be courteous
• Base URL: https://www.ebi.ac.uk/gwas/rest/api
• Response format: HAL+JSON with _embedded data and _links for pagination
• Pagination: default 20 results per page; max 500 via size parameter

Quick Start

import requests
import time

BASE = "https://www.ebi.ac.uk/gwas/rest/api"

def gwas_get(endpoint, params=None):
    """GWAS Catalog REST API helper with rate limiting and pagination support."""
    url = f"{BASE}/{endpoint}"
    resp = requests.get(url, params=params or {})
    resp.raise_for_status()
    time.sleep(0.2)
    return resp.json()

# Find studies for a trait keyword. Study records have no top-level `title`
# — the publication title lives at `publicationInfo.title`; the trait label
# lives at `diseaseTrait.trait`.
data = gwas_get("studies/search/findByDiseaseTrait", {"diseaseTrait": "diabetes"})
studies = data["_embedded"]["studies"]
print(f"Found {len(studies)} studies for 'diabetes'")
for s in studies[:3]:
    title = (s.get("publicationInfo") or {}).get("title", "N/A")
    trait = (s.get("diseaseTrait") or {}).get("trait", "N/A")
    print(f"  {s['accessionId']} | {trait[:40]:<40} | {title[:60]}")

Core API

Module 1: Study Search

Search GWAS studies by disease trait keyword or PubMed ID.

# Search studies by disease trait
data = gwas_get("studies/search/findByDiseaseTrait", {"diseaseTrait": "breast cancer"})
studies = data["_embedded"]["studies"]
for s in studies[:5]:
    pi = s.get("publicationInfo") or {}
    print(f"  {s['accessionId']} | PMID:{pi.get('pubmedId','N/A')} | {pi.get('title','')[:60]}")

time.sleep(0.2)

# Search by PubMed ID. NOTE: the older `findByPubmedId` 404s on /studies/;
# the working endpoint is `findByPublicationIdPubmedId`.
data = gwas_get("studies/search/findByPublicationIdPubmedId", {"pubmedId": "25673413"})
studies = data["_embedded"]["studies"]
print(f"Studies from PMID 25673413: {len(studies)}")
for s in studies:
    trait = (s.get("diseaseTrait") or {}).get("trait", "N/A")
    print(f"  {s['accessionId']}: {trait}")

Module 2: Association Queries

Retrieve SNP-trait associations filtered by trait (EFO term), variant, or p-value.

# Associations by EFO trait. The old path `efoTraits/{shortForm}/associations`
# also works *if* you have the current shortForm — but trait shortForms have
# been re-mapped to MONDO (e.g. EFO_0000249 → MONDO_0004975). The most reliable
# path is `associations/search/findByEfoTrait?efoTrait=<canonical trait name>`.
data = gwas_get("associations/search/findByEfoTrait",
                {"efoTrait": "type 2 diabetes mellitus", "size": 50})
assocs = data["_embedded"]["associations"]
print(f"Associations for 'type 2 diabetes mellitus': {len(assocs)}")

for a in assocs[:5]:
    pval = a.get("pvalue", None)
    genes = []
    for locus in a.get("loci", []) or []:
        for gene in locus.get("authorReportedGenes", []) or []:
            genes.append(gene.get("geneName", ""))
    loci = a.get("loci") or [{}]
    snps = [r.get("snps", [{}])[0].get("rsId", "N/A")
            for r in (loci[0].get("strongestRiskAlleles") or [])]
    print(f"  rs={snps} | p={pval} | genes={genes}")

# Associations for a specific variant. NOTE: association records do not embed
# `efoTraits` inline — they expose them via the `_links.efoTraits.href`
# HAL link. Follow the link (cached if needed) to resolve trait names.
data = gwas_get("singleNucleotidePolymorphisms/rs7903146/associations", {"size": 5})
assocs = data["_embedded"]["associations"]
print(f"Associations for rs7903146 (first page): {len(assocs)}")

def association_traits(assoc):
    """Resolve efoTraits via the HAL link on an association record."""
    href = (assoc.get("_links") or {}).get("efoTraits", {}).get("href")
    if not href:
        return []
    r = requests.get(href, timeout=15)
    if not r.ok:
        return []
    return [t.get("trait") for t in r.json().get("_embedded", {}).get("efoTraits", [])]

for a in assocs[:5]:
    traits = association_traits(a)
    print(f"  p={a.get('pvalue')} | OR={a.get('orPerCopyNum', 'N/A')} | traits={traits}")
    time.sleep(0.1)

Module 3: Variant Lookup

Query variant details by rsID, chromosomal region, or cytogenetic band.

# Lookup single variant
data = gwas_get("singleNucleotidePolymorphisms/rs7903146")
loc = data.get("locations", [{}])[0]
print(f"rs7903146: chr{loc.get('chromosomeName', '?')}:{loc.get('chromosomePosition', '?')}")
print(f"  Functional class: {data.get('functionalClass', 'N/A')}")
print(f"  Merged into: {data.get('merged', 'N/A')}")

time.sleep(0.2)

# Search variants by chromosomal region
data = gwas_get("singleNucleotidePolymorphisms/search/findByChromBpLocationRange",
                {"chrom": "10", "bpStart": "114750000", "bpEnd": "114800000", "size": 50})
snps = data["_embedded"]["singleNucleotidePolymorphisms"]
print(f"Variants in chr10:114750000-114800000: {len(snps)}")
for v in snps[:5]:
    print(f"  {v['rsId']}: {v.get('functionalClass', 'N/A')}")

# Search variants by gene name (the cytogenetic-band endpoint
# `findByCytogeneticBand` was removed — use gene or chromosome-range instead).
data = gwas_get("singleNucleotidePolymorphisms/search/findByGene",
                {"geneName": "TCF7L2", "size": 5})
snps = data["_embedded"]["singleNucleotidePolymorphisms"]
print(f"Variants in TCF7L2: {len(snps)}")

Module 4: Trait Search

Browse and search EFO-mapped traits in the GWAS Catalog.

# Search traits by exact name (the older `findByDescription` endpoint was
# removed — search/efoTrait now expects the canonical trait label).
data = gwas_get("efoTraits/search/findByEfoTrait", {"trait": "Alzheimer disease"})
traits = data["_embedded"]["efoTraits"]
print(f"Traits matching 'Alzheimer disease': {len(traits)}")
for t in traits[:5]:
    print(f"  {t['shortForm']}: {t['trait']}  (uri={t['uri']})")

time.sleep(0.2)

# Get specific trait by shortForm. NOTE: many legacy EFO IDs have been
# re-mapped to MONDO (e.g. old `EFO_0000249` for Alzheimer is now
# `MONDO_0004975` — `efoTraits/EFO_0000249` returns 404). Resolve via search
# above first, then use the current shortForm:
short_form = traits[0]["shortForm"]   # e.g. 'MONDO_0004975'
data = gwas_get(f"efoTraits/{short_form}")
print(f"Trait: {data['trait']}")
print(f"  URI       : {data['uri']}")
print(f"  shortForm : {data['shortForm']}")

Module 5: Summary Statistics

Access study-level summary statistics for downstream analysis (meta-analysis, PRS).

# List studies with available summary statistics
data = gwas_get("studies/search/findByFullPvalueSet", {"fullPvalueSet": True, "size": 10})
studies = data["_embedded"]["studies"]
print(f"Studies with summary stats (first page): {len(studies)}")
for s in studies[:5]:
    trait = (s.get("diseaseTrait") or {}).get("trait", "N/A")
    print(f"  {s['accessionId']}: {trait[:50]}")

time.sleep(0.2)

# Summary statistics metadata is NOT exposed via the REST API
# (`studies/{acc}/summaryStatistics` returns 404). Use the GWAS Catalog FTP
# directly — paths are predictable by study accession:
ftp_base = "http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics"
acc = studies[0]["accessionId"]
ftp_url = f"{ftp_base}/{acc[:-3]}001-{acc[:-3]}999/{acc}/"
print(f"Summary stats FTP directory: {ftp_url}")

# Download summary statistics file (FTP)
# Summary statistics are hosted on the GWAS Catalog FTP, not the REST API
import urllib.request

study_id = "GCST006867"  # Example study
ftp_base = "http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics"
# Actual paths vary by study; check the study page for the download link
# Example: ftp_base/{study_id}/{study_id}.tsv.gz
url = f"{ftp_base}/{study_id}"
print(f"Summary stats FTP directory: {url}")
# Use requests.get() or urllib to download the .tsv.gz file

Module 6: Gene and Publication Search

Find GWAS associations by gene name or retrieve publication metadata.

# Search associations by gene
data = gwas_get("singleNucleotidePolymorphisms/search/findByGene",
                {"geneName": "BRCA1", "size": 50})
snps = data["_embedded"]["singleNucleotidePolymorphisms"]
print(f"Variants near BRCA1: {len(snps)}")
for v in snps[:5]:
    locs = v.get("locations", [{}])
    pos = locs[0].get("chromosomePosition", "?") if locs else "?"
    print(f"  {v['rsId']}: chr{locs[0].get('chromosomeName', '?')}:{pos}")

time.sleep(0.2)

# Get study publication details
data = gwas_get("studies/GCST000392")
pub = data.get("publicationInfo", {})
print(f"Study: {data['accessionId']}")
print(f"  Author: {pub.get('author', {}).get('fullname', 'N/A')}")
print(f"  Journal: {pub.get('publication', 'N/A')}")
print(f"  PMID: {pub.get('pubmedId', 'N/A')}")
print(f"  Date: {pub.get('publicationDate', 'N/A')}")

Key Concepts

Data Entities and Relationships

The GWAS Catalog organizes data as interconnected entities:

Relationships: Study --(reports)--> Association --(involves)--> Variant + Trait. Variants map to genomic positions and nearby genes.

HAL+JSON Response Structure

All API responses follow HAL (Hypertext Application Language) format:

Genome-wide Significance

The standard genome-wide significance threshold is p <= 5 x 10^-8, correcting for approximately 1 million independent tests across the human genome. Associations below this threshold are considered suggestive. The GWAS Catalog includes associations at various significance levels -- always check p-values when filtering results.

Key Identifiers

• GCST IDs: GWAS Catalog study accessions (e.g., GCST000392)
• rs numbers: dbSNP reference SNP identifiers (e.g., rs7903146)
• EFO terms: Experimental Factor Ontology for trait standardization (e.g., EFO_0001360 for type 2 diabetes)
• Cytogenetic bands: Chromosomal location notation (e.g., 10q25.2)

Common Workflows

Workflow 1: Disease Genetic Architecture

Goal: Map the genetic landscape of a disease by collecting all genome-wide significant loci.

import requests, time

BASE = "https://www.ebi.ac.uk/gwas/rest/api"

def gwas_get(endpoint, params=None):
    url = f"{BASE}/{endpoint}"
    resp = requests.get(url, params=params or {})
    resp.raise_for_status()
    time.sleep(0.2)
    return resp.json()

# Step 1: Resolve trait → current shortForm via findByEfoTrait
# (the older `findByDescription` endpoint was removed).
traits = gwas_get("efoTraits/search/findByEfoTrait", {"trait": "schizophrenia"})
efo_id = traits["_embedded"]["efoTraits"][0]["shortForm"]
print(f"Using EFO: {efo_id}")

# Step 2: Get all associations for this trait — nested path
# `efoTraits/{shortForm}/associations` works once you have the canonical shortForm.
all_assocs = []
page = 0
while True:
    data = gwas_get(f"efoTraits/{efo_id}/associations",
                    {"size": 500, "page": page})
    assocs = data["_embedded"]["associations"]
    all_assocs.extend(assocs)
    if page >= data["page"]["totalPages"] - 1:
        break
    page += 1

# Step 3: Filter genome-wide significant
significant = [a for a in all_assocs if a.get("pvalue") and a["pvalue"] < 5e-8]
print(f"Total associations: {len(all_assocs)}, genome-wide significant: {len(significant)}")

# Step 4: Extract variant and effect details
for a in significant[:10]:
    risk_alleles = a.get("loci", [{}])[0].get("strongestRiskAlleles", [])
    snp = risk_alleles[0].get("snps", [{}])[0].get("rsId", "N/A") if risk_alleles else "N/A"
    or_val = a.get("orPerCopyNum", "N/A")
    beta = a.get("betaNum", "N/A")
    print(f"  {snp} | p={a['pvalue']:.2e} | OR={or_val} | beta={beta}")

Workflow 2: Variant Pleiotropy Analysis

Goal: Determine how many distinct traits a single variant is associated with.

import requests, time

BASE = "https://www.ebi.ac.uk/gwas/rest/api"

def gwas_get(endpoint, params=None):
    url = f"{BASE}/{endpoint}"
    resp = requests.get(url, params=params or {})
    resp.raise_for_status()
    time.sleep(0.2)
    return resp.json()

rs_id = "rs7903146"  # Well-known pleiotropic variant

# Get all associations for this variant
data = gwas_get(f"singleNucleotidePolymorphisms/{rs_id}/associations", {"size": 500})
assocs = data["_embedded"]["associations"]

# Association records don't embed efoTraits inline — follow the HAL
# `_links.efoTraits.href` link per association. Cache per-href to avoid
# duplicate fetches.
trait_set = {}
href_cache = {}
def fetch_traits(href):
    if href in href_cache:
        return href_cache[href]
    r = requests.get(href, timeout=15)
    href_cache[href] = (r.json().get("_embedded", {}).get("efoTraits", [])) if r.ok else []
    return href_cache[href]

for a in assocs:
    href = (a.get("_links") or {}).get("efoTraits", {}).get("href")
    if not href:
        continue
    for t in fetch_traits(href):
        tid = t.get("shortForm", "unknown")
        if tid not in trait_set:
            trait_set[tid] = {"trait": t.get("trait", "N/A"),
                              "best_pval": a.get("pvalue", 1), "count": 0}
        trait_set[tid]["count"] += 1
        if a.get("pvalue") and a["pvalue"] < trait_set[tid]["best_pval"]:
            trait_set[tid]["best_pval"] = a["pvalue"]
    time.sleep(0.1)

print(f"{rs_id} is associated with {len(trait_set)} distinct traits:")
for tid, info in sorted(trait_set.items(), key=lambda x: x[1]["best_pval"]):
    print(f"  {tid}: {info['trait']} (best p={info['best_pval']:.2e}, n={info['count']})")

Workflow 3: Summary Statistics Manhattan Plot

Goal: Download summary statistics for a study and create a Manhattan plot.

import numpy as np
import pandas as pd

# Summary statistics as a CHR/BP/P table (simulated here for demonstration).
# In practice: download from GWAS Catalog FTP, e.g.
# url = "http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCSTXXXXXX/..."
# import gzip, requests; data = gzip.decompress(requests.get(url).content)
np.random.seed(42)
n_snps = 5000
pvalues = np.random.uniform(0, 1, size=n_snps)
pvalues[:20] = 10 ** np.random.uniform(-15, -8, size=20)   # add some "hits"
sumstats = pd.DataFrame({
    "CHR": np.random.choice(range(1, 23), size=n_snps),
    "BP": np.random.randint(1, 250_000_000, size=n_snps),
    "P": pvalues,
})
sumstats.to_csv("gwas.csv", index=False)
print(f"Summary stats: {len(sumstats)} SNPs, {(sumstats['P'] < 5e-8).sum()} genome-wide sig -> gwas.csv")
# Render with the omics-plotting SKILL (`skills/data-visualization/omics-plotting/SKILL.md`) "Manhattan" recipe -> figures/manhattan_plot.png

Key Parameters

Best Practices

1. Paginate large result sets: Default page size is 20; set size=500 and loop over pages for complete data retrieval. Check page.totalElements to know the full count before iterating.

2. Use EFO IDs for precise trait queries: Free-text search may return related but different traits. Look up the exact EFO ID first, then query associations by EFO ID for precision.

3. Always check p-values: The catalog contains associations at various significance levels. Filter to p < 5e-8 for genome-wide significant results unless you specifically need suggestive associations.

4. Be ancestry-aware: Effect sizes and allele frequencies vary across populations. Check the initialSampleSize and replicationSampleSize fields to understand the ancestry composition of each study.

5. Add rate-limiting delays: Although no official limit exists, time.sleep(0.2) between requests prevents server overload and avoids temporary blocks.

6. Cache frequently accessed data: Study and trait metadata rarely change. Cache results locally when running batch analyses to reduce redundant API calls.

7. Anti-pattern -- Don't rely solely on reported genes: Author-reported genes may not be the causal gene. Cross-reference with functional annotation tools and eQTL databases for biological interpretation.

Common Recipes

Recipe 1: Cross-Reference with PGS Catalog

Identify polygenic scores available for a GWAS trait.

import requests, time

# Step 1: Get EFO/MONDO shortForm from GWAS Catalog
# (the old `findByDescription` endpoint was removed; use `findByEfoTrait`)
BASE = "https://www.ebi.ac.uk/gwas/rest/api"
traits = requests.get(f"{BASE}/efoTraits/search/findByEfoTrait",
                      params={"trait": "coronary artery disease"}).json()
efo_id = traits["_embedded"]["efoTraits"][0]["shortForm"]
time.sleep(0.2)

# Step 2: Query PGS Catalog for scores using same EFO
pgs_url = f"https://www.pgscatalog.org/rest/score/search?trait_id={efo_id}"
pgs_data = requests.get(pgs_url).json()
print(f"PGS scores for {efo_id}: {pgs_data.get('count', 0)}")
for score in pgs_data.get("results", [])[:5]:
    print(f"  {score['id']}: {score['name']} (variants: {score.get('variants_number', '?')})")

Recipe 2: Regional Association Summary

Collect all GWAS associations in a genomic region.

import requests, time

BASE = "https://www.ebi.ac.uk/gwas/rest/api"

# Query region: chr9:21,900,000-22,200,000 (CDKN2A/2B locus)
data = requests.get(f"{BASE}/singleNucleotidePolymorphisms/search/findByChromBpLocationRange",
                    params={"chrom": "9", "bpStart": "21900000",
                            "bpEnd": "22200000", "size": 200}).json()
time.sleep(0.2)
snps = data["_embedded"]["singleNucleotidePolymorphisms"]
print(f"Variants in CDKN2A/2B locus: {len(snps)}")

# Get associations for each variant
region_assocs = []
for v in snps[:10]:  # Limit for demo
    rs = v["rsId"]
    try:
        a_data = requests.get(f"{BASE}/singleNucleotidePolymorphisms/{rs}/associations",
                              params={"size": 50}).json()
        time.sleep(0.2)
        for a in a_data["_embedded"]["associations"]:
            for t in a.get("efoTraits", []):
                region_assocs.append({"rsId": rs, "trait": t["trait"],
                                      "pvalue": a.get("pvalue")})
    except Exception:
        continue

print(f"Total associations in region: {len(region_assocs)}")

Recipe 3: Effect Size Forest Plot

Visualize effect sizes across studies for a single variant.

import re
import pandas as pd
import requests, time

BASE = "https://www.ebi.ac.uk/gwas/rest/api"
data = requests.get(f"{BASE}/singleNucleotidePolymorphisms/rs1801282/associations",
                    params={"size": 100}).json()
time.sleep(0.2)
assocs = data["_embedded"]["associations"]

def parse_ci(text):
    """Pull [low-high] confidence bounds out of the GWAS Catalog 'range' string."""
    nums = re.findall(r"[0-9]*\.?[0-9]+", text or "")
    return (float(nums[0]), float(nums[1])) if len(nums) >= 2 else (float("nan"), float("nan"))

# Build a forest-plot table: label, estimate (OR), ci_low, ci_high
rows = []
for a in assocs:
    or_val = a.get("orPerCopyNum")
    trait = a.get("efoTraits", [{}])[0].get("trait", "N/A") if a.get("efoTraits") else "N/A"
    if or_val and or_val > 0:
        lo, hi = parse_ci(a.get("range", ""))
        rows.append({"label": trait[:30], "estimate": or_val, "ci_low": lo, "ci_high": hi})

effects = pd.DataFrame(rows)
effects.to_csv("effects.csv", index=False)
print(f"Effect sizes: {len(effects)} entries -> effects.csv")
# Render with the omics-plotting SKILL (`skills/data-visualization/omics-plotting/SKILL.md`) "Forest" recipe (null line at OR=1) -> figures/forest_plot.png

Troubleshooting

Bundled Resources

references/api_endpoints.md

Covers: Complete endpoint catalog for all 6 REST API groups (studies, associations, variants, traits, genes, summary statistics) with query parameters, plus response field tables for the main entity types (association fields, study fields, variant fields), HTTP error codes, and pagination patterns.

Relocated inline: Core query patterns and the most common endpoints are demonstrated in Core API modules with full code examples. HAL+JSON structure and key identifier formats are in Key Concepts.

Omitted from original api_reference.md (794 lines): Advanced query composition patterns (multi-parameter chaining beyond what Core API shows), child/parent trait traversal details, detailed changelog/versioning notes. These are specialized and covered by EBI documentation.

Related Skills

• gget-genomic-databases -- gene lookups, variant annotation, BLAST (upstream ID resolution)
• ensembl-database (planned) -- variant effect prediction, regulatory annotations
• opentargets-database (planned) -- drug target validation from GWAS hits
• clinvar-database (planned) -- clinical variant interpretation
• bioservices-multi-database -- cross-database queries integrating GWAS with pathway data

References

• GWAS Catalog website -- main portal for browsing and searching
• GWAS Catalog REST API documentation -- official API reference
• GWAS Catalog GitHub -- source code and issue tracker
• Summary Statistics FTP -- bulk summary statistics downloads
• PGS Catalog -- polygenic score database using GWAS data