Imported from bg-szy/TOP-SKILLS (skills/awesome-skills/clinvar-database/SKILL.md). Install upstream with npx skills add bg-szy/TOP-SKILLS --skill clinvar-database. Copyright stays with the author (CC0-1.0).

ClinVar Clinical Variants Database

Overview

ClinVar is NCBI's public archive of interpretations of variants submitted by clinical laboratories, researchers, and expert panels. It contains 2M+ variants with clinical significance classifications (Pathogenic, Likely Pathogenic, VUS, Likely Benign, Benign) for over 6,000 conditions. Access is free and requires no authentication via NCBI E-utilities.

When to Use

• Checking whether a specific variant (rsID, HGVS, or genomic position) has a clinical significance classification
• Retrieving all pathogenic/likely-pathogenic variants in a gene of interest
• Identifying conflicting interpretations between submitting laboratories
• Pulling condition/phenotype associations for a variant (MIM, MeSH, HPO terms)
• Building variant filtering pipelines that prioritize clinically actionable variants
• For somatic cancer variants, also check cosmic-database; for GWAS associations use gwas-database

Prerequisites

• Python packages: requests, xml.etree.ElementTree (stdlib)
• Data requirements: gene symbols, rsIDs, HGVS strings, or ClinVar Variation IDs
• Environment: internet connection; NCBI Entrez email required (set email parameter)
• Rate limits: 3 requests/second unauthenticated; 10/second with API key (free at https://www.ncbi.nlm.nih.gov/account/)

pip install requests
# No additional packages required; xml.etree is part of Python stdlib

Quick Start

import requests

EMAIL = "your@email.com"  # required by NCBI policy

def clinvar_search(query, retmax=10):
    """Search ClinVar and return a list of ClinVar Variation IDs."""
    r = requests.get(
        "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi",
        params={"db": "clinvar", "term": query, "retmax": retmax,
                "retmode": "json", "email": EMAIL}
    )
    r.raise_for_status()
    return r.json()["esearchresult"]["idlist"]

# Find pathogenic BRCA1 variants
ids = clinvar_search("BRCA1[gene] AND pathogenic[clinsig]", retmax=5)
print(f"Found variation IDs: {ids}")

Core API

Query 1: Search Variants by Gene and Clinical Significance

Use ESearch to find ClinVar Variation IDs matching a structured query.

import requests

EMAIL = "your@email.com"
BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils"

def esearch(query, retmax=200):
    r = requests.get(f"{BASE}/esearch.fcgi",
                     params={"db": "clinvar", "term": query,
                             "retmax": retmax, "retmode": "json", "email": EMAIL})
    r.raise_for_status()
    result = r.json()["esearchresult"]
    return result["idlist"], int(result["count"])

# Gene-specific pathogenic variants
ids, total = esearch("BRCA2[gene] AND (pathogenic[clinsig] OR likely pathogenic[clinsig])")
print(f"Pathogenic/LP BRCA2 variants: {total} total, retrieved {len(ids)}")
print(f"First 5 IDs: {ids[:5]}")

# By rsID
ids, _ = esearch("rs80357906[rs]")
print(f"Variant IDs for rs80357906: {ids}")

# By condition name
ids, total = esearch("breast cancer[dis] AND pathogenic[clinsig]")
print(f"Pathogenic variants for breast cancer: {total}")

Query 2: Fetch Variant Summary Records

Retrieve structured summary data (JSON) for a list of Variation IDs.

import requests, json

EMAIL = "your@email.com"
BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils"

def esummary(ids):
    """Fetch ESummary records for a list of ClinVar variation IDs."""
    r = requests.post(f"{BASE}/esummary.fcgi",
                      data={"db": "clinvar", "id": ",".join(ids),
                            "retmode": "json", "email": EMAIL})
    r.raise_for_status()
    return r.json()["result"]

ids, _ = esearch_func = lambda q: requests.get(
    f"{BASE}/esearch.fcgi",
    params={"db": "clinvar", "term": q, "retmax": 5, "retmode": "json", "email": EMAIL}
).json()["esearchresult"]["idlist"]

# Manual example with known IDs
sample_ids = ["12375", "17684", "54270"]
result = esummary(sample_ids)

for vid in result.get("uids", []):
    rec = result[vid]
    # ClinVar 2024 schema: clinical_significance was replaced by germline_classification
    # (also: clinical_impact_classification, oncogenicity_classification — same shape, often empty)
    gc = rec.get("germline_classification", {})
    print(f"\nVariation {vid}: {rec.get('title')}")
    print(f"  ClinSig  : {gc.get('description')}")
    print(f"  Review   : {gc.get('review_status')}")
    print(f"  Gene     : {rec.get('genes', [{}])[0].get('symbol')}")

Query 3: Fetch Full XML Records

Retrieve the complete variant record in XML for detailed submitter and condition data.

import requests
import xml.etree.ElementTree as ET

EMAIL = "your@email.com"
BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils"

def efetch_xml(variation_ids):
    # ClinVar 2024 XML overhaul: "clinvarset" rettype returns an empty stub.
    # Use rettype="vcv" + is_variationid="true" to get the new <VariationArchive> records.
    r = requests.post(f"{BASE}/efetch.fcgi",
                      data={"db": "clinvar", "id": ",".join(variation_ids),
                            "rettype": "vcv", "is_variationid": "true",
                            "retmode": "xml", "email": EMAIL})
    r.raise_for_status()
    return ET.fromstring(r.text)

root = efetch_xml(["17677"])  # BRCA1 c.5266dupC (rs80357906)

# Aggregate (germline) classification — one per VariationArchive
for va in root.iter("VariationArchive"):
    name = va.get("VariationName")
    gc = va.find("./ClassifiedRecord/Classifications/GermlineClassification")
    desc = gc.find("Description") if gc is not None else None
    rstat = gc.find("ReviewStatus") if gc is not None else None
    print(f"{name}: {desc.text if desc is not None else 'n/a'} "
          f"({rstat.text if rstat is not None else 'n/a'})")

    # Per-submitter assertions
    for ca in va.iter("ClinicalAssertion"):
        acc = ca.find("ClinVarAccession")
        cls = ca.find("Classification/GermlineClassification")
        if acc is not None and cls is not None:
            print(f"  {acc.get('SubmitterName', '?')}: {cls.text}")

Query 4: ClinVar FTP Bulk Data

For large-scale queries, download and parse the full variant summary file.

import urllib.request
import gzip, csv, io

# Full summary (tab-separated, ~300 MB compressed)
URL = "https://ftp.ncbi.nlm.nih.gov/pub/clinvar/tab_delimited/variant_summary.txt.gz"

# Stream and parse without full download
with urllib.request.urlopen(URL) as resp:
    with gzip.open(resp, "rt", encoding="utf-8") as f:
        reader = csv.DictReader(f, delimiter="\t")
        pathogenic_brca1 = []
        for row in reader:
            if row["GeneSymbol"] == "BRCA1" and "Pathogenic" in row["ClinicalSignificance"]:
                pathogenic_brca1.append({
                    "name": row["Name"],
                    "clinsig": row["ClinicalSignificance"],
                    "condition": row["PhenotypeList"],
                    "rsid": row["RS# (dbSNP)"],
                })
        print(f"Pathogenic BRCA1 variants: {len(pathogenic_brca1)}")
        for v in pathogenic_brca1[:3]:
            print(f"  {v['name']} | {v['clinsig']} | rs{v['rsid']}")

Query 5: Review Status and Conflicting Interpretations

Filter variants by review status (evidence quality) and find conflicts.

import requests

EMAIL = "your@email.com"
BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils"

# Stars correspond to review levels:
# 0 = no assertion criteria, 1 = criteria provided (single),
# 2 = criteria provided (multiple), 3 = expert panel, 4 = practice guideline

def search_by_review_stars(gene, min_stars=2):
    """Search for variants with at least min_stars review status."""
    star_terms = {1: "criteria provided, single submitter",
                  2: "criteria provided, multiple submitters, no conflicts",
                  3: "reviewed by expert panel",
                  4: "practice guideline"}
    terms = [f'"{star_terms[s]}"[review status]' for s in range(min_stars, 5) if s in star_terms]
    query = f"{gene}[gene] AND (" + " OR ".join(terms) + ")"
    r = requests.get(f"{BASE}/esearch.fcgi",
                     params={"db": "clinvar", "term": query, "retmax": 100,
                             "retmode": "json", "email": EMAIL})
    return r.json()["esearchresult"]

result = search_by_review_stars("BRCA1", min_stars=3)
print(f"Expert-reviewed BRCA1 variants: {result['count']}")

Query 6: Variant-to-Condition Mapping

Extract condition (phenotype) data from ClinVar records.

import requests, json

EMAIL = "your@email.com"
BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils"

def get_conditions(variation_ids):
    """Return condition data for a list of ClinVar variation IDs."""
    r = requests.post(f"{BASE}/esummary.fcgi",
                      data={"db": "clinvar", "id": ",".join(variation_ids),
                            "retmode": "json", "email": EMAIL})
    r.raise_for_status()
    result = r.json()["result"]
    conditions = {}
    for vid in result.get("uids", []):
        rec = result[vid]
        # trait_set moved under germline_classification in the 2024 ClinVar JSON
        trait_set = rec.get("germline_classification", {}).get("trait_set", [])
        conditions[vid] = [t.get("trait_name") for t in trait_set]
    return conditions

sample_ids = ["12375", "17684", "54270"]
cond_map = get_conditions(sample_ids)
for vid, conds in cond_map.items():
    print(f"Variation {vid}: {', '.join(conds)}")

Key Concepts

ClinVar Variation ID vs. rsID

ClinVar assigns its own stable Variation ID (integer) to each interpreted variant record. This differs from dbSNP rsIDs. A single rsID can correspond to multiple ClinVar Variation IDs if different alleles or interpretations are submitted separately.

Review Stars and Evidence Quality

ClinVar's "review status" encodes the level of evidence:

• 0 stars: No assertion criteria provided
• 1 star: Criteria provided, single submitter
• 2 stars: Multiple submitters, no conflict
• 3 stars: Reviewed by expert panel (e.g., ENIGMA, ClinGen)
• 4 stars: Practice guideline

Common Workflows

Workflow 1: Gene Pathogenicity Report

Goal: Retrieve all high-confidence pathogenic variants in a gene and export to CSV.

import requests, json, time, pandas as pd

EMAIL = "your@email.com"
BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils"

def search_gene_pathogenic(gene, clinsig="pathogenic"):
    query = f"{gene}[gene] AND {clinsig}[clinsig]"
    r = requests.get(f"{BASE}/esearch.fcgi",
                     params={"db": "clinvar", "term": query, "retmax": 500,
                             "retmode": "json", "email": EMAIL})
    return r.json()["esearchresult"]["idlist"]

def fetch_summaries(ids):
    records = []
    for i in range(0, len(ids), 100):
        batch = ids[i:i+100]
        r = requests.post(f"{BASE}/esummary.fcgi",
                          data={"db": "clinvar", "id": ",".join(batch),
                                "retmode": "json", "email": EMAIL})
        result = r.json()["result"]
        for vid in result.get("uids", []):
            rec = result[vid]
            # ClinVar 2024 schema: clinical_significance → germline_classification; trait_set nested inside it
            gc = rec.get("germline_classification", {})
            records.append({
                "variation_id": vid,
                "name": rec.get("title"),
                "clinsig": gc.get("description"),
                "review_status": gc.get("review_status"),
                "gene": ",".join(g.get("symbol", "") for g in rec.get("genes", [])),
                "conditions": "; ".join(t.get("trait_name", "") for t in gc.get("trait_set", [])),
            })
        time.sleep(0.15)
    return records

gene = "BRCA1"
ids = search_gene_pathogenic(gene)
print(f"Found {len(ids)} pathogenic variants in {gene}")

records = fetch_summaries(ids)
df = pd.DataFrame(records)
df.to_csv(f"{gene}_pathogenic_variants.csv", index=False)
print(f"Saved {len(df)} records → {gene}_pathogenic_variants.csv")
print(df[["name", "clinsig", "review_status"]].head())

Workflow 2: Variant Classification Check

Goal: Check ClinVar status for a list of user-provided rsIDs or HGVS notations.

import requests, time, pandas as pd

EMAIL = "your@email.com"
BASE = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils"

variants = ["rs80357906", "rs80357220", "rs28897672"]
results = []

for rsid in variants:
    r = requests.get(f"{BASE}/esearch.fcgi",
                     params={"db": "clinvar", "term": f"{rsid}[rs]",
                             "retmax": 5, "retmode": "json", "email": EMAIL})
    ids = r.json()["esearchresult"]["idlist"]
    if not ids:
        results.append({"rsid": rsid, "variation_id": None, "clinsig": "Not in ClinVar"})
        continue

    r2 = requests.post(f"{BASE}/esummary.fcgi",
                       data={"db": "clinvar", "id": ",".join(ids[:1]),
                             "retmode": "json", "email": EMAIL})
    rec = r2.json()["result"][ids[0]]
    gc = rec.get("germline_classification", {})  # 2024 ClinVar JSON
    results.append({
        "rsid": rsid,
        "variation_id": ids[0],
        "clinsig": gc.get("description", "Unknown"),
        "review_status": gc.get("review_status"),
    })
    time.sleep(0.15)

df = pd.DataFrame(results)
print(df.to_string(index=False))

Key Parameters

Best Practices

1. Always set email: NCBI requires an email in all E-utility calls for rate-limit attribution and policy compliance.

2. Use FTP bulk download for large queries: For more than ~1000 variants, download variant_summary.txt.gz from the ClinVar FTP rather than looping over EFetch — it's faster and avoids rate limits.

3. Filter by review status: Automated pipelines should filter to ≥2-star variants to reduce noise from single-submitter assertions without peer review.

4. Use API key for production: Register at https://www.ncbi.nlm.nih.gov/account/ to get a free API key (api_key parameter) and triple your rate limit (3 → 10 req/s).

5. Handle VUS separately: "Conflicting interpretations of pathogenicity" is its own ClinSig category — don't combine it with "VUS" in filters; they have different implications for clinical decision-making.

Common Recipes

Recipe: Check if rsID Is in ClinVar

When to use: Quick lookup for a single known variant.

import requests

EMAIL = "your@email.com"
rsid = "rs80357906"

r = requests.get(
    "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi",
    params={"db": "clinvar", "term": f"{rsid}[rs]",
            "retmax": 1, "retmode": "json", "email": EMAIL}
)
count = int(r.json()["esearchresult"]["count"])
print(f"{rsid}: {'found' if count else 'NOT'} in ClinVar ({count} records)")

Recipe: Download Variant Summary TSV

When to use: Bulk analysis — load entire ClinVar into a pandas DataFrame.

import pandas as pd

url = "https://ftp.ncbi.nlm.nih.gov/pub/clinvar/tab_delimited/variant_summary.txt.gz"
# Only human GRCh38 pathogenic variants
df = pd.read_csv(url, sep="\t", compression="gzip",
                 usecols=["#AlleleID", "Name", "GeneSymbol", "ClinicalSignificance",
                          "ReviewStatus", "PhenotypeList", "Assembly", "RS# (dbSNP)"])
df = df[(df["Assembly"] == "GRCh38") & (df["ClinicalSignificance"].str.contains("Pathogenic", na=False))]
print(f"Pathogenic variants (GRCh38): {len(df)}")
df.to_csv("clinvar_pathogenic_grch38.csv", index=False)

Recipe: Search by OMIM Disease ID

When to use: Find all ClinVar variants associated with a specific OMIM condition.

import requests

EMAIL = "your@email.com"
omim_id = "604370"  # BRCA1-associated breast-ovarian cancer

r = requests.get(
    "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi",
    params={"db": "clinvar", "term": f"{omim_id}[MIM]",
            "retmax": 20, "retmode": "json", "email": EMAIL}
)
result = r.json()["esearchresult"]
print(f"Variants for OMIM {omim_id}: {result['count']} total")
print(f"First IDs: {result['idlist'][:5]}")

Troubleshooting

Related Skills

• gwas-database — GWAS Catalog for population-level SNP-trait associations (complement to ClinVar's clinical assertions)
• ensembl-database — Ensembl VEP for predicting variant consequences without requiring prior clinical curation
• cosmic-database — Somatic cancer variant database (complementary to ClinVar's germline focus)
• pubmed-database — Retrieve supporting publications cited in ClinVar submissions

References

• ClinVar official site — Browse and download ClinVar data
• NCBI E-utilities documentation — Full E-utilities API reference
• ClinVar FTP downloads — Bulk data files (variant_summary.txt.gz, etc.)
• ClinVar data model — Understanding review status stars and ClinSig categories