Search and Cross-Reference NCBI GEO with ENCODE

When to Use

• User wants to find complementary datasets in NCBI GEO to supplement ENCODE data
• User asks about "GEO", "Gene Expression Omnibus", "supplementary data", or "find related datasets"
• User needs to cross-reference ENCODE experiments with GEO series for additional replicates or conditions
• User wants to link ENCODE accessions to GEO/SRA identifiers for data sharing or citation
• Example queries: "find GEO datasets for pancreatic islet RNA-seq", "link this ENCODE experiment to GEO", "search GEO for complementary ATAC-seq data"

Query the Gene Expression Omnibus programmatically to find complementary datasets, cross-reference ENCODE experiments, and download metadata.

Scientific Rationale

The question: "What additional expression or epigenomic datasets exist in GEO that complement my ENCODE analysis?"

GEO hosts >200,000 series across all organisms and assay types. Many ENCODE experiments are deposited in GEO as secondary archives (ENCODE Portal is primary). GEO also contains vast amounts of non-ENCODE data — disease cohorts, perturbation experiments, time courses — that complement ENCODE's reference epigenomes.

GEO ↔ ENCODE Relationship

• ENCODE processed data is deposited at GEO as standard GSE submissions
• Raw sequencing data goes to SRA (linked from both GEO and ENCODE)
• The ENCODE Portal is canonical; GEO is secondary archive
• GEO accessions are stored in ENCODE's dbxrefs field as GEO:GSExxxxx
• NCBI maintains a dedicated ENCODE listing: https://www.ncbi.nlm.nih.gov/geo/encode/

GEO Entity Hierarchy

Series (GSE) — An experiment/study
  ├── Sample (GSM) — Individual measurements
  │     ├── references → Platform (GPL)
  │     ├── has → Supplementary files (raw data)
  │     └── has → Data table (normalized values)
  │
  └── curated into → DataSet (GDS) [not all GSE get curated]
        └── generates → Profiles (gene-level summaries)

Step 1: Find GEO Accessions for ENCODE Experiments

From ENCODE → GEO

ENCODE experiments may have GEO cross-references in their metadata. After tracking an experiment:

encode_track_experiment(accession="ENCSR...")

Check the experiment's dbxrefs field for GEO:GSExxxxx entries. If found, link it:

encode_link_reference(
    experiment_accession="ENCSR...",
    reference_type="geo_accession",
    reference_id="GSE12345"
)

From GEO → ENCODE

Search GEO for ENCODE-deposited data:

# Via NCBI E-utilities
curl "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=gds&term=ENCODE[KEYWORD]+AND+gse[ETYP]&retmax=100&usehistory=y&tool=encode_mcp&email=YOUR_EMAIL"

Step 2: Search GEO for Complementary Datasets

E-utilities Search Syntax

Base URL: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi

Required parameters: db=gds, term=QUERY, tool=encode_mcp, email=YOUR_EMAIL

Rate limit: 3 req/sec without API key, 10 req/sec with key. Get a key at https://www.ncbi.nlm.nih.gov/account/

Search Field Qualifiers

Example Searches

# Human pancreas ATAC-seq datasets with BED files
curl "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=gds&term=pancreas+AND+ATAC-seq+AND+%22Homo+sapiens%22[ORGN]+AND+gse[ETYP]+AND+bed[suppFile]&retmax=50&tool=encode_mcp&email=YOUR_EMAIL"

# ChIP-seq datasets from a specific year
curl "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=gds&term=ChIP-seq+AND+H3K27ac+AND+gse[ETYP]+AND+2024[PDAT]&retmax=50&tool=encode_mcp&email=YOUR_EMAIL"

# Datasets associated with a PubMed ID
curl "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&db=gds&id=PMID&tool=encode_mcp&email=YOUR_EMAIL"

Step 3: Retrieve GEO Metadata

Get Summary for GEO Records

# Step 1: Search (returns UIDs, NOT accessions)
curl "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=gds&term=GSE12345[ACCN]&tool=encode_mcp&email=YOUR_EMAIL"

# Step 2: Get summary (use UID from step 1)
curl "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=gds&id=UID&version=2.0&tool=encode_mcp&email=YOUR_EMAIL"

Direct Record Access (acc.cgi)

# Get full SOFT-format record
curl "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE12345&targ=self&view=full&form=text"

# Get XML (MINiML) format
curl "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE12345&targ=self&view=full&form=xml"

# Get all sample metadata for a series
curl "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE12345&targ=gsm&view=brief&form=text"

Step 4: Download GEO Data Files

FTP Directory Convention

GEO uses a "nnn" directory pattern: replace last 3 digits with "nnn".

Key Download Paths

Download Commands

# Download series matrix (fastest for expression data)
wget "https://ftp.ncbi.nlm.nih.gov/geo/series/GSE12nnn/GSE12345/matrix/GSE12345_series_matrix.txt.gz"

# Download all supplementary files
wget "https://ftp.ncbi.nlm.nih.gov/geo/series/GSE12nnn/GSE12345/suppl/GSE12345_RAW.tar"

Format Selection Guide

Step 5: Cross-Reference Workflow

ENCODE + GEO Integration Pattern

1. Find ENCODE experiments of interest:
   encode_search_experiments(assay_title="total RNA-seq", organ="pancreas")

2. For each experiment, check for GEO accession:
   encode_get_experiment(accession="ENCSR...")
   → Look in dbxrefs for "GEO:GSExxxxx"

3. If GEO accession found, link it:
   encode_link_reference(
       experiment_accession="ENCSR...",
       reference_type="geo_accession",
       reference_id="GSE12345"
   )

4. Search GEO for complementary non-ENCODE datasets:
   E-utils search for same tissue + different assay or condition

5. Download GEO metadata for comparison:
   acc.cgi or E-utils esummary

6. Log the cross-reference:
   encode_log_derived_file(
       file_path="/path/to/comparison.tsv",
       source_accessions=["ENCSR...", "GSE12345"],
       description="ENCODE-GEO cross-tissue comparison"
   )

Finding SRA Accessions from GEO

For sequencing data, raw reads are in SRA, not GEO:

# Link GEO to SRA
curl "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=gds&db=sra&id=GDS_UID&tool=encode_mcp&email=YOUR_EMAIL"

Python alternative using pysradb:

from pysradb.search import SraSearch
# Convert GSE to SRP
pysradb gse-to-srp GSE12345
# Get all SRR run accessions
pysradb gsm-to-srr GSM12345

Pitfalls and Caveats

1. E-utils return UIDs, not accessions: GEO search returns numeric UIDs. You must call ESummary to get the actual GSE/GDS accession numbers.
2. Not all ENCODE experiments have GEO accessions: The dbxrefs field may be empty. ENCOD