GEO Database

Overview

The Gene Expression Omnibus (GEO) is NCBI's public repository for high-throughput gene expression and functional genomics data. It holds hundreds of thousands of studies (Series) and millions of samples from both array-based and sequence-based experiments. For current totals see the GEO browser at https://www.ncbi.nlm.nih.gov/geo/.

When to Use This Skill

Use this skill when searching for gene expression datasets, retrieving experimental data, downloading raw and processed files, querying expression profiles, or integrating GEO data into computational analysis workflows.

Core Workflow

1. Search for relevant studies (E-utilities / Bio.Entrez) → see references/searching.md.
2. Retrieve a series with GEOparse: GEOparse.get_GEO(geo="GSE...", destdir="./data") → see references/geoparse_usage.md.
3. Extract the expression matrix via gse.pivot_samples('VALUE') (genes × samples).
4. Analyze — QC, differential expression, clustering, meta-analysis → see references/expression_analysis.md.

For bulk downloads, skip GEOparse and pull files directly over FTP → see references/data_access.md.

GEO Data Organization

GEO organizes data hierarchically using different accession types:

• Series (GSE) — a complete experiment with related samples (e.g. GSE123456). Largest organizational unit. Holds experimental design, samples, study info.
• Sample (GSM) — a single experimental sample / biological replicate (e.g. GSM987654). Linked to platforms and series.
• Platform (GPL) — the microarray or sequencing platform (e.g. GPL570, Affymetrix Human Genome U133 Plus 2.0 Array). Shared across experiments.
• DataSet (GDS) — curated, consistently-formatted collections (e.g. GDS5678) processed for differential analysis. Ideal for quick comparative analyses, but the curated GDS set is legacy and no longer growing — most recent studies exist only as GSE Series, so prefer GSE for new work.
• Profiles — gene-specific expression data linked to sequence features, queryable by gene name, cross-referenced to Entrez Gene.

Access Methods (routing)

• GEOparse (recommended) — easiest series-level access; download, parse metadata, extract expression matrices, supplementary files, filter samples. → references/geoparse_usage.md
• E-utilities (Bio.Entrez) — metadata searching and batch queries across gds, geoprofiles. Always set Entrez.email. → references/searching.md
• Direct FTP / wget / curl — bulk downloads of series matrix, SOFT, MINiML, and supplementary files; no rate limits. → references/data_access.md
• GEO2R web tool — no-code differential expression analysis in the browser at https://www.ncbi.nlm.nih.gov/geo/geo2r/?acc=GSExxxxx; generates reproducible R scripts. Useful for exploratory analysis before downloading.

Installation and Setup

uv pip install GEOparse        # primary GEO access (recommended)
uv pip install biopython       # E-utilities / programmatic NCBI access
uv pip install pandas numpy scipy statsmodels scikit-learn  # analysis
uv pip install matplotlib seaborn                           # visualization

Configure NCBI E-utilities access (email is required by NCBI; an API key raises the rate limit from 3 to 10 requests/second):

from Bio import Entrez

Entrez.email = "your.email@example.com"   # required
# Optional API key — https://www.ncbi.nlm.nih.gov/account/
Entrez.api_key = "your_api_key_here"

Key Concepts

• SOFT (Simple Omnibus Format in Text): GEO's primary text format with metadata and data tables. Easily parsed by GEOparse.
• MINiML (MIAME Notation in Markup Language): XML format for programmatic access and data exchange.
• Series Matrix: tab-delimited expression matrix (samples as columns, genes/probes as rows). Fastest format for getting expression data.
• MIAME Compliance: minimum standardized annotation GEO enforces on submissions.
• Expression Value Types: raw signal, normalized, or log-transformed — always check platform and processing methods.
• Platform Annotation: maps probe/feature IDs to genes; essential for biological interpretation.

Common Use Cases

• Transcriptomics research — download expression data, compare profiles across studies, identify DEGs, run meta-analyses.
• Drug response studies — analyze post-treatment expression changes, identify response biomarkers, build sensitivity models.
• Disease biology — disease vs. normal expression, disease signatures, patient subgroup/stage comparisons, expression–outcome correlation.
• Biomarker discovery — screen diagnostic/prognostic markers, validate across cohorts, integrate with clinical data.

Rate Limiting and Best Practices

• E-utilities limits: 3 req/s without an API key, 10 req/s with one. Insert delays: time.sleep(0.34) (no key) or time.sleep(0.1) (with key).
• FTP: no rate limits — preferred for bulk downloads (wget -r for whole directories).
• GEOparse caching: files cached in destdir; subsequent calls reuse them.
• Method selection: GEOparse for series-level access, E-utilities for search / batch metadata, FTP for direct/bulk file pulls. Cache locally; always set Entrez.email with Biopython.

Important Notes

• Data quality: GEO accepts user-submitted data of varying quality. Check platform annotation and processing methods, verify metadata and design, watch for batch effects, and consider reprocessing raw data.
• File sizes: series matrix files can exceed 1 GB; supplementary files (e.g. CEL) can be very large. Plan disk space; download incrementally.
• Citation: GEO data is free for research. Cite original studies plus the GEO database (Barrett et al. 2013, Nucleic Acids Research). Check per-dataset usage restrictions and follow NCBI guidelines.
• Common pitfalls: platforms use different probe IDs (need annotation mapping); values may be raw/normalized/log-transformed (check metadata); sample metadata is inconsistently formatted; older submissions may lack series matrix files; platform annotations may be outdated.

Reference Index

• references/searching.md — E-utilities search (datasets, profiles, advanced queries, search→summary→fetch workflow, batch metadata).
• references/geoparse_usage.md — GEOparse install/usage, expression matrices, supplementary files, sample filtering, caching.
• references/data_access.md — direct FTP access (ftplib, wget, curl) and the accession-to-path rule.
• references/expression_analysis.md — QC/preprocessing, differential expression, correlation/clustering, batch processing, cross-study meta-analysis.
• references/geo_reference.md — in-depth technical reference: E-utilities API endpoints, SOFT/MINiML format docs, FTP directory structure, normalization pipelines, platform quirks, and error-handling troubleshooting.

Additional Resources

• GEO Website: https://www.ncbi.nlm.nih.gov/geo/
• GEO Submission Guidelines: https://www.ncbi.nlm.nih.gov/geo/info/submission.html
• GEOparse Documentation: https://geoparse.readthedocs.io/
• E-utilities Documentation: https://www.ncbi.nlm.nih.gov/books/NBK25501/
• GEO FTP Site: ftp://ftp.ncbi.nlm.nih.gov/geo/
• GEO2R Tool: https://www.ncbi.nlm.nih.gov/geo/geo2r/
• NCBI API Keys: https://ncbiinsights.ncbi.nlm.nih.gov/2017/11/02/new-api-keys-for-the-e-utilities/
• Biopython Tutorial: https://biopython.org/DIST/docs/tutorial/Tutorial.html

Scripts

scripts/query_geo.py — runnable helper for GEO DataSets via NCBI E-utilities (no key; NCBI_API_KEY lifts the rate limit). Has a PEP 723 inline-dependency header, so uv run installs requests automatically:

uv run scripts/query_geo.py search "breast cancer AND Homo sapiens[ORGN]" --retmax 5
uv run scripts/query_geo.py summary 200000001,200000002