Skill: Search GEO

Use When

• A user wants to find expression datasets, RNA-seq series, microarray experiments, or epigenomics datasets in NCBI GEO.
• A user needs to look up GEO accessions: GSE (series), GSM (sample), GPL (platform), or GDS (curated dataset).
• A user wants to search GEO by organism, tissue, disease, treatment, assay type, or free-text keyword.
• A user wants to retrieve series metadata: number of samples, associated publication (PMID), submission date, platform.
• A user wants a list of GSM sample accessions for a given GSE series.
• A user wants to find GEO datasets suitable for a specific analysis (e.g., bulk RNA-seq of a specific cell type).

Inputs

• Required:
  • one of: free-text search query, GEO accession (GSE/GSM/GPL/GDS), organism name, or PubMed ID
• Optional:
  • GEO entity type: series (GSE), sample (GSM), platform (GPL), or dataset (GDS)
  • organism filter (e.g., "Homo sapiens"[Organism])
  • experiment type filter (e.g., "Expression profiling by high throughput sequencing"[DataSet Type])
  • date range: mindate / maxdate in YYYY/MM/DD
  • maximum results (retmax, default 20)
  • output format: json, xml, or soft
  • output directory

Workflow

1. Determine the NCBI database:
   • GSE, GSM, GPL → geo (use esearch -db geo)
   • GDS → gds (use esearch -db gds)
   • For most keyword searches, use gds which indexes curated datasets and series.
2. Build the query string with field tags when helpful:
   • organism: "Homo sapiens"[Organism]
   • data type: "Expression profiling by high throughput sequencing"[DataSet Type]
   • platform: GPL570[Platform]
   • publication: <PMID>[PubMed ID]
3. Run esearch -db gds -query "<term>" | esummary -format json to retrieve series metadata.
4. For a known GSE accession, fetch metadata via the GEO accession API:

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=<GSE>&targ=self&form=text&view=brief
   

5. To enumerate all samples in a series:
   • esearch -db gds -query "GSE12345[Accession]" | elink -target geosamples | esummary
   • Or fetch the series matrix from ftp://ftp.ncbi.nlm.nih.gov/geo/series/
6. Parse metadata to extract: accession, title, organism, type, sample count, platform, publication PMID, submission date, supplementary file URLs.
7. Report a summary table and, if requested, supplementary download links.

Output Contract

• Search query used (exact E-utilities command or API URL)
• Number of hits found
• Summary table with columns: Accession (GSE), Title, Organism, Type, Samples (#), Platform, PMID, Date
• List of GSM sample accessions for a queried series
• FTP/HTTPS links to raw data (supplementary files, series matrix) when available
• Saved output files when an output directory is provided:
  • esearch.json
  • esummary.json
  • samples.txt (one GSM per line, for series-level queries)

Limits

• This skill depends on live access to https://eutils.ncbi.nlm.nih.gov/entrez/eutils/ and https://www.ncbi.nlm.nih.gov/geo/.
• NCBI rate limit: 3 requests/second without API key, 10/second with API key.
• GEO gds database indexes curated GEO DataSets; not every GSE series becomes a GDS entry.
• Supplementary raw data may be large; this skill returns metadata and links, not the data itself.
• Controlled-access GEO datasets (e.g., human genotype data) require dbGaP authorization.
• The SOFT format is the canonical GEO text format; XML and JSON are available via E-utilities summary.
• Common failure cases:
  • querying the geo database instead of gds for series-level searches (use gds for keyword search)
  • confusing GSE (series) with GDS (curated dataset) — not all GSEs have a GDS entry
  • expecting retmode=json for efetch on GEO (only SOFT and XML are supported by efetch)
  • missing sample-level detail because only the series summary was fetched