geotcha 0.1.0

Extract and harmonize RNA-seq metadata from NCBI GEO

GEOtcha

Extract and harmonize RNA-seq metadata from NCBI GEO.

GEOtcha is a CLI tool that helps researchers search GEO by disease keyword, filter to human RNA-seq datasets, extract structured metadata at both series (GSE) and sample (GSM) levels, and harmonize the results into standardized output files.

Installation

pip install geotcha

For development:

git clone https://github.com/shantanubafna/GEOtcha.git
cd geotcha
pip install -e ".[dev]"

Quick Start

Search for datasets

geotcha search "inflammatory bowel disease"

Full pipeline with subset testing

geotcha run "IBD" --subset 5 --output ./results/ --harmonize

Extract from specific GSE IDs

geotcha extract GSE12345 GSE67890 --output ./results/

With LLM harmonization

pip install geotcha[llm]
geotcha run "IBD" --harmonize --llm --llm-provider anthropic

Configuration

# Set your NCBI API key (recommended for higher rate limits)
geotcha config set ncbi_api_key "YOUR_KEY"

# Set your email for NCBI Entrez
geotcha config set ncbi_email "[e-mail]"

# View current configuration
geotcha config show

Configuration priority: CLI flags > environment variables (GEOTCHA_*) > config file (~/.config/geotcha/config.toml) > defaults.

Output

GEOtcha produces:

• gse_summary.csv — One row per GSE with series-level metadata
• gsm/<GSE_ID>_samples.csv — Per-GSE file with sample-level metadata
• With --harmonize: additional _harmonized columns with standardized values

Fields extracted

Level	Fields
GSE	ID, URL, title, organism, experiment type, platform, sample counts, PubMed links, tissue, disease, treatment, timepoint, gender, age, responder info
GSM	ID, title, source, organism, platform, instrument, library strategy, tissue, cell type, disease, gender, age, treatment, timepoint, responder status

Interactive Flow

$ geotcha run "IBD"
Searching GEO for: IBD, ulcerative colitis, Crohn's disease...
Found 347 datasets. After filtering (Homo sapiens + RNA-seq): 182 datasets.

Run on a subset first? [Y/n]: Y
Subset size [5]: 5

Processing 5/182 datasets...
 [████████████████] 5/5 complete

Results: ./output/gse_summary.csv (5 rows), ./output/gsm/ (5 files)

Proceed with remaining 177 datasets? [Y/n]:

Disease Expansion

GEOtcha automatically expands disease keywords to capture related terms:

• IBD → inflammatory bowel disease, ulcerative colitis, Crohn's disease (abbreviations like UC, CD are used for relevance filtering)
• SLE → systemic lupus erythematosus, lupus
• RA → rheumatoid arthritis

Harmonization

Rule-based normalization for:

• Gender: male/M/man → "male"
• Age: "45 years", "45yo" → "45"
• Tissue: mapped to UBERON ontology terms
• Disease: mapped to Disease Ontology terms
• Timepoint: "week 8", "W8" → "W8"

Optional LLM-assisted harmonization (--llm) for ambiguous free-text values.

License

MIT