jaechang-hits/gget-genomic-databases
Unified CLI/Python interface to 20+ genomic databases. Gene lookups (Ensembl search/info/seq), BLAST/BLAT, AlphaFold, Enrichr enrichment, OpenTargets disease/drug, CELLxGENE single-cell, cBioPortal/COSMIC cancer, ARCHS4 expression. Spans genomics, proteomics, disease. For batch/advanced BLAST use biopython; for multi-DB Python SDK use bioservices.
This is a copy. The original lives at biotender-max/gget-genomic-databases.
npx skills add https://github.com/jaechang-hits/SciAgent-Skills --skill gget-genomic-databases
gget is a command-line and Python package providing unified access to 20+ genomic databases and analysis methods. Query gene information, sequences, protein structures, expression data, and disease associations through a consistent interface. All modules work as both CLI tools and Python functions, returning DataFrames (Python) or JSON/CSV (CLI).
biopython insteadbioservices insteadggetgget setup <module> before first use (alphafold, cellxgene, elm, gpt)time.sleep(). Databases update biweekly; keep gget updated. Max ~1000 Ensembl IDs per gget.info() callpip install gget
# Optional: setup modules that need additional dependencies
gget setup alphafold # ~4GB model parameters, requires OpenMM
gget setup cellxgene # cellxgene-census package
gget setup elm # local ELM database
import gget
# Search for genes by keyword
results = gget.search(["BRCA1", "tumor suppressor"], species="homo_sapiens")
print(f"Found {len(results)} genes")
# Get detailed gene information (Ensembl + UniProt + NCBI)
info = gget.info(["ENSG00000012048"])
print(f"Gene: {info.iloc[0]['primary_gene_name']}")
# Enrichment analysis on a gene list
enrichment = gget.enrichr(["ACE2", "AGT", "AGTR1"], database="ontology")
print(f"Enriched terms: {len(enrichment)}")
Query Ensembl for gene references, search by keywords, retrieve gene metadata, and fetch sequences.
import gget
# Search for genes by keyword
results = gget.search(["BRCA1", "tumor suppressor"], species="homo_sapiens")
print(f"Found {len(results)} genes")
print(results[["ensembl_id", "gene_name", "biotype"]].head())
# Get detailed gene information (Ensembl + UniProt + NCBI)
info = gget.info(["ENSG00000012048", "ENSG00000139618"])
print(f"Gene info columns: {list(info.columns)}")
import gget
# Retrieve sequences
nucleotide_seqs = gget.seq(["ENSG00000012048"])
protein_seqs = gget.seq(["ENSG00000012048"], translate=True, isoforms=True)
print(f"Retrieved {len(protein_seqs)} isoform sequences")
# Download reference genome files (specify release for reproducibility)
ref_links = gget.ref("homo_sapiens", which="gtf", release=112)
print(f"GTF download link: {ref_links}")
BLAST/BLAT remote searches, multiple sequence alignment, and fast local alignment.
import gget
import time
# BLAST against SwissProt (remote API — add delay for batch queries)
blast_results = gget.blast(
"MKWMFKEDHSLEHRCVESAKIRAKYPDRVPVIVEKVSGSQIVDIDKRKYLVPSDITVAQFMWIIRKRIQLPSEKAIFLFVDKTVPQSR",
database="swissprot", limit=10
)
print(f"Top hit: {blast_results.iloc[0]['Description']}, E-value: {blast_results.iloc[0]['e-value']}")
time.sleep(2) # Rate-limit between BLAST queries
# BLAT — find genomic position (UCSC)
blat_results = gget.blat("ATCGATCGATCGATCGATCG", assembly="human")
print(f"Genomic location: chr{blat_results.iloc[0]['chromosome']}:{blat_results.iloc[0]['start']}")
import gget
# Multiple sequence alignment with Muscle5
aligned = gget.muscle("sequences.fasta", save=True)
# Fast local alignment with DIAMOND (local, no rate limit needed)
diamond_results = gget.diamond(
"GGETISAWESQME",
reference="reference.fasta",
sensitivity="very-sensitive",
threads=4
)
print(f"Alignments found: {len(diamond_results)}")
Download PDB structures, predict structures with AlphaFold2, find linear motifs.
import gget
# Download PDB structure
pdb_data = gget.pdb("7S7U", save=True)
# Predict structure with AlphaFold2 (requires gget setup alphafold)
structure = gget.alphafold(
"MKWMFKEDHSLEHRCVESAKIRAKYPDRVPVIVEKVSGSQIVDIDKRKYLVPSDITVAQFMWIIRKRIQLPSEKAIFLFVDKTVPQSR",
plot=True, show_sidechains=True
)
print("Structure prediction complete, PDB file saved")
import gget
# Find Eukaryotic Linear Motifs (requires gget setup elm)
ortholog_df, regex_df = gget.elm("LIAQSIGQASFV")
print(f"Ortholog motifs: {len(ortholog_df)}, Regex motifs: {len(regex_df)}")
Gene expression, tissue expression, correlated genes, single-cell data.
import gget
# Tissue expression from ARCHS4
tissue_expr = gget.archs4("ACE2", which="tissue")
print(f"Expression across {len(tissue_expr)} tissues")
# Correlated genes from ARCHS4
correlated = gget.archs4("ACE2", which="correlation")
print(f"Top correlated gene: {correlated.iloc[0]['gene_symbol']}")
import gget
# Single-cell data from CELLxGENE (requires gget setup cellxgene)
adata = gget.cellxgene(
gene=["ACE2", "TMPRSS2"],
tissue="lung",
cell_type="epithelial cell",
census_version="2023-07-25" # pin version for reproducibility
)
print(f"Cells: {adata.n_obs}, Genes: {adata.n_vars}")
# Orthologs and expression from Bgee
orthologs = gget.bgee("ENSG00000169194", type="orthologs")
print(f"Orthologs in {len(orthologs)} species")
Disease associations, drug targets, enrichment analysis.
import gget
# Disease associations from OpenTargets
diseases = gget.opentargets("ENSG00000169194", resource="diseases", limit=10)
print(f"Associated diseases: {len(diseases)}")
# Drug associations
drugs = gget.opentargets("ENSG00000169194", resource="drugs", limit=10)
print(f"Associated drugs: {len(drugs)}")
# OpenTargets resources: diseases, drugs, tractability, pharmacogenetics,
# expression, depmap, interactions
import gget
# Enrichment analysis via Enrichr
# Database shortcuts: 'pathway' (KEGG), 'transcription' (ChEA),
# 'ontology' (GO_BP), 'diseases_drugs' (GWAS), 'celltypes' (PanglaoDB)
enrichment = gget.enrichr(
["ACE2", "AGT", "AGTR1", "TMPRSS2", "DPP4"],
database="ontology"
)
print(f"Enriched terms: {len(enrichment)}")
print(enrichment[["Term", "Adjusted P-value"]].head())
Cancer mutations, copy number alterations, and somatic mutation databases.
import gget
# Search cBioPortal studies
studies = gget.cbio_search(["breast", "lung"])
print(f"Studies found: {len(studies)}")
# Plot cancer genomics heatmap
gget.cbio_plot(
["msk_impact_2017"],
["AKT1", "ALK", "BRAF"],
stratification="tissue",
variation_type="mutation_occurrences"
)
import gget
# COSMIC: requires account + local database download
# First-time: gget.cosmic(searchterm="", download_cosmic=True,
# email="[email protected]", password="xxx", cosmic_project="cancer")
cosmic_results = gget.cosmic("EGFR", cosmic_tsv_path="cosmic_data.tsv", limit=10)
print(f"COSMIC mutations: {len(cosmic_results)}")
Generate mutated sequences and manage module dependencies.
import gget
import pandas as pd
# Generate mutated sequences from mutation annotations
mutations_df = pd.DataFrame({
"seq_ID": ["seq1", "seq1"],
"mutation": ["c.4G>T", "c.10del"]
})
mutated = gget.mutate(["ATCGCTAAGCTGATCG"], mutations=mutations_df)
print(f"Generated {len(mutated)} mutated sequences")
gget organizes 20+ modules by domain. Python interface uses gget.<module>():
| Domain | Modules | Primary Database |
|--------|---------|-----------------|
| Gene reference | ref, search, info, seq | Ensembl, UniProt, NCBI |
| Sequence alignment | blast, blat, muscle, diamond | NCBI BLAST, UCSC, local |
| Protein structure | pdb, alphafold, elm | RCSB PDB, AlphaFold2, ELM |
| Expression | archs4, cellxgene, bgee | ARCHS4, CZ CELLxGENE, Bgee |
| Disease/drugs | opentargets, enrichr | OpenTargets, Enrichr |
| Cancer | cbio, cosmic | cBioPortal, COSMIC |
| Utilities | mutate, setup, gpt | local / OpenAI |
| Context | Default Format | Alternatives |
|---------|---------------|-------------|
| Python | DataFrame or dict | json=True for JSON; save=True to file |
| CLI | JSON | -csv for CSV; -o file to save |
| Sequences | FASTA (seq, mutate) | -- |
| Structures | PDB file (pdb, alphafold) | JSON alignment error data |
| Single-cell | AnnData object (cellxgene) | meta_only=True for metadata only |
| Visualization | PNG (cbio plot) | show=True for interactive display |
| Shortcut | Full Database Name |
|----------|-------------------|
| 'pathway' | KEGG_2021_Human |
| 'transcription' | ChEA_2016 |
| 'ontology' | GO_Biological_Process_2021 |
| 'diseases_drugs' | GWAS_Catalog_2019 |
| 'celltypes' | PanglaoDB_Augmented_2021 |
Custom libraries: pass any Enrichr library name directly (e.g., "Jensen_TISSUES").
| Resource | Description |
|----------|------------|
| diseases | Disease associations with evidence scores |
| drugs | Drug associations and clinical trial data |
| tractability | Target tractability assessment |
| pharmacogenetics | Pharmacogenetic variants |
| expression | Baseline tissue expression |
| depmap | DepMap gene-disease effects |
| interactions | Protein-protein interactions |
Pin database versions for consistent results across analyses:
import gget
# Pin Ensembl release
ref = gget.ref("homo_sapiens", release=112)
# Pin CELLxGENE Census version
adata = gget.cellxgene(gene=["ACE2"], census_version="2023-07-25")
# Always record gget version
print(f"gget version: {gget.__version__}")
Goal: Find genes of interest, get their sequences, and perform enrichment analysis.
import gget
# 1. Search for genes
results = gget.search(["GABA", "receptor"], species="homo_sapiens")
gene_ids = results["ensembl_id"].tolist()[:10]
# 2. Get detailed information
info = gget.info(gene_ids)
print(f"Retrieved info for {len(info)} genes")
# 3. Get protein sequences
sequences = gget.seq(gene_ids, translate=True)
# 4. Find correlated genes
correlated = gget.archs4(info.index[0], which="correlation")
# 5. Enrichment analysis on correlated genes
gene_list = correlated["gene_symbol"].tolist()[:50]
enrichment = gget.enrichr(gene_list, database="ontology")
print(f"Top enriched term: {enrichment.iloc[0]['Term']}")
Goal: Investigate a gene's disease associations, druggability, and cancer mutations.
import gget
gene_id = "ENSG00000169194" # ZBTB16
# 1. Disease associations
diseases = gget.opentargets(gene_id, resource="diseases", limit=20)
# 2. Drug associations
drugs = gget.opentargets(gene_id, resource="drugs")
# 3. Tractability assessment
tractability = gget.opentargets(gene_id, resource="tractability")
# 4. Protein interactions
interactions = gget.opentargets(gene_id, resource="interactions")
print(f"Diseases: {len(diseases)}, Drugs: {len(drugs)}, Interactions: {len(interactions)}")
# 5. Cancer genomics
gget.cbio_plot(["msk_impact_2017"], ["ZBTB16"], stratification="cancer_type")
Goal: Compare a gene across species using orthologs and sequence alignment.
import gget
# 1. Find orthologs
orthologs = gget.bgee("ENSG00000169194", type="orthologs")
# 2. Get sequences for human and mouse
human_seq = gget.seq("ENSG00000169194", translate=True)
mouse_seq = gget.seq("ENSMUSG00000026091", translate=True)
# 3. Align sequences
alignment = gget.muscle([human_seq, mouse_seq])
# 4. Get human protein structure from PDB
pdb_structure = gget.pdb("7S7U")
print("Comparative analysis complete")
| Parameter | Module(s) | Default | Range / Options | Effect |
|-----------|-----------|---------|-----------------|--------|
| species | search, archs4, cellxgene, enrichr | "homo_sapiens" | Any Ensembl species; shortcuts: 'human', 'mouse' | Target organism |
| limit | blast, opentargets, cosmic | 50 / 100 | 1-1000 | Maximum results returned |
| database | blast, enrichr | varies | blast: nt/nr/swissprot/pdbaa; enrichr: shortcuts or library names | Target database for query |
| which | ref, archs4 | varies | ref: gtf,cdna,dna,cds,pep; archs4: correlation,tissue | Data type to retrieve |
| translate | seq | False | True/False | Return amino acid instead of nucleotide sequences |
| resource | opentargets | "diseases" | diseases, drugs, tractability, pharmacogenetics, expression, depmap, interactions | OpenTargets data type |
| release | ref, search | latest | Integer Ensembl release number | Pin database version for reproducibility |
| census_version | cellxgene | "stable" | "stable", "latest", date string | Pin CELLxGENE Census version |
| sensitivity | diamond, elm | "very-sensitive" | fast to ultra-sensitive | Alignment sensitivity vs speed |
| threads | diamond, elm | 1 | 1-N | CPU threads for alignment |
| multimer_recycles | alphafold | 3 | 3-20 | Higher = more accurate multimer prediction |
release=112 for Ensembl and census_version="2023-07-25" for CELLxGENE to ensure consistent results across analyses.time.sleep(2) between BLAST/BLAT queries in loops. For gget.info(), limit to ~1000 IDs per call.pip install --upgrade gget regularly to avoid breakage from schema changes.-csv is better for quick one-off lookups.gget.pdb() is instant; AlphaFold prediction takes minutes to hours. Always check if the structure already exists in PDB.'pathway', 'ontology', etc.) map to curated Enrichr libraries. For custom analyses, pass any Enrichr library name directly.data_dir="./cache" parameter to avoid re-downloading large cancer genomics datasets.When to use: Need information for many genes at once (up to ~1000 IDs per call).
import gget
import time
gene_ids = ["ENSG00000012048", "ENSG00000139618", "ENSG00000141510"]
info = gget.info(gene_ids)
info.to_csv("gene_info_batch.csv")
print(f"Saved info for {len(info)} genes")
# For >1000 genes, batch with rate limiting
all_ids = [f"ENSG{i:011d}" for i in range(2000)]
results = []
for i in range(0, len(all_ids), 500):
batch = all_ids[i:i+500]
results.append(gget.info(batch))
time.sleep(1)
When to use: Running enrichment against a custom background gene set.
import gget
# Use specific Enrichr library with background genes
enrichment = gget.enrichr(
["ACE2", "AGT", "AGTR1"],
database="Jensen_TISSUES",
background_list=["ACE2", "AGT", "AGTR1", "TP53", "BRCA1", "MYC"]
)
print(enrichment[["Term", "Adjusted P-value"]].head())
When to use: Predicting and visualizing protein structures with confidence coloring.
import gget
# Predict with visualization (PAE + 3D structure)
result = gget.alphafold(
"MKWMFKEDHSLEHRCVESAKIRAKYPDRVPVIVEKVSGSQIVDIDKRKYLVPSDITVAQFMWIIRKRIQLPSEKAIFLFVDKTVPQSR",
plot=True,
show_sidechains=True,
relax=True # AMBER relaxation for final structure
)
# Output: PDB file + predicted aligned error (PAE) JSON
# PAE heatmap auto-generated with plot=True
When to use: Setting up reference files for RNA-seq alignment pipelines.
# Download GTF and cDNA for human (specific release)
gget ref -w gtf -w cdna -d -r 112 homo_sapiens
# Download genome DNA
gget ref -w dna -d homo_sapiens
| Problem | Cause | Solution |
|---------|-------|----------|
| ModuleNotFoundError: gget | Package not installed | pip install gget in clean virtual environment |
| gget setup alphafold fails | Python version incompatibility | Use Python 3.8-3.10; check gget --version |
| Empty BLAST results | Sequence too short or no matches | Try longer sequence, different database, or megablast_off=True |
| cellxgene gene not found | Case-sensitive gene symbols | Use 'ACE2' for human, 'Ace2' for mouse (exact capitalization required) |
| gget info timeout | Too many IDs at once | Limit to ~1000 Ensembl IDs per call; batch with time.sleep() |
| Database structure changed | gget databases update biweekly | pip install --upgrade gget |
| COSMIC authentication error | Missing or expired credentials | Re-enter email/password; check COSMIC account status |
| AlphaFold out of memory | Protein too long for GPU memory | Use shorter sequences or split into domains |
| Different results on re-run | Database updated between runs | Pin versions: release=112 for Ensembl, census_version for CELLxGENE |
2 reference files provide extended coverage of capabilities from the original 3 reference files and 3 script files:
references/module_parameters.md — Consolidates module_reference.md (468 lines). Covers: detailed parameter tables for all 15+ modules with types, defaults, and return value descriptions; CLI vs Python interface differences; setup requirements per module. Relocated inline: most-used module parameters (Core API code blocks), output format summary (Key Concepts table). Omitted: gget gpt module details — trivial OpenAI wrapper, not genomics-specific.references/databases_workflows.md — Consolidates database_info.md (301 lines) and workflows.md (815 lines). Covers: complete database directory with update frequencies and citation info, extended workflow examples (building reference indices, disease-drug pipeline, multi-species comparative analysis), data consistency and reproducibility guidance. Relocated inline: core database overview (Key Concepts table), top 3 workflows (Common Workflows), reproducibility patterns (Key Concepts). Omitted: scripts/ content (3 files, 590 lines total) — thin wrappers around gget API calls for CLI automation; core patterns absorbed into Core API and Common Workflows.gget.cellxgene()Take jaechang-hits/gget-genomic-databases from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.
The instructions reference pip.
Without those the skill loads but fails at the first command.