seqout

Clients for seqout.org, which aggregates study metadata from GEO, SRA, ENA, DDBJ, ArrayExpress, GEA and GSA.

Seqout client can be used search across all the databases, resolve an accession to its records in every archive, and read GEO supplementary files as counts matrices.

Python

uv add seqout | seqout.org/cli/python

from seqout import connect

sq = connect()
sq.search("liver cancer scRNA", organism="Homo sapiens")
sq.get("GSE151530").samples

The package also installs a seqout command:

seqout search "liver cancer scRNA" --organism "Homo sapiens"
seqout show GSE151530

seqout can also be installed system-wide as a CLI tool:

uv tool install seqout

R

pak::pak("saketlab/seqout/R") or r-universe | seqout.org/cli/R

library(seqout)

# No connection needed; every function reads the REST API by default
project("GSE151530")
project_samples("GSE151530")

seqout_search("liver cancer scRNA", organism = "Homo sapiens")

# Opt in to Parquet for SQL over the whole index
con <- seqout_connect("parquet")
query("SELECT count(*) FROM unified_metadata WHERE n_samples >= 10", con = con)

From a GEO accession to a single-cell matrix

Seqout library (both python and R) can be used to directly fetch counts, cell-level and sample-level metadata at once, i.e., you do not need to leave your programming environment.

Supplementary files are grouped into units that read as one matrix: a 10x triplet, a CellRanger .h5, an .h5ad, an .rds or a table. The manifest resolves them without downloading anything, and the donor covariates come from the same accession.

from seqout import seqout_counts

counts = seqout_counts(gse="GSE297547")
counts.manifest()                       
m = counts.matrix(sample="GSM8994520")  
counts.design                           

counts <- seqout_counts("GSE297547")
manifest(counts)
m <- seqout_matrix(counts, sample = "GSM8994520")
sample_frame(project_samples("GSE297547"))