Data access

Download release metadata and retained-cell tables

The portal serves lightweight release indexes and per-library Parquet tables without duplicating the underlying matrices.

Download catalog JSON Download vocabulary JSON

Find a library

Use species, GEO accession or library identifiers in the browser. The catalog records retained cells, display status and available embeddings.

Open browser

Inspect evidence

Review cluster-level labels, confidence, marker support, candidates and ontology mappings before selecting an export.

Download tables

Library pages provide retained-cell and marker Parquet files. Full expression and accessibility matrices remain in the registered H5MU collection.

Available files

Portal-ready outputs

catalog.json

Release summary and one record per formal library.

vocabulary.json

Stable type IDs, colors, evidence statuses and ontology links.

cells.parquet

Retained cell identities, QC fields, paired barcodes and RNA/ATAC/WNN coordinates.

markers.parquet

Full archived marker selection for retained clusters.

Python example

import pandas as pd

cells = pd.read_parquet("library.cells.parquet")
markers = pd.read_parquet("library.markers.parquet")

print(cells[["cell_id", "display_cell_type"]].head())

Matrix access

The web release intentionally avoids copying hundreds of gigabytes of matrices. Full paired H5MU files are registered under multiome_data/02_h5mu/official/. Apply the retained-cell selection before RNA or ATAC matrix analysis.

Open API documentation