CELLxGENE Census-derived networks

Download v1.0.0 on Zenodo · Try the interactive example

Precomputed exploratory gene regulatory networks for 704 human Cell Ontology labels, generated with the Huang Lab at Mount Sinai from CELLxGENE Census 2025-11-08. Reuse the results without HPC access or rerunning inference.

The collection includes 137,766,796 retained TF-to-target edges and 51,527 motif-supported regulons, drawing on 565 contributing datasets. Each label uses up to 5,000 cells balanced by dataset and donor.

Find and load results

  1. Use the JSON catalogue or Parquet index to find a label and its cohort_index.
  2. Download the corresponding archive from Zenodo. Eight archives each contain 88 cohorts, with index ranges in their filenames. For example, endothelial cells (CL:0000115, index 43) are in rustscenic-cellxgene-human-networks-v1.0.0-cohorts-001-088.tar.gz.
  3. Extract the archive, enter rustscenic-cellxgene-human-networks-v1.0.0, then run:
python -m pip install -r requirements-reuse.txt
python reuse_example.py --cell-type-id CL:0000115

Use Python 3.11 or later. For the complete collection, extract all eight archives into the same directory and run python verify_package.py --require-complete inside the extracted folder. The release README includes full instructions.

Each cohort includes network tables, motif evidence, regulons, aggregate AUCell activity summaries, metadata and source citations. Methods, checksums and the exact development workflow are archived. Raw expression matrices, cell-level records and input motif databases are excluded.

Interpretation

All 704 outputs passed file-integrity and content-consistency checks. Biological accuracy and cross-study replication have not been established. Labels include broad annotations; cohorts pool tissues and disease states and use one fixed random seed. Motif support does not experimentally confirm an interaction. Activity scores are not directly comparable across cohorts with different gene sets. See the published methods.

Citation and credits

Kahraman, E., Huangfu, L., El Moussaoui, M., & Huang, K.-L. (2026). RustScenic: exploratory human gene regulatory networks for 704 CELLxGENE Census labels (Version 1.0.0) [Data set]. Zenodo. doi:10.5281/zenodo.23548341.

See Contributions for project roles.

Also cite CELLxGENE and the contributing studies listed in each cohort's sources.json. See the attribution notes.

Derived data and resource documentation: CC BY 4.0. Included software: Apache-2.0. See the dataset licence. This resource is independently hosted on Zenodo. CELLxGENE is the source platform, not the host or endorser. The dataset DOI is separate from the RustScenic software DOI.