scPerturb: single-cell perturbation data
Peidli, Green, Shen, et al.
A harmonized collection of single-cell experiments where genes or drugs were perturbed, in one format.
RNA and protein single-cell perturbation data sets from scPerturb.org, saved as h5ad files for easy analysis in Python. Each cell has expression measurements and the perturbation applied, such as a gene knockout or a drug. It is used to benchmark methods for predicting effects of perturbations.
Access
- Cost
- FreeFree download from Zenodo; no account needed.
- License
- Open licenseCC BY 4.0 (underlying studies have their own terms)
- Sign-up
- No sign-up
- How you get it
- Bulk downloadH5AD
What's in it
- One row is
- One cell with expression data and perturbation label
- Covers
- Not geographic
- Years
- Experiments published through about 2022
- Updated
- Irregular
- Size
- Dozens of datasets, millions of cells
Quality
Peer-reviewed in Nature Methods with harmonized processing; reflects published experiments only.
- Heterogeneity across labs and assays
- Large files
Connects to
- Perturb-seq datasets (Replogle et al.): Large genome-wide perturbation screens included in the collection
- Human Cell Atlas: Reference cell states for context
- Gene Ontology Data Archive: Annotations to interpret perturbation effects
Questions it could answer
- How well can models predict a gene knockout's effect on expression?
- Which perturbation types produce the strongest shifts?