Boom5426/iPSC-CRISPRi-PerturbSeq

Human iPSC CRISPRi Perturb-seq: mirror of two public datasets

This repository re-hosts processed data from two independent CRISPRi Perturb-seq studies in human induced pluripotent stem cells (iPSCs). Each study has its own folder, its own license, and its own citation.

This is not original work by the uploader. All data were produced and released by the original authors listed below. They are mirrored here only to make them easier to download. Please cite the original papers and datasets.

Folder	Study	Source	License
doctor2024_kolf2.1j_atlas/	KOLF2.1J Perturbation Cell Atlas (Doctor, Nourreddine, Mali)	figshare+ 10.25452/figshare.plus.27261219.v1	CC BY 4.0
feng2026_ipsc_crispri_map/	Genome-scale CRISPRi map across human iPSC lines (Feng et al., Cell Genomics 2026)	figshare 10.6084/m9.figshare.26819743.v1, 10.6084/m9.figshare.27989294.v1	MIT (lfc/), CC BY 4.0 (counts/)

Every folder contains a checksums.md5 in md5sum -c format. The checksums are the ones published by the original source, and every mirrored file was checked against them before upload.

Repository history. Until 2026-10-01 this repository was named Boom5426/KOLF and held only the KOLF2.1J .h5ad files at the repository root. Hugging Face redirects the old repository name, but the old root-level file paths (for example .../resolve/main/KOLF_Strong_Perturbations.h5ad) no longer exist. Those files are now under doctor2024_kolf2.1j_atlas/, with identical content (same LFS SHA-256).

---

1. KOLF2.1J Perturbation Cell Atlas (Doctor et al.)

A genome-scale CRISPRi Perturb-seq atlas in the KOLF2.1J iPSC line.

Original title	A genome-scale CRISPRi perturbation atlas of human induced pluripotent stem cells
Authors	Yesh Doctor, Sami Nourreddine, Prashant Mali
Source dataset	figshare+, DOI 10.25452/figshare.plus.27261219.v1
License	CC BY 4.0
Associated preprint	10.1101/2024.11.03.621734
Analysis code	https://github.com/y-doctor/KOLF2.1J_Perturbation_Cell_Atlas (MIT)

All files are AnnData .h5ad (HDF5).

File (under doctor2024_kolf2.1j_atlas/)	Size	MD5
KOLF_Metabolic_Enzymes_Strong_Perturbations.h5ad	1.24 GiB	95f57fd679b98ba4f060d143959c3a66
KOLF_Chromatin_Modifiers_Strong_Perturbations.h5ad	1.42 GiB	fbd62eee013ad01be10f8ff291c8f8f5
KOLF_Chromatin_Modifiers_QC_Filtered.h5ad	4.32 GiB	246eda29efaf8ae8da86c39840678582
KOLF_Metabolic_Enzymes_QC_Filtered.h5ad	6.50 GiB	b5e17ab928cf263f679cf63bbb311a25
KOLF_Strong_Perturbations.h5ad	43.51 GiB	28bcfff0679e7c6c35bdd584f3626362
KOLF_Pan_Genome_QC_Filtered.h5ad	176.39 GiB	afd30fde1e6ad32969c29868394385d1

KOLF_Pan_Genome_QC_Filtered.h5ad is about 176 GiB. Use backed reads (anndata.read_h5ad(path, backed="r")) unless you have enough RAM to load it.

---

2. Genome-scale CRISPRi map across iPSC lines (Feng et al.)

Processed outputs of two Perturb-seq experiments run in human iPSC lines from healthy donors: a genome-wide screen and a targeted screen. Per the paper, the study used over 20,000 guide RNAs targeting 7,226 genes across 34 cell lines from 26 genetic backgrounds, with expression data from nearly 2 million cells.

Paper	Feng C, Peets EM, Zhou Y, et al. (22 authors; Parts L, Stegle O, Velten B senior). A genome-scale single-cell CRISPRi map of trans gene regulation across human pluripotent stem cell lines. Cell Genomics 6(2):101076 (2026). DOI 10.1016/j.xgen.2025.101076
Preprint	bioRxiv, DOI 10.1101/2024.11.28.625833
Source: lfc/	figshare article 26819743, version 1, (Transcriptional Changes), DOI 10.6084/m9.figshare.26819743.v1. License: MIT
Source: counts/	figshare article 27989294, version 1, (Count Data), DOI 10.6084/m9.figshare.27989294.v1. License: CC BY 4.0 for v1. figshare v2 of the same article re-licenses it as MIT; the files in v1 and v2 are identical (same figshare file IDs and MD5).

The raw figshare API record for each article version is included as figshare_<article>_v1_metadata.json in the corresponding folder. File names are kept exactly as published on figshare.

feng2026_ipsc_crispri_map/lfc/: transcriptional changes

Descriptions follow the figshare record. Columns are as read from each file's header.

File	Size	Content
GenomeWideScreen_LFC_byGene.tsv.gz	757.8 MiB	Genome-wide screen: mean transcriptional fold-change per target and expressed gene. Columns: Target, Expressed_Gene_Symbol, Expressed_Gene_Ens_ID, lfc, pval_adj.
GenomeWideScreen_LFC_byGuide.tsv.gz	1.69 GiB	Genome-wide screen: mean transcriptional fold-change per guide and expressed gene. Columns: Target, Guide, Expressed_Gene_Symbol, Expressed_Gene_Ens_ID, lfc, pval_adj.
GenomeWideScreen_Coregulation-Target-Correlation.tsv.gz	148.0 MiB	Genome-wide screen: co-regulation of target pairs. Columns: Target_1, Target_2, R (signed).
GenomeWideScreen_Coperturbation-Expressed-Gene-Correlation.tsv.gz	505.0 MiB	Genome-wide screen: co-perturbation of expressed-gene pairs. Columns: Expressed_Gene_1, Expressed_Gene_2, R.
TargetedScreen_LFC_byGene.tsv.gz	74.0 MiB	Targeted screen: mean fold-change per target across cell lines, per expressed gene. Columns: Target, Expressed_Gene_Symbol, Expressed_Gene_Ens_ID, lfc, pval_adj.
TargetedScreen_LFC_byGuide.tsv.gz	220.9 MiB	Targeted screen: mean fold-change per guide across cell lines, per expressed gene. Columns: Target, Expressed_Gene_Symbol, Expressed_Gene_Ens_ID, lfc, pval_adj. As published, this file has no guide identifier column: it has 3x the rows of TargetedScreen_LFC_byGene over the same 444 targets.
TargetedScreen_LFC_byGene-perLine.tsv.gz	1.79 GiB	Targeted screen: mean fold-change per target, per cell line, per expressed gene. Columns: Target, Expressed_Gene_Symbol, Expressed_Gene_Ens_ID, Cell_Line, wt_expr, lfc, pval_adj.
TargetedScreen_LFC_byGuide-perLine.tsv.gz	3.35 GiB	Targeted screen: fold-change per guide, per cell line, per expressed gene. Columns: Guide, Target, Expressed_Gene_Symbol, Expressed_Gene_Ens_ID, Cell_Line, lfc, pval_adj.
00_raw_data_id_mapping.csv	120 KiB	Mapping from raw sequencing sample IDs to screen, pool, day, inlet and the cell lines used. Columns: sample_title, experiment_description, rna_type, experiment_description_shorthand, lines_used.
scripts.zip	3.2 MiB	Snapshot of the authors' analysis code (crispri_scrnaseq_hipsci-main/, R and R Markdown).
Peer Review.pdf	2.3 MiB	Peer review file, as deposited by the authors.

feng2026_ipsc_crispri_map/counts/: count data

For each screen (genome-wide screen split into fitness and non-fitness target genes, plus the targeted screen), figshare provides RNA UMI counts, guide UMI counts, and cell metadata.

File	Size	Content
GenomeWideScreen_FitnessGenes_RNA-UMI-Counts.csv.gz	1.35 GiB	RNA UMI counts, dense CSV, genes (rows, ENSG:symbol:Gene-Expression) x cells (columns).
GenomeWideScreen_FitnessGenes_Guide-UMI-Counts.csv.gz	47.9 MiB	Guide UMI counts, dense CSV, guides (rows, TARGET-protospacer) x cells (columns).
GenomeWideScreen_FitnessGenes_Cell-Metadata.tsv.gz	3.4 MiB	Per-cell metadata. Columns: Cell_ID, Batch, Guide_Call, Cell_Line.
GenomeWideScreen_NonFitnessGenes_RNA-UMI-Counts.csv.gz	1.63 GiB	RNA UMI counts, dense CSV, genes (rows, ENSG:symbol:Gene-Expression) x cells (columns).
GenomeWideScreen_NonFitnessGenes_Guide-UMI-Counts.csv.gz	99.8 MiB	Guide UMI counts, dense CSV, guides (rows, TARGET-protospacer) x cells (columns).
GenomeWideScreen_NonFitnessGenes_Cell-Metadata.tsv.gz	4.0 MiB	Per-cell metadata. Columns: Cell_ID, Batch, Guide_Call, Cell_Line.
TargetedScreen_RNA-UMI-Counts.csv.gz	4.17 GiB	RNA UMI counts, dense CSV, genes (rows, ENSG:symbol:Gene-Expression) x cells (columns).
TargetedScreen_Guide-UMI-Counts.csv.gz	70.4 MiB	Guide UMI counts, dense CSV, guides (rows, TARGET-protospacer) x cells (columns).
TargetedScreen_Cell-Metadata.tsv.gz	12.9 MiB	Per-cell metadata. Columns: Cell_ID, Batch, Guide_Call, Cell_Line.

Cell barcodes are not written identically across the three file types of a screen (for example MP-ST-P3-D3_I1_AAACCTGAGACGCAAC-1 in the metadata, ST-P3-D3_I1_AAACCTGAGACGCAAC-1 in the RNA counts, and MP.ST.P3.D3_I1_AAACCTGAGACGCAAC.1 in the guide counts), so they need to be harmonized before joining.

---

Usage

from huggingface_hub import hf_hub_download, snapshot_download
import anndata as ad
import pandas as pd

REPO = "Boom5426/iPSC-CRISPRi-PerturbSeq"

# One KOLF2.1J AnnData file
path = hf_hub_download(
    repo_id=REPO,
    repo_type="dataset",
    filename="doctor2024_kolf2.1j_atlas/KOLF_Chromatin_Modifiers_Strong_Perturbations.h5ad",
)
adata = ad.read_h5ad(path, backed="r")

# One Feng et al. LFC table
path = hf_hub_download(
    repo_id=REPO,
    repo_type="dataset",
    filename="feng2026_ipsc_crispri_map/lfc/GenomeWideScreen_LFC_byGene.tsv.gz",
)
lfc = pd.read_csv(path, sep="\t")

# A whole folder
local_dir = snapshot_download(
    repo_id=REPO,
    repo_type="dataset",
    allow_patterns=["feng2026_ipsc_crispri_map/counts/*"],
)

Verify a download with md5sum -c checksums.md5 inside the folder.

Citation

If you use the KOLF2.1J atlas (doctor2024_kolf2.1j_atlas/):

@article{doctor2024kolf,
  title   = {A genome-scale CRISPRi perturbation atlas of human induced pluripotent stem cells},
  author  = {Doctor, Yesh and Nourreddine, Sami and Mali, Prashant},
  journal = {bioRxiv},
  year    = {2024},
  doi     = {10.1101/2024.11.03.621734}
}

Dataset: Doctor Y., Nourreddine S., Mali P. KOLF2.1J Perturbation Cell Atlas. figshare+. DOI 10.25452/figshare.plus.27261219.v1. CC BY 4.0.

If you use the Feng et al. data (feng2026_ipsc_crispri_map/):

@article{feng2026crispri,
  title   = {A genome-scale single-cell {CRISPRi} map of trans gene regulation across human pluripotent stem cell lines},
  author  = {Feng, Claudia and Peets, Elin Madli and Zhou, Yan and Crepaldi, Luca and Usluer, Sunay and Dunham, Alistair and Braunger, Jana M. and Su, Jing and Strauss, Magdalena E. and Muraro, Daniele and Xian Cheam, Kimberly Ai and Bonder, Marc Jan and Nogales, Edgar Garriga and Cooper, Sarah and Bassett, Andrew and Leonard, Steven and Gu, Yong and Fussing, Bo and Burke, David and Parts, Leopold and Stegle, Oliver and Velten, Britta},
  journal = {Cell Genomics},
  volume  = {6},
  number  = {2},
  pages   = {101076},
  year    = {2026},
  doi     = {10.1016/j.xgen.2025.101076}
}

Datasets: Feng, Claudia (2025). A genome-scale single cell CRISPRi map of trans gene regulation across many human pluripotent stem cell lines (Transcriptional Changes). figshare. DOI 10.6084/m9.figshare.26819743.v1. And: (Count Data). figshare. DOI 10.6084/m9.figshare.27989294.v1.