Single-Cell Biology · Xaira-Therapeutics
X-Atlas-Orion
A genome-wide Perturb-seq atlas of 8 million human cells across 18,903 protein-coding gene knockdowns.
What X-Atlas-Orion is
X-Atlas: Orion edition (X-Atlas/Orion) is a large-scale single-cell transcriptomics perturbation atlas generated by Xaira Therapeutics. It comprises two genome-wide Fix-Cryopreserve-ScRNAseq (FiCS) Perturb-seq screens targeting all 18,903 human protein-coding genes.
The dataset contains eight million cells across HCT116 and HEK293T cell lines, deeply sequenced to a median depth of 16,000 unique molecular identifiers (UMIs) per cell. It is designed to capture gene-dosage effects and support the training of perturbation-aware biological foundation models while mitigating common scalability and variability challenges in single-cell data generation.
How it was made
Data were generated using a Fix-Cryopreserve-ScRNAseq (FiCS) Perturb-seq platform across HCT116 and HEK293T cell lines, achieving a median of at least 140 cells per perturbation. The median on-target knockdown efficiency reached 75.4% in HCT116 cells and 51.5% in HEK293T cells.
Sequencing reads were aligned to the 10x Genomics GRCh38 2024-A pre-built reference genome. Official gene symbols and Ensembl identifiers were extracted from the corresponding genes.gtf annotation file. The data are structured with sparse non-zero gene expression counts, gene token identifiers, cell barcodes, guide identifications, and quality metrics including mitochondrial read percentages.
Key facts
- Cell lines
- HCT116 and HEK293T
- Number of cells
- 8 million
- Targeted genes
- 18,903 human protein-coding genes
- Sequencing depth
- Median 16,000 UMIs per cell
- Perturbation coverage
- Median of at least 140 cells per perturbation
- Knockdown efficiency
- 75.4% median in HCT116; 51.5% median in HEK293T
- Reference genome
- 10x Genomics GRCh38 2024-A
- File formats
- Parquet and h5ad
What you can use it for
- Foundation model pre-training
- Train biological foundation models that incorporate perturbation responses and gene-dosage effects at single-cell resolution.
- Systematic gene function discovery
- Assess transcriptomic phenotypes resulting from the knockdown of any of the 18,903 human protein-coding genes.
- Cross-cell-line perturbation comparison
- Compare regulatory networks and knockdown responses across distinct biological backgrounds using the matched HCT116 and HEK293T screens.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from Xaira-Therapeutics/X-Atlas-Orion at the moment you asked. Nothing is cached or stored — every row above came from that request.
Strengths and limits
Good at
- Provides genome-wide coverage targeting all 18,903 human protein-coding genes.
- Offers deep single-cell sequencing with a median of 16,000 UMIs per cell.
- Ensures statistical power with a median of at least 140 cells captured per perturbation.
- Delivers substantial on-target knockdown efficiency, reaching a median of 75.4% in HCT116 cells.
Watch out for
- Knockdown efficiency varies across cell contexts, falling to a median of 51.5% in HEK293T cells.
- Cellular diversity is restricted to two immortalised cell lines, HCT116 and HEK293T.
Questions
Can I look inside X-Atlas-Orion without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is X-Atlas-Orion under?
Its authors declare cc-by-nc-sa-4.0. Read the terms on the original page before reusing the data.
Which cell lines are included in the dataset?
The dataset contains single-cell profiles from HCT116 and HEK293T cell lines.
What reference genome was used to process the reads?
All samples were aligned to the 10x Genomics GRCh38 2024-A reference genome.
How are gene expression levels represented?
Expression is provided as raw counts for genes with non-zero expression, indexed using gene token identifiers mapped to Ensembl IDs and official gene symbols.
How can the dataset be accessed?
It can be loaded in streaming mode or split by cell line (HCT116 or HEK293T) via Parquet datasets, or accessed as processed h5ad files.
How to cite it
Huang, Ann C and Hsieh, Tsung-Han S and Zhu, Jiang and Michuda, Jackson and Teng, Ashton and Kim, Soohong and Rumsey, Elizabeth M and Lam, Sharon K and Anigbogu, Ikenna and Wright, Philip and Ameen, Mohamed and You, Kwontae and Graves, Christopher J and Kim, Hyunsung John and Litterman, Adam J and Sit, Rene V and Blocker, Alex and Chu, Ci. X-Atlas/Orion: Genome-wide Perturb-seq Datasets via a Scalable Fix-Cryopreserve Platform for Training Dose-Dependent Biological Foundation Models. bioRxiv (2025). https://doi.org/10.1101/2025.06.11.659105v1
Written from X-Atlas-Orion's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.