---
license: cc-by-4.0
language: en
tags: [biology, single-cell, transcriptomics, anndata, h5ad, raw-data]
size_categories: ['1M<n<10M']
---

# ConvergeCELL Source Datasets — v1.0.0

Cached H5ADs of every single-cell RNA-seq dataset registered in the
ConvergeCELL data catalog. Mirrors the original sources (GEO, figshare,
Tabula Sapiens, CellxGene) so downstream code has a single, fast, versioned
endpoint to fetch from.

Each row of every h5ad is one cell; each column is one gene (HGNC symbol).
The exact obs/var schema follows whatever the original source provided —
this bundle does not re-annotate, harmonize, or QC. For a harmonized,
pseudobulked variant see [converge-bio/pseudobulk-panel-v1](https://huggingface.co/datasets/converge-bio/pseudobulk-panel-v1).

## Datasets (18)

| Name | Tissue | Disease | # Cells | # Cell Types | Accession | Publication |
|---|---|---|---:|---:|---|---|
| `bladder_gouin` | bladder | urothelial_carcinoma | 67,988 | 5 | GSE169379 | Gouin et al., Nature Communications 2021 |
| `bladder_gouin_normal` | bladder | healthy | 25,102 | — | GSE169379 | Gouin et al., Nature Communications 2021 |
| `bladder_lai` | bladder | urothelial_carcinoma | 36,788 | — | GSE135337 | Lai et al., Int J Cancer 2021 |
| `melanoma_jerby_arnon` | skin | melanoma | 7,186 | — | GSE115978 | Jerby-Arnon et al., Cell 2018 |
| `melanoma_sade_feldman` | skin | melanoma | 16,291 | — | GSE120575 | Sade-Feldman et al., Cell 2018 |
| `melanoma_tirosh` | skin | melanoma | 4,645 | 7 | GSE72056 | Tirosh et al., Science 2016 |
| `pbmc3k` | blood | healthy | 2,700 | — | N/A | 10X Genomics |
| `skin_healthy` | skin | healthy | 50,000 | 127 | CellxGene | CellxGene |
| `ts_blood` | blood | healthy | 10,000 | 107 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_bone_marrow` | bone_marrow | healthy | 10,000 | 119 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_eye` | eye | healthy | 10,000 | 105 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_heart` | heart | healthy | 10,000 | 67 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_kidney` | kidney | healthy | 10,000 | 114 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_liver` | liver | healthy | 10,000 | 117 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_lung` | lung | healthy | 10,000 | 167 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_pancreas` | islet | healthy | 10,000 | 54 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_skin` | skin | healthy | 10,000 | 105 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |
| `ts_spleen` | spleen | healthy | 10,000 | 90 | figshare:14267219 | Tabula Sapiens Consortium, Science 2022 |

## How to load

```python
from huggingface_hub import hf_hub_download
import anndata as ad

path = hf_hub_download(
    repo_id="nicolas-lynn/vcell-perturbation-source-data",
    filename="ts_lung.h5ad",     # any name from the table above
    repo_type="dataset",
)
adata = ad.read_h5ad(path)
```

Or, programmatically via convergecell-data which will fall back to HF when
the local cache is empty:

```python
import convergecell_data as cd
adata = cd.load_dataset("ts_lung")
```

## License

CC BY 4.0 for the manifest + dataset card. Each individual h5ad retains its
upstream license — consult the publication column for the canonical source.