File size: 3,178 Bytes
6546c58
c589708
092211f
 
 
 
 
 
 
 
 
6546c58
092211f
c589708
092211f
c589708
 
 
 
092211f
c589708
 
 
 
 
092211f
c589708
 
 
092211f
c589708
092211f
c589708
 
 
 
 
 
 
 
 
 
 
092211f
c589708
 
092211f
c589708
092211f
c589708
 
 
 
 
 
 
 
 
 
 
 
092211f
 
 
c589708
 
 
 
 
 
092211f
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
---
pretty_name: CellClip public normalized perturbation single-cell release
license: other
task_categories:
  - feature-extraction
tags:
  - single-cell
  - transcriptomics
  - perturb-seq
  - h5ad
  - human
---

# CellClip public normalized perturbation single-cell release

This public repository contains the source-cleared, human, cell-level portion
of CellClip Stage 1: **254 H5AD files**, **12,718,270 cells**, and
**260,859,318,437 payload bytes**. These are processed derivatives rather than
the original raw download archives.

| Source partition | Units | Cells | Reference cells | Effect cells |
|---|---:|---:|---:|---:|
| scPerturb (23 genepert + 229 chempert) | 252 | 4,774,802 | 397,528 | 4,377,274 |
| XCell / X-Atlas Orion (genepert) | 2 | 7,943,468 | 384,615 | 7,558,853 |
| **Total** | **254** | **12,718,270** | **782,143** | **11,936,127** |

The repository intentionally excludes all PerturBase-derived files. Those
files have no verified package-level redistribution permission and are held in
a separate private repository. The OT-center release is also separate.

## Data contract

- `X` is per-cell `normalize_total(target_sum=1e4)` followed by natural
  `log1p`.
- Confirmed raw UMI counts are retained in `layers["counts"]`.
- `obs["is_reference"]` is the authoritative reference mask; do not infer
  controls from filenames or loose regular expressions.
- Gene perturbation targets are standardized in `target_gene_symbol`; consult
  `target_gene_symbol_status` before foundation-model vocabulary filtering.
- `obsm["X_scVI"]` is a 30-dimensional representation independently trained
  within each H5AD. It is not a shared coordinate system across files.
- Machine-readable unit, batch, source, transformation, license, and SHA-256
  metadata are under `manifest/`.

The path spelling `scPeturb/` is retained for release compatibility, while the
machine-readable source value is `scPerturb`.

## License boundary

There is no single permissive license for this repository, so the dataset card
uses `license: other`:

- `scPeturb/` is adapted from scPerturb v1.4 under **CC BY 4.0**. Reuse must
  attribute scPerturb and the original studies, link the license, and identify
  CellClip's transformations.
- `xcell/` is adapted from Xaira Therapeutics X-Atlas Orion under
  **CC BY-NC-SA 4.0**. This partition is noncommercial, requires attribution
  and change indication, and must remain ShareAlike. See `xcell/LICENSE.md`.

Keeping these path partitions together is aggregation, not a relicensing of
either source. Users must apply the terms of the partition they download.

## Integrity and selective download

`manifest/UPLOAD_MANIFEST.tsv` contains exactly the 254 H5AD objects present in
this repository. `manifest/RELEASE_FILES.tsv` and `SHA256SUMS` are generated
specifically for this repository and contain no PerturBase or OT-center paths.

The two XCell files are very large. Prefer selective downloads and backed or
sliced reads instead of cloning the full repository:

```bash
hf download cfy2yue/perturbseq_normalized \
  scPeturb/model_ready/genepert/<training-unit>.h5ad \
  --repo-type dataset --local-dir perturbseq_normalized
```