Published July 3, 2026 | Version v2

The Human Skin Cell Atlas - Data Objects

  • 1. ROR icon Beiersdorf (Germany)
  • 2. EDMO icon University of Hamburg

Description

The Human Skin Cell Atlas (HSCA) integrates 34 public healthy skin scRNA-seq datasets (~820,000 cells) with curated metadata and a harmonized cell type nomenclature. We constructed a HSCA core atlas from PSU-containing datasets and applied transfer learning to integrate 21 additional PSU-free datasets, resulting in the HSCA extended. The HSCA provides a collection of data objects designed to facilitate downstream analyses in both R and Python environments.

For methodological details, benchmarking, and atlas construction, please refer to:
Düz et al. An Integrated Skin Cell Atlas Decodes the Pilosebaceous Unit. Journal of Investigative Dermatology (2026). DOI: https://doi.org/10.1016/j.jid.2026.06.1282 

The available objects are:

scRNA-seq reference atlases:

  • HSCA core (HSCA_core.rds, HSCA_core.h5ad)
  • HSCA extended (HSCA_extended.rds, HSCA_extended.h5ad)
  • Pilosebaceous unit core (psu_core.rds, psu_core.h5ad)
  • Lower hair follicle core (lower_hf_core.rds, lower_hf_core.h5ad)
  • Sebaceous gland core (sg_core.rds, sg_core.h5ad)
  • Interfollicular Epidermis extended (ife_extended.h5ad)
  • Fibroblast extended (fb_extended.h5ad)
  • Immune extended (immune_extended.h5ad)
  • Other extended (other_extended.h5ad)

Reference model for transfer learning:

  • model.pt
  • HSCA_core_scarches_emb.h5ad

Spatial transcriptomics datasets:

  • VisiumHD_full.rds
  • VisiumHD_lower_hairfollicle.rds

Each object contains raw and normalized counts, embeddings, cell type annotations, and extensive metadata. A detailed description of all slots and stored information can be found in Supplementary Table 4 of the associated manuscript.

The Python objects were stored using adata.write_h5ad(filename, compression="gzip"), ensuring efficient storage and fast loading with read_h5ad.

⚠️ Integrity check: Please verify all downloads with the provided md5sum values to ensure data integrity.

For detailed usage instructions, the latest updates, and implementations, please refer to our GitHub repository: https://github.com/TolgaDuz/HSCA  

Files

Files (49.7 GB)

Name Size
md5:57b0e250d1ce271d98d3d730a5b6af39
1.7 GB Download
md5:012ea36c390babbd7648bb25a6e0d50f
6.3 GB Download
md5:3ef5bcd0d514f88ab81d03b86d4050d1
9.0 GB Download
md5:4081295a4732d59fc54e3f434da446ad
125.2 MB Download
md5:95f83e6562e514d435f70bb84d76c88c
9.2 GB Download
md5:e023df73b2c1aa986ccdd4ee26711e90
9.0 GB Download
md5:b022e0b5af6851a4f9d6e8b2a94156a4
3.4 GB Download
md5:9d4f6f78269426f116a2f24c3c85b54a
1.2 GB Download
md5:f3ab3f0b64d6b77ecfe9c1f50c7bc148
182.4 MB Download
md5:140c6909acc7e85b2d130f4ebe217e92
283.8 MB Download
md5:1eb96eb85d2ee1bda4588a7ce1fcb56b
8.3 MB Download
md5:aeb1f5d20cd1936a831ded1c3e978705
506.3 MB Download
md5:7b8087338ad1a673e4629defe77d49ef
672.9 MB Download
md5:b6ceb9fd4bf994e5c814cb01e8eeb5c5
1.0 GB Download
md5:e4bbc809829dcdc752f7fb5d6b17d7b7
129.1 MB Download
md5:59031b5d4011418cda9f4fb5f2462ed7
210.8 MB Download
md5:415d66792830d60f88d9a43c0218cb83
5.1 GB Download
md5:4cb6182f7a1c161d930f9b136fd24aed
1.7 GB Download