Human Skin Cell Atlas data objects

Human Skin Cell Atlas This upload contains processed data objects (.h5ad files) generated for the Human Skin Cell Atlas (HSCA) project and associated publication. NOTE! The most up-to-date per-cell metadata is provided in the Metadata_HSCA_CellMetadata_Core_Extended.csv data table. Individual .h5ad files may contain metadata entries that differ from those in Metadata_HSCA_CellMetadata_Core_Extended.csv. In cases where metadata values differ between files, the information in Metadata_HSCA_CellMetadata_Core_Extended.csv should be considered the authoritative version. Metadata_HSCA_CellMetadata_Core_Extended.csv contains per-cell metadata for both the core and extended HSCA datasets, including: Study and experiment information. Donor (including bodysite annotation), sample, library preparation, sequencing, and mapping metadata. Quality-control metrics. Annotations: Original author annotations Integrated HSCA annotation levels 1–4 for core HSCA cells. Level 1 annotation for extended HSCA cells. The deposited datasets include: Separate AnnData objects for different annotation levels and cell-type subsets. Mapping and integration results with external datasets. An extended HSCA dataset object. Spatial transcriptomics mapping results for Stereo-seq and Xenium datasets. Metadata tables, cluster annotation, color mapping, and differential gene expression analysis files. Filename prefixes and suffixes: Metadata: files containing metadata regarding cells, samples or used cluster colors HSCA: Human Skin Cell Atlas core or extended datasets. Depending on the full filename, these files may represent the core HSCA atlas, a specific HSCA cell-type subset, a subcluster-level object, or the extended HSCA dataset. Steele: Data derived from the Steele et al. dataset after integration with, or mapping onto, the HSCA reference. Forsthuber: Data derived from the Forsthuber et al. cancer-associated fibroblast (CAF) dataset after integration with, or mapping onto, the HSCA reference. ObRe: Data derived from the Ober-Reynolds et al. dataset after integration with, or mapping onto, the HSCA reference. ObRe is used as a shorthand for Ober-Reynolds. Stereoseq: Stereo-seq data with Stereoscope-based mapping to the HSCA reference. Corresponding nuclei images for the included samples are provided in Stereoseq_nuclei_images.zip. Xenium: Xenium data with cell2location-based mapping to the HSCA reference. Corresponding images are included within the uploaded Xenium .zarr archive. Additional files: Metadata_Cluster_annotation_and_colors.py contains cluster-number mappings, full cluster names, abbreviated cluster names, and associated cluster colors. DGEA_L1_L2_L3_L4_Subclusters.zip contains differential gene expression analysis results for all subclustering levels, from level 1 to level 4. Metadata_SupplementaryTable_SampleMetadata_Extended.xlsx contains per-sample metadata for the core and extended HSCA dataset, including study information, donor identifiers, basic donor metadata such as sex and age, and anatomical-region annotations. Metadata_HSCA_CellMetadata_Core_Extended.csv contains the most up-to-date per-cell metadata for the core and extended HSCA datasets. This includes study, experiment, donor, sample, library preparation, sequencing, mapping, anatomical site, quality-control, original annotation, and integrated HSCA annotation metadata. README_general_annotation_metadata.md contains detailed information about metadata column contents and .h5ad data structure Differential gene expression analysis: For each analysis group, each subcluster was compared against all other cells within the same group. For example, a level 2 keratinocyte cluster was compared against all other keratinocytes. For each level and cell type, the differential expression results are provided in three formats: .csv files for programmatic access. .parquet files for efficient programmatic access. .xlsx files for manual browsing and exploration, with each subcluster provided as a separate sheet. Gene filtering: All AnnData objects were filtered to retain a curated set of standard genes obtained from BioMart. The retained genes include protein-coding and non-coding genes, while spike-ins, technical controls, and non-standard constructs were removed. This filtering resulted in approximately 35,000–40,000 genes per dataset. AnnData structure: Raw counts are stored in .X as a sparse matrix. Log-normalized counts are stored in adata.layers as a sparse matrix.