Data and analysis resources for cross-cohort reconstruction of multicellular remodeling in atopic dermatitis
Description
This repository contains the processed data, analysis-ready tables, figure-source data, and reproducible analysis code supporting the study “Cross-cohort reconstruction reveals multicellular spatial remodeling of skin neighborhoods in atopic dermatitis.” The study integrates publicly available single-cell and spatial transcriptomic resources to investigate multicellular lesion-associated remodeling in atopic dermatitis (AD) using biological donors or patients as the principal inferential units. Donor-aware pseudobulk analysis of paired lesional and non-lesional skin in GSE204762 was used to define cell-state-specific transcriptional programs, which were subsequently evaluated in the independent Reynolds human skin atlas. Regulatory analyses assessed transcription-factor activity together with program-level target coherence. Spatial analyses used the multicellular neighborhoods reported in a normal-human-skin atlas as fixed reference expression profiles. Lesion-associated changes in neighborhood-reference scores were defined in GSE197023 and independently evaluated in GSE206391. Additional graph-based analyses assessed spatial autocorrelation, inter-neighborhood coupling, cross-compartment organization, and patterns shared across AD, psoriasis, and lichen planus. The repository is intended to support transparency, reproducibility, and reuse of the analytical workflow. It includes derived numerical results, statistical summaries, source data for manuscript figures and tables, and scripts required to reproduce the principal analyses and visualizations. Original sequencing and spatial-transcriptomic datasets are not redistributed here; they remain available from their respective public repositories and published sources under the accession identifiers cited in the manuscript and accompanying documentation. All files are provided for research and reproducibility purposes. Users should cite both the associated manuscript and the original studies/data resources when reusing these materials.
Files
Steps to reproduce
1. Clone or download this repository and review the README and accompanying file manifests for the expected directory structure and required inputs. 2. Obtain the original public datasets from their respective repositories using the accession identifiers reported in the manuscript. The principal datasets are GSE204762 for donor-aware single-cell discovery, the Reynolds human skin atlas for independent transcriptomic evaluation, GSE197023 for spatial discovery, and GSE206391 for independent spatial evaluation and cross-disease comparison. GSE288946 is used only for the conditional regulatory follow-up described in the manuscript. Original sequencing and spatial-transcriptomic data are not redistributed in this repository. 3. Place the downloaded source files in the input directories specified in the repository documentation. Do not rename or modify source files unless explicitly instructed by the accompanying metadata or mapping tables. 4. Recreate the software environment using the versions documented in the repository. The analyses use R and Python-based workflows; package and environment information is provided with the analysis files. 5. Run the analysis scripts in their numbered workflow order. The main workflow reproduces: (i) donor-aware pseudobulk discovery and signed disease programs; (ii) external evaluation in the Reynolds cohort; (iii) regulatory activity and TF–program coherence analyses; (iv) healthy-skin reference-score projection and spatial analysis; (v) independent evaluation in GSE206391; and (vi) graph-based spatial and cross-disease analyses. 6. Use the supplied plotting and table-generation scripts to regenerate the manuscript figures, supplementary figures, and publication-facing numerical tables from the derived analysis outputs. 7. Compare reproduced outputs with the provided result tables and figure-source files. Small numerical differences attributable to software/platform implementation should be documented; the predefined analysis definitions, donor-level inferential structure, and program/reference specifications should not be altered during reproduction.
Institutions
- Second Military Medical UniversityShanghai, Shanghai