Cross-Cohort Transcriptomic Analysis of Keloid Identifies Candidate Lesion Signals and Fibroblast-State Heterogeneity — Frozen Release

Frozen computational-research release accompanying the manuscript "Cross-Cohort Transcriptomic Analysis of Keloid Identifies Candidate Lesion Signals and Fibroblast-State Heterogeneity". Contains the complete, version-locked analysis outputs for a cross-cohort transcriptomic synthesis of keloid: nine public GEO datasets (136 bulk samples) integrated with donor-level pseudobulk single-cell RNA-seq (40,655 cells); a prespecified random-effects meta-analysis of the mature-lesion contrast (five cohorts, Hartung–Knapp–Sidik–Jonkman inference, metafor REML); a four-layer (spatial-mechanistic) disease-signature decomposition (S2 local-lesion, S3 nonlesional, S4 longitudinal trauma response, plus cell-intrinsic fibroblast C4); donor-level single-cell pseudobulk DEGs across eight fibroblast subclusters; and an auditable 7-dimension drug-prioritization (MCDA) framework with explicit 0–5 evidence anchors and weight-sensitivity analysis. Includes: environment/sessionInfo (R 4.3.3 + package versions); metadata/ (prespecified statistical analysis plan, contrast matrix, cohort definitions, inclusion/exclusion criteria, dataset dependency map); scripts/ (numbered pipeline 01→15 plus HKSJ rerun and figure-regeneration); results/ (HKSJ meta-analysis tables, sensitivity analyses, exploratory candidate signals, MCDA matrix, per-cohort DEG tables, four-signature gene lists, S2 replication, GSE113619 longitudinal sample maps and full-rank design matrices, scRNA donor×cluster QC, frozen CMap query signature, and supplementary_tables.xlsx with one sheet per final object); figures/ (JID-compliant <=180 mm Arial); and a SHA-256 manifest (manifest_sha256.csv) of every released file. All inputs are public GEO accessions; no individual-level human data are included. License: CC-BY-4.0.