Manuscript: "A keloid-derived fibroproliferative transcriptional program shows cross-disease enrichment in Dupuytren's disease"
This repository contains the complete analysis code, cohort metadata, gene sets, and results for the above study: a transcriptomic re-analysis testing whether a keloid-derived 125-gene fibroproliferative program is portable to Dupuytren's disease and to independent keloid cohorts, with donor-level statistical inference and specificity benchmarking against generic ECM/EMT/TGF-β and random gene sets.
scripts/ All R analysis scripts (01–11b, recurrence, audit_20260818 reconstitution)
metadata/ Sample-level manifests for all cohorts (GEO-derived) + recurrence metadata
data/gene_sets/ program125 / core47 / tail32 gene lists
results/ Full differential tables, donor-level scores, permutation results,
specificity benchmarks, recurrence QC (audit_20260818/)
docs/ Re-analysis audit report and Phase-A forensic memo
| Cohort | Role | Design | Donor-level n |
|---|---|---|---|
| GSE44270 | Discovery (keloid fibroblasts) | keloid vs adjacent(4) vs healthy(3) | 9 / 4 / 3 |
| GSE75152 | Cross-disease test (Dupuytren fascia) | 12 vs 12 | 12 / 12 |
| GSE190626 | Same-tissue RNA-seq (keloid skin) | 3 vs 3 | 3 / 3 |
| GSE117887 | Same-cell-type RNA-seq (dermal fibroblasts) | 3 keloid vs 3 normal cell lines, day1/day5 repeated | 3 / 3 |
| GSE145725 | Donor-level non-circularity check | 5 vs 5 donors (19 arrays) | 5 / 5 |
| GSE163973 | Single-cell descriptive anchoring | 3 keloid vs 3 normal-scar donors | 3 / 3 |
| GSE113620 | Exploratory miRNA time course (paired) | 13 paired Day0/Day42 | 13 |
| PRJNA813172 | Exploratory recurrence test | 7 recurrent vs 70 non-recurrent (male primary 6 vs 32) | 77 |
- Gene-set inference is at the sample/donor level (ssGSEA + exact permutation of cohort labels); gene-direction proportions are descriptive only.
- GSE117887 is analysed at the donor level (3+3 cell lines; duplicateCorrelation sensitivity); GSE145725 at the donor level (5+5).
- Specificity benchmarks in GSE75152: program125 Cohen's d = 4.07, exceeding NABA Matrisome (3.60), GO extracellular matrix (3.67), Hallmark EMT (3.01), Hallmark TGF-β (2.03), and 100 size-matched random gene sets (max |d| = 2.89).
- Recurrence (PRJNA813172): multi-gene sex QC; male-only primary analysis; robust log2-CPM checks (IL1RN not robustly supported).
Environment: R 4.3.3, WSL2/Ubuntu. All paths are resolved at runtime from a project-root .env file via path_config.R; no script contains machine-specific hard-coded paths.
Setup:
- Copy
.env.exampleto.envat the repository root. - Set
DATASET_PATHto the directory holding the raw (non-reproducible) GEO series matrices and the GSE163973 Seurat object, andDATASET_PATH_TEMPto a writable directory for reproducible outputs. - Run each script with
Rscript scripts/<name>.R(all scripts locatepath_config.Rby walking up from their own location, so they run identically from the repo checkout or the original analysis tree).
Key entry points:
scripts/01_bulk_discovery.R– discovery program assemblyscripts/recon_bulk.R– three-group model, donor-level cohort analyses, specificity benchmarksscripts/recon_singlecell.R– donor-level single-cell statisticsscripts/recon_recurrence.R– recurrence sex QC + male-only DESeq2scripts/il1rn_robust.R– robust non-parametric checks
MIT (code). All input data are from public repositories (GEO / SRA); cite the original studies (see manuscript references).