Analysis code, frozen gene signatures and result tables for: Proximal tubule stress and repair programs are consistently elevated across independent diabetic kidney disease transcriptomic cohorts and track cross-sectional clinical severity

Reproducible Analysis Artefacts for a Secondary Study of Human Kidney Transcriptomic Data in Diabetic Kidney Disease (DKD) The deposit contains: l (i) the pre-specified, hash-frozen single-nucleus-derived proximal tubule stress (aPT_VCAM1, 36 genes) and repair (repair_PT, 43 genes) gene signatures, with the SHA-256 of the frozen object (56ae9b41f3a873f7d4a8692b94f8008df9f83ac0c45f63e23a175cee5f1f0afb); l (ii) the complete R and Python pipeline used to derive the signatures from 8-donor snRNA-seq (GSE195460) and to score them in four independent bulk cohorts — GSE96804 (41 DKD / 20 controls, glomeruli), GSE142025 (RNA-seq), GSE30528 (9 / 13, glomeruli) and the Schmid Diabetes TubInt cohort (11 / 7, tubulointerstitium, accessed via Nephroseq); l (iii) all result tables (CSV/JSON) and the five vector figures reported in the manuscript; l (iv) the data-audit trail, comprising SHA-256 receipts for every downloaded primary file, a retrieval log, the per-sample metadata map and the audit report including documented protocol deviations; and l (v) the analysis plans written before the analyses were performed. Principal quantities that the code reproduces: group differences for both programs in all four cohorts (Cohen's d = 1.00, 2.70, 2.12 and 0.67-1.29), an expression-matched null with one-sided empirical P = 0.011, signature-length sensitivity across top20/50/100, a merged within-DKD inverse association with eGFR (Stouffer Z = -2.39, P = 0.0084 for aPT_VCAM1; Z = -1.96, P = 0.0248 for repair_PT; n = 32), and an association of repair_PT with proteinuria (rho = 0.65, P = 0.030). Negative result retained: a pre-registered "profibrotic PT" signature failed its own pre-specified rule (3 genes surviving, threshold >= 20) and was never frozen; this negative result is documented rather than removed. Pathway-level controls caveat: positive-control gene sets (TGF-beta, hypoxia) were recalled only in the tubulointerstitial and staged cohorts and failed or reversed in the glomerular microarray cohorts, which is reported as a compartment/platform-matching caveat for pathway-level controls. Data availability and redistribution: no primary expression data are redistributed. GSE195460, GSE96804, GSE142025 and GSE30528 are public through GEO; Schmid sample-level clinical variables were obtained through Nephroseq under its terms of use and are not included — only aggregate association statistics are provided. Statistical scope: all clinical associations are cross-sectional and exploratory, with nominal P values except where explicitly merged; no prognostic, predictive or incremental-value claim is made. Licensing: data, tables, figures and text are licensed CC BY 4.0; the analysis scripts are additionally available under the MIT licence.