A stratified, direction-aware transcriptomic meta-analysis framework identifies conserved and context-dependent programs in diabetic peripheral neuropathy animal models

This repository contains the derived-data archive and the analysis and figure-rendering code supporting the study "A stratified, direction-aware transcriptomic meta-analysis framework identifies conserved and context-dependent programs in diabetic peripheral neuropathy animal models" . The framework integrates 12 publicly available animal-model (mouse/rat) transcriptomic datasets of diabetic peripheral neuropathy (DPN). Per-study differential expression (microarray and RNA-seq) was harmonized using MaxMean probe-to-gene collapse and mouse/rat→human orthology mapping (Ensembl Compara), then combined by random-effects meta-analysis (REML primary; DerSimonian–Laird and fixed-effect inverse-variance-weighted as comparators), with Bayesian prioritization (BayesMP), Morris + eFAST global sensitivity analysis, platform-offset and leave-one-out robustness checks, two-stage functional enrichment (ORA + ranked GSEA; MSigDB v2026.1.Hs), and a cross-stratum directional-concordance classification (A_core / B_robust / C_heterogeneous / D_weak). Animal A_core signals are projected onto a human DPN sural-nerve dataset (Tavares-Ferreira et al., 2022) as an external directional reference, not a validation cohort. Contents include the meta-analysis inputs and outputs, per-study DE and orthology tables, BayesMP, sensitivity and platform-offset outputs, enrichment results, the cross-stratum / A_core / biphasic classification, a consolidated per-gene master summary table, Supplementary Tables S1–S7, supplementary-figure source tables, the full analysis pipeline (R), and the figure-rendering scripts, with per-file SHA-256 checksums. Raw datasets are available from GEO under the accessions listed in DATA_AVAILABILITY.md.