Code and data for: A reproducible inflammatory, redox, and hypoxia injury program in diabetic animal retina with partial human macular concordance: a direction-aware transcriptomic meta-analysis

Code and derived data accompanying the manuscript "A reproducible inflammatory, redox, and hypoxia injury program in diabetic animal retina with partial human macular concordance: a direction-aware transcriptomic meta-analysis" (under review). This archive contains the analysis code, the frozen derived-result tables, the dataset-curation manifests, and a package-version manifest needed to reproduce the study. No raw sequencing or platform-native files are redistributed; every primary dataset is public and is retrieved from its original repository (NCBI GEO and ArrayExpress/BioStudies; search performed 20 June 2026). The study has two layers with deliberately different evidential weight. Primary (discovery): a direction-aware, accession-conscious random-effects meta-analysis of animal (mouse and rat) retinal transcriptomes — 16 comparison units from 11 accessions, 16,538 genes, per-gene REML synthesis (metafor), Hedges' g standardization, and leave-one-accession-out stability, yielding 333 direction-stable core genes (144 higher, 189 lower in diabetic/DR-model retina). Secondary (bounded concordance): a donor-aware test of whether the animal signals correspond, in direction and at the pathway level, in one human macular dataset (GSE160306; 36 donors, 26 cases, 10 controls; 60.1% directional concordance). A cell-type localization layer maps the core onto a human retinal single-cell atlas (Menon et al. 2019). The animal stable core is a within-animal reproducibility statement. It is not a definitive human diabetic-retinopathy gene core, and nothing here asserts causality, temporal order, cellular origin, or therapeutic efficacy. See README.md for details.