Endocrine. 2026 Aug 15;91(1):279. doi: 10.1007/s12020-026-04754-w.
ABSTRACT
BACKGROUND: Diabetic foot ulcers (DFUs) arise from interacting clinical, microbial, and host processes, yet public datasets differ substantially in design, scale, and evidentiary strength.
METHODS: We curated eight public DFU cohorts and defined an evidence hierarchy. PRJNA287759 was reprocessed from raw 16 S reads; outcome tests used one baseline sample per patient (74 healed and 15 non-healed/adverse). GSE134431 compared 13 DFU with 8 diabetic foot skin samples using limma. Host discrimination underwent fold-confined leave-one-sample-out validation, 100 repeated nested five-fold analyses, 200 label permutations, and independent rank-score evaluation in GSE80178.
RESULTS: Baseline diversity and community structure did not differ by outcome (Shannon P = 0.159; Observed ASVs P = 0.669; PERMANOVA R2 = 0.0157, P = 0.142). Rothia was the only nominal genus (P = 0.0307), and none survived FDR correction across prevalence thresholds. GSE134431 yielded 2,873 differentially expressed genes with coherent epidermal repair enrichment. Host leave-one-sample-out AUC was 0.990 (95% CI 0.964-1.000), repeated nested-CV median AUC was 0.981 (95% interval 0.952-1.000), and permutation P was 0.00995. The GSE80178 score separated 6 DFU from 3 diabetic foot skin samples (exact P = 0.0238; AUC = 1.000), although this external comparison was small.
CONCLUSIONS: Host expression provides the strongest evidence, whereas microbiome findings remain exploratory. The repair framework synthesizes parallel evidence without demonstrating causal coupling or a clinically deployable biomarker.
PMID:42603246 | DOI:10.1007/s12020-026-04754-w