Mapping Nutritional Repair Capacity in Diabetic Foot Ulcer Healing Using Single-Cell and Bulk Transcriptomic Analysis with Exploratory Spatial-Context Assessment
Description
Diabetic foot ulcer (DFU) healing failure reflects the convergence of neuropathy, vascular impairment, infection, mechanical loading, hypoxia, chronic inflammation, and metabolic dysregulation. Although nutrition is clinically relevant to wound repair, nutritional factors are often treated as supportive measures rather than being linked to wound-cell programs in a measurable transcriptomic framework. This data release accompanies a framework-driven exploratory transcriptomic study using an author-defined nutrition-repair framework termed Nutritional Repair Capacity (NRC). The study mapped nutrition-relevant repair programs in public DFU datasets and assessed their cell-class localization, healing-status-associated directional patterns, bulk tissue-level contextual support, and exploratory GeoMx spatial-context patterns. A structured repository search and feasibility audit identified GSE165816 as the discovery scRNA-seq dataset and GSE231643 as an independent comparator scRNA-seq dataset. GSE143735, GSE134431, and GSE199939 were retained as bulk RNA-seq contextual-support datasets, and GSE166120 GeoMx Digital Spatial Profiling data were analyzed as an exploratory spatial-context layer. Eight predefined NRC_primary_v1 modules represented ECM/collagen repair, epithelial repair, angiogenesis/endothelial repair, inflammatory burden, inflammatory resolution, hypoxia/metabolic stress, oxidative-stress/antioxidant response, and antimicrobial/wound ecology response. Single-cell NRC scores were calculated using deterministic mean normalized expression and aggregated at the sample × cell-class × module level to reduce cell-level pseudoreplication. A total of 72,262 filtered cells were analyzed, including 71,906 cells in the primary NRC-ready layer. NRC modules showed biologically interpretable cell-class localization, with stromal, endothelial, myeloid/macrophage, and neutrophil compartments carrying major repair-related signals. Cross-dataset analysis identified 10 Tier A directionally consistent descriptive signals, all showing lower module-score direction in dataset-defined Non-Healer samples, involving ECM/collagen repair, hypoxia/metabolic stress, inflammatory burden, and oxidative-stress/antioxidant-response programs. Bulk RNA-seq analyses provided partial module-level contextual support. The GeoMx spatial-context layer showed concordant inflammatory-burden direction but mixed or discordant directions for ECM/collagen repair and hypoxia/metabolic stress. These materials support transparency, reproducibility auditing, and manuscript review. They represent exploratory evidence mapping of nutrition-relevant repair programs, not clinical nutritional-status measurement, diagnostic or prognostic biomarker validation, intervention-response assessment, or causal mechanistic proof.
Files
Steps to reproduce
1. Download and extract `DFU_NRC_Mendeley_public_release_v1.rar`. This archive contains the main analysis tables, figures, supplementary data files, data-description files, and archived R-code records. 2. Read `README.md` first to understand file structure, dataset roles, interpretation boundaries, and code limitations. The R-code files are stepwise reproducibility records extracted from project documentation, not a single automated pipeline. 3. For dataset-identification transparency, download and extract `DFU_NRC_repository_search_audit_v1.rar`. This companion archive contains GEO search records, candidate dataset tables, deduplicated search outputs, and SRA/BioProject/RunInfo traceability checks. These files document dataset search and feasibility audit but are not required to rerun module-scoring outputs. 4. To reproduce or audit the workflow, open `code/R_code_archive/`. Start with `00_README_manifest.txt`, then review `00_ALL_EXTRACTED_R_CODE_CONSOLIDATED.txt` or the step-specific TXT files. 5. Before running code, update local paths, verify required GEO processed files or manually downloaded supplementary files, and confirm expected input filenames. The original workflow used public GEO processed expression files rather than FASTQ-level reprocessing. 6. Reproduce the workflow in stepwise order: metadata audit, scRNA-seq object construction, initial QC, filtering-threshold design, filtered-object generation, post-filter QC, annotation-readiness review, dataset-wise preprocessing, marker-based annotation review, final major cell-class locking, NRC-ready preparation, NRC_primary_v1 module construction, single-cell NRC scoring, sample × cell-class aggregation, support-filtered contrast analysis, bulk RNA-seq contextual scoring, and GeoMx spatial-context assessment. 7. For single-cell analyses, verify Seurat object compatibility, sample metadata fields, cell-class labels, and NRC-ready flags before scoring. Use the frozen `NRC_primary_v1` gene-module tables included in the release. Aggregate cell-level NRC scores to the sample × cell-class × module level to reduce cell-level pseudoreplication. 8. For bulk and GeoMx analyses, use the provided parsing, scoring, and audit tables to confirm sample grouping, gene identifiers, module coverage, and interpretation boundaries. Treat bulk and spatial-context analyses as contextual support layers, not validation of cell-class-specific scRNA-seq signals. 9. Compare regenerated outputs with the supplied result tables and figures. Small differences may occur if R package versions, Seurat object handling, normalization defaults, or local file parsing behavior differ. Document any deviations. 10. Interpret all outputs as exploratory, descriptive, and directionally oriented transcriptomic mapping results. They should not be used as evidence of a validated nutritional-status score, diagnostic signature, prognostic model, intervention-response biomarker, or causal mechanistic proof.