Abstract
BACKGROUND: Spinal cord injury (SCI) causes permanent neurological deficits and is associated with complex pathological processes such as angiogenesis and programmed cell death (PCD). However, the crosstalk between angiogenesis and PCD in SCI remains unclear. METHODS: We integrated angiogenesis-related and PCD-related datasets to identify angiogenesis- and PCD-associated differentially expressed genes (ARPRDEGs). Weighted gene coexpression network analysis (WGCNA), protein–protein interaction (PPI), and machine learning were used to screen hub genes, followed by time-course qRT-PCR and Western blot validation in a transection mouse model. We also performed immune infiltration analysis, drug–gene interaction prediction, Mendelian randomization (MR) analysis using GWAS and eQTL data from SCI patients, and single-cell transcriptomics with CellChat communication analysis. RESULTS: A total of 134 ARPRDEGs were identified and enriched in cell adhesion, leukocyte migration, and wound healing. Three hub genes (Cd44, Icam1, and Tnf) were identified and validated by qRT-PCR and Western blot. Time-course analysis revealed Cd44 was consistently upregulated, Icam1 showed early elevation followed by decline, while Tnf exhibited stage-dependent downregulation, highlighting the impact of injury stage. MR analysis provided genetic evidence supporting potential causal associations: genetically predicted Cd44 and Icam1 expression were suggested as risk factors, while Tnf showed a potential protective effect. A diagnostic model based on these hub genes demonstrated good predictive performance across species. Single-cell transcriptomics further revealed that VISFATIN signaling mediated endothelial–myeloid communication in the high-expression hub-gene group, reinforcing their biological plausibility. In addition, acetaminophen was predicted as a potential candidate through computational drug screening, with exploratory in vivo findings suggesting Cd44-related effects; its therapeutic role requires further pharmacological validation. CONCLUSION: This study systematically revealed the role of angiogenesis-PCD interplay in SCI and identified Cd44, Icam1, and Tnf as promising candidate diagnostic and therapeutic targets, whose clinical translation requires further validation. By integrating transcriptomic, protein-level, single-cell, clinical, and genetic evidence, our findings provide novel insights into stage-specific mechanisms and potential therapeutic directions, while highlighting the need for further pharmacological validation for SCI. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12967-025-07405-2.