Integrative bulk and single-cell transcriptomics links amino acid metabolism-related biomarkers to endothelial immune remodeling in periodontitis

Background Periodontitis (PD) is a chronic inflammatory disease driven by microbial dysbiosis and immune-mediated periodontal tissue destruction. Although amino acid metabolic dysregulation has been implicated in inflammatory diseases, its diagnostic relevance and cellular context in PD remain unclear. We aimed to identify amino acid metabolism-related diagnostic biomarkers and characterize their potential roles in PD. Methods Bulk transcriptomic datasets (GSE16134, GSE10334, and GSE223924) and a single-cell RNA sequencing dataset (GSE164241) were analyzed. Amino acid metabolism-related differentially expressed genes were identified and functionally annotated through enrichment analyses. LASSO regression, Boruta, and random forest jointly screened diagnostic biomarkers. A nomogram was constructed and assessed by ROC, calibration, and decision curve analyses. Immune infiltration, consensus clustering, and weighted gene co-expression network analysis were performed; cell-type localization, cell-cell communication, and pseudotime analyses used single-cell data. Human gingival tissue samples were used for validation through histological analysis, immunohistochemistry, and quantitative real-time PCR (RT-qPCR). Results Six amino acid metabolism-related biomarkers showed concordant expression changes across the three bulk datasets. TDO2, ICAM2, VWF and FABP4 were upregulated in periodontitis, whereas ACAT2 and AHNAK2 were downregulated. The six-gene nomogram discriminated PD from controls in GSE16134 (AUC = 0.956), while cross-platform evaluation revealed calibration limitations. Enrichment analysis linked these biomarkers to chemokine and cytokine signaling, cell adhesion and leukocyte transendothelial migration. Immune infiltration and consensus clustering revealed broad immune associations and three expression-based subgroups with distinct inflammatory features. Single-cell analysis localized the highest biomarker activity to endothelial cells, which showed the strongest diagnostic biomarker module score despite a relative reduction in their proportion. Pseudotime analysis revealed biomarker expression variation along an inferred endothelial trajectory, while CellChat predicted altered endothelial–immune communication. Experimental validation confirmed these changes for TDO2, ICAM2, VWF, FABP4 and ACAT2 at both mRNA and protein levels in clinical samples. Conclusion We identify six amino acid metabolism-related candidate biomarkers with diagnostic potential for PD—five supported by clinical gingival tissue expression data—and link them to endothelial cell state changes and immune microenvironment remodeling. These findings provide an immune-metabolic framework for biomarker development and mechanistic studies in PD.