Investigation of the Molecular Mechanisms Underlying Sepsis Through Integration of Single‐Cell Sequencing and Transcriptome Sequencing

Background Sepsis remains a major challenge in critical care due to its complex pathophysiology and cellular heterogeneity. Understanding the molecular underpinnings and immune profiles of sepsis could advance diagnostic and therapeutic strategies. This study aimed to dissect the cellular and molecular landscape of sepsis, delineate the differential gene expression profiles, identify novel biomarkers through advanced machine learning algorithms, and assess the immune infiltration levels for potential diagnostic and prognostic utility. Methods Single‐cell RNA sequencing data were analyzed to perform subtype clustering, identifying subtype‐specific marker genes using the FindAllMarkers function. Integration of three GEO datasets facilitated the differential expression analysis via the limma package. Single‐cell data are noisier due to technical dropouts and cellular heterogeneity; hence, a stricter cutoff (logFC > 1) was applied to avoid spurious hits. In contrast, bulk data are more stable with higher signal‐to‐noise ratios, so a more relaxed cutoff (log2|FC| > 0. 5) is appropriate to capture moderate but consistent differences. Four feature selection algorithms—LASSO, RF, SVM–RFE, and XGB—were employed to pinpoint hub genes followed by diagnostic modeling. ROC analysis validated the models’ efficacy. Finally, we investigated the relationship between hub genes and immune cell infiltration, and the signaling pathways associated with key genes were elucidated. Results We identified 493 cell subtype‐specific marker genes and 19 clusters within six major immune cell types. The integrated RNA‐seq analysis revealed 140 upregulated and 98 downregulated genes. The XGB diagnostic model outperformed others with an AUC of 0. 884. Five hub genes (MMP9, KLRB1, S100A8, S100A12, and MMP8) showed strong diagnostic potential and were significantly correlated with various immune cells. Pathway enrichment analysis linked these genes to crucial immunological pathways, and specific signaling pathways, such as AUTOIMMUNETHYROIDDISEASE and GRAFTVERSUSHOSTDISEASE, were associated with key genes. Conclusion This study provides a comprehensive cellular and molecular characterization of sepsis. The identified biomarkers and immune profiles from single‐cell data, coupled with robust diagnostic models, provide a foundation for further investigation to improve sepsis diagnosis and understand the immune evasion mechanisms. The correlations between hub genes and immune infiltration underscore their potential as targets for therapeutic intervention. The detailed pathway analyses offer insights into the systemic effects of sepsis and may guide personalized therapy development.