ABSTRACT Sepsis, a life‐threatening condition triggered by dysregulated host response to infection, poses significant global health challenges. Identifying lipopolysaccharide (LPS)‐related biomarkers and underlying mechanisms remains critical, yet underexplored. We integrated bulk microarray datasets and single‐cell RNA‐seq data from the Gene Expression Omnibus to identify LPS‐related genes associated with sepsis. scRNA‐seq was used for cell clustering, annotation, AUCell scoring, and cell‐cell communication analysis. Differentially expressed LRGs were screened from both bulk and single‐cell datasets and intersected. Three machine learning algorithms—least absolute shrinkage and selection operator regression, support vector machine–recursive feature elimination, and extreme gradient boosting—were applied to select robust diagnostic biomarkers. Gene expression was validated via qRT‐PCR. Diagnostic and prognostic models were constructed and validated in independent cohorts. Seven key LRGs were identified. The diagnostic model achieved high AUCs (> 0.89) across validation cohorts, while the prognostic model effectively stratified patients into distinct survival groups. High‐risk groups showed increased myeloid‐derived suppressor cell and macrophage infiltration, activation of inflammatory pathways, and unique intercellular communication networks. scRNA‐seq revealed cell‐type‐specific LRGs expression, particularly in myeloid populations. We established and validated a robust LPS‐related biomarker model that integrates bulk microarray and single‐cell transcriptomics, offering novel diagnostic, prognostic, and therapeutic insights for sepsis.
Zhang et al. (Fri,) studied this question.