Multi-omics machine learning identifies diagnostic gene signatures and functionally supports PRKACB involvement in macrophage inflammatory responses in sepsis.

Sepsis is a life-threatening condition caused by a dysregulated immune response, often leading to organ failure and death. Diagnosis and therapy remain challenging. This study aimed to identify biomarkers for sepsis through multi-omics analysis and experimental validation. A total of 1,166 samples from the GEO repository underwent differential analysis, WGCNA, and logistic regression to identify sepsis-associated features. After SVM-RFE screening, a 28-gene signature distinguishing sepsis from healthy controls achieved an AUC of 0.970 (sensitivity 0.939, specificity 1.000) in an independent cohort and 0.870 (sensitivity 0.906, specificity 0.700) in qRT-PCR validation. A 13-gene signature distinguishing sepsis from SIRS achieved an AUC of 1.000 (sensitivity 1.000, specificity 1.000) and 0.745 (sensitivity 0.500, specificity 0.900), respectively. PRKACB, consistently decreased in sepsis, emerged as a myeloid-associated hub gene with functional support in macrophages. Single-cell analysis revealed that PRKACB is predominantly expressed in myeloid cells, with macrophages and neutrophils showing the most notable compositional changes. Pathway enrichment indicated distinct PRKACB-related functions between sepsis and healthy controls, and across myeloid subtypes. Reducing PRKACB expression significantly upregulated TNF-α and IL-1β release and decreased THP-1 macrophage viability, suggesting that excessive inflammation may trigger apoptosis or metabolic dysfunction. These data support PRKACB as a myeloid hub linked to macrophage inflammatory output and viability, while defining the precise downstream signaling as a priority for future studies.