Integration of multi-omics and machine learning reveals sodium overload related molecular subtypes and biomarkers in sepsis

Sepsis is a life-threatening syndrome driven by a dysregulated host response to infection, and sodium dysregulation has recently emerged as a potential contributor to immune dysfunction and organ injury, yet its molecular signatures remain poorly characterized. In this study, we integrated multiple public transcriptomic datasets and employed weighted gene co‑expression network analysis, machine learning algorithms, consensus molecular subtyping, single‑cell RNA sequencing, and in vitro experiments in LPS‑stimulated THP‑1 cells to identify sodium overload-related candidate genes (SORGs) signatures and validate their functional roles. A five‑biomarker signature ( TXN , ADRB2 , DPP4 , MYC , SMAD3 ) was established, and a nomogram incorporating these genes achieved excellent diagnostic performance (AUC = 0.997). Two distinct molecular subtypes with divergent immune infiltration profiles were identified. Thioredoxin (TXN) , the most prominent hub gene, was significantly upregulated in sepsis and positively correlated with pro‑inflammatory immune cells. Knockdown of TXN attenuated LPS‑induced inflammatory cytokines (TNF‑α, IL‑1β, IL‑6), reduced intracellular sodium accumulation, and downregulated sodium transport‑related proteins (NHE1, NCX1, calpain), thereby mitigating sodium overload‑associated cell death (NECSO). Our findings suggest that TXN may represent a molecular feature associated with sodium-related cellular responses and immune dysregulation in sepsis, highlighting its potential value for further mechanistic investigation.