Integrated multi omics and machine learning reveal mitochondrial immunometabolic networks in sepsis associated encephalopathy

Sepsis-associated encephalopathy (SAE) is a major complication in intensive care units, characterized by diffuse brain dysfunction due to systemic inflammation. Despite advances in critical care medicine, SAE remains a key factor in poor patient outcomes, with its pathogenesis closely related to mitochondrial damage and the release of mitochondrial DNA (mtDNA). In this study, we integrated multiple transcriptomic and single-cell sequencing datasets to comprehensively analyze mitochondrial-associated differentially expressed genes (MitoDEGs) in SAE brain tissues. Using machine learning algorithms, we identified three core biomarkers (ALDH7A1, HOGA1, and AA467197). Functional enrichment analysis showed that the upregulated genes in SAE were mainly involved in immune and inflammatory responses, while the downregulated genes were associated with mitochondrial metabolism and vascular functions. Based on MitoDEGs, clinical subtype analysis shows that changes in mitochondrial function can effectively distinguish three sepsis subtypes (Cluster 1-3). Among these, Cluster 3 had worse prognosis due to enhanced mitochondrial function and activated inflammatory pathways. Immune microenvironment analysis revealed that MitoDEGs were closely associated with damage-associated molecular patterns (DAMPs) signaling and the expression of mitochondrial respiratory chain complexes. Experimental validation showed that exogenous mtDNA significantly increased the levels of inflammatory cytokines (TNF-α, IL-1β, and IL-6), thereby aggravating brain tissue pathological damage.