Screening of mitochondrial-related biomarkers connected with immune infiltration for acute respiratory distress syndrome through WGCNA and machine learning

Septic acute respiratory distress syndrome (ARDS) is a complex and noteworthy type, but its molecular mechanism has not been fully elucidated. The aim is to explore specific biomarkers to diagnose sepsis-induced ARDS. Gene expression data of sepsis alone and sepsis-induced ARDS were downloaded from public databases, and the differential immune cells and differential expressed genes between the 2 groups were screened. Weighted gene co-expression network analysis was used to identify immune cells-related module genes, and then integrated with mitochondrial genes to obtain common genes. Next, least absolute shrinkage and selection operator, random forest, and support vector machine-recursive feature elimination were utilized to construct a nomogram model. Meanwhile, the biological function and targeted drugs of biomarkers were analyzed. The abundance of 3 immune cells (macrophage, neutrophils, and monocytes) was significantly different between the 2 groups. Weighted gene co-expression network analysis and machine learning identified 5 biomarkers were up-regulated in ARDS and had diagnostic significance. Next, the nomogram based on these genes had good confidence and clinical application value. Gene set enrichment analysis showed that phenylalanine metabolism pathway was increased in ARDS samples and had positive correlation with diagnostic genes. Drug prediction analysis exhibited that chlorzoxazone, ajmaline, and clindamycin could target multiple diagnostic genes. Overall, the diagnostic signature screened in this study can effectively predict the possibility of ARDS in sepsis patients, which can deepen the understanding of ARDS pathogenesis and targeted therapy development.