Advancing sepsis diagnosis and immunotherapy machine learning-driven identification of stable molecular biomarkers and therapeutic targets

Sepsis represents a significant global health challenge, necessitating early detection and effective treatment for improved outcomes. While traditional inflammatory markers facilitate the diagnosis of sepsis, the aspect of immune suppression remains poorly addressed. This study aimed to identify critical immune-related genes (IIRGs) associated with sepsis through genomic analysis and machine learning techniques, thereby enhancing diagnostic and treatment response predictions. Analyses of two extensive datasets were conducted, identifying significant immune genes using the ESTIMATE algorithm, Weighted Gene Correlation Network Analysis (WGCNA), and five machine learning methods. Prediction models were constructed and validated using six machine learning algorithms, achieving high accuracy (AUC > 0.75). Eleven key IIRGs were identified as active in immune pathways, such as the JAK-STAT signaling pathway, and were significantly correlated with immune cell infiltration in sepsis. Additionally, drug sensitivity analysis indicated that IIRGs correlated with responses to anticancer drugs. These results underscore the potential of these genes in enhancing sepsis diagnosis and treatment, highlighting the imperative for further validation across diverse populations.