Advancing Sepsis Diagnosis and Immunotherapy Machine Learning-Driven Identification of Stable Molecular Biomarkers and Therapeutic Targets

Abstract This research presents a novel integrated approach combining genomic analysis and machine learning to identify biomarkers and drug sensitivities specific to sepsis, aiming to facilitate personalized treatment strategies. We comprehensively examined gene expression profiles from sepsis patients and healthy controls by utilizing the Gene Expression Omnibus (GEO) database, specifically datasets GSE154918 and GSE134347. Through the application of the ESTIMATE algorithm, weighted gene co-expression network analysis (WGCNA), and a range of machine learning techniques, we identified crucial Sepsis-Related Genes (SRGs), Immune-Related Differentially Expressed Genes (IRDEGs), and Important Immune-related genes (IIRGs). Our analysis revealed significant differences in immune and stromal scores between sepsis patients and controls, highlighting the altered immune landscape in sepsis. The study also uncovered specific genes associated with drug sensitivity, providing insights into potential therapeutic targets. The predictive model developed demonstrated high accuracy in sepsis diagnosis and prognosis, validated by independent datasets. These findings contribute to understanding sepsis at a molecular level and open new avenues for developing personalized therapeutic interventions, underscoring the potential of integrating genomic analysis and machine learning in sepsis research.