586. Classification of patients with sepsis according to immune cell characteristics: a bioinformatic analysis of two cohort studies

Abstract

Background: Sepsis is well known to alter innate and adaptive immune responses for sustained periods after initiated by an invading pathogen. Identification the immune cell characteristics may shed light on the immune signature of patients with sepsis and further appropriate immune-modulatory therapy for distinct popula-ent immune endotypes via transcriptomics data analysis of previous published cohort studies.Methods: Datasets from two observational cohort studies that included 585 consecutive sepsis patients admitted to two intensive care units were downloaded as training cohort and external validation cohort. We analyzed genome-wide blood gene expression profiles from these patients by machine learning and bioinformatics.Results: The train cohort and the validation cohort had 479 and 106 patients re-spectively. Principal component analysis indicated that two immune sub-phenotypes for sepsis, designated immunoparalysis endotype and immunocompetent endotype could be distinguished clearly. In the train cohort, the worse prognosis was found in patients classified as immunoparalysis endotype and its hazard ratio is 2.32 (95% CI: 1.53 to 3.46 vs immunocompetent endotype). External validation furthermore demonstrates that present model could categorize sepsis into immunoparalysis and immunocompetent status precisely and efficiently. The percentage of 4 immune cells (Macrophages M0, Macrophages M2, B cells naïve, T cells CD4 naive) were found that associated with 28-day cumulative mortality significantly(P < 0.05).Conclusion: The present study developed a comprehensive tool to identify immunoparalysis endotype and immunocompetent status in sepsis be hospitalized and provides novel clues for further targeting of therapeutic approaches.