To identify immunophenotypes in sepsis pneumonia based on transcriptome data analysis.

Datasets from observational cohort studies in GEO public database that included consecutive sepsis patients admitted to intensive care units were downloaded. We analyzed genome-wide gene expression profiles in blood from sepsis patients by using machine learning and bioinformatics.

A total of 479 sepsis patients, including 183 with septic pneumonia, were enrolled. The 28-day cumulative mortality of sepsis patients was linked with the expression of 257 immunological molecules (P<0.05). Patients with pneumonia-mediated sepsis had a 28-day cumulative death rate that was connected with the expression of 87 immune molecules (P<0.05), of which 31 immune molecules had a substantially higher correlation (P<0.01). Cluster A (286 instances of sepsis, 117 cases of septic pneumonia) and Cluster B, two immunophenotypes, were found (193 cases of sepsis, 66 cases of septic pneumonia). Sepsis patients in the Cluster B group had a higher 28-day cumulative death rate than those in the Cluster A group [HR 3.173 95%CI (2.117, 4.457), P<0.001]. The 28-day cumulative mortality of Cluster B in septic pneumonia was higher than that of Cluster A [HR 3.523 95%CI (1.699, 7.035)], P<0.001].

The present study developed a comprehensive tool to identify the immunoparalysis endotype and immunocompetent status in hospitalized patients with sepsis and provides novel clues for further targeting of therapeutic approaches.