A machine learning framework, built to deploy a quantitative sepsis response signature, can identify individuals with dysfunctional immune profiles, according to a study published in Science Translational Medicine. Eddie Cano-Gamez, PhD, and colleagues examined the use of whole-blood transcriptomics for stratification of patients with severe infection by integrating data from 3,149 samples from patients with sepsis due to community-acquired pneumonia or fecal peritonitis, as well as from healthy individuals, into a gene expression reference map. The
researchers used the gene expression reference map to derive a quantitative sepsis response signature (SRSq), which was reflective of immune dysfunction and predictive of clinical outcomes; these could be estimated using a seven- or 12-gene signature. A machine learning framework, SepstratifieR, was built to deploy SRSq in adult and pediatric bacterial and viral sepsis, H1N1 influenza, and COVID-19. The framework exhibited clinically relevant stratification across diseases and exposed certain physiological alterations that associate immune dysregulation with mortality. The method allowed early identification of individuals with dysfunctional immune profiles. “SepstratifieR enables stratification of patients with acute infection and can model their responses as a continuum,” Dr. Cano-Gamez and
colleagues wrote. “In combination with clinical biomarkers, SepstratifieR could improve risk estimation of immune dysfunction and clinical outcomes, as well as inform clinical trial design, bringing us closer to precision medicine for severe infection.”
