The Host Response Map: a Cutting-Edge AI/ML Approach to Characterizing Human Immune Response

B. Reddy, Jr.
Prenosis, Inc., Illinois, United States

Poster stand number: W125

Keywords: sepsis, host response, infection, ai, machine learning

Sepsis is life-threatening dysregulated host response to infection. To elucidate the complex underlying biology, Prenosis built the Host Response Map (HRM), a cutting edge AI/ML approach to characterizing host response based on the NOSIS dataset/biobank. NOSIS is one of the largest biobanks ever assembled for infection, including over 70,000 samples from over 13,000 patients. The HRM was constructed from biomarker and clinical data with unsupervised ML procedures. Patient location on the map corresponds to a patient-specific biological profile. Agglomeration of patients on the HRM indicates similarity of host response biology, suggesting the existence of host response endotypes. The map is highly prognostic, as demonstrated by the association of 8 endotypes of host response with clinical outcomes (Fig 1) of interest and a longitudinal study of 125 high-risk patients (Fig 2). Many scientific insights become apparent using the map, which corroborate biomarker relationships reported in the literature (Fig 3). The HRM can also be used to identify heterogeneity of treatment response (Fig 4). The Prenosis Host Response Map is a powerful platform for elucidating host response to reduce morbidity and mortality caused by sepsis and other life-threatening complications of infection to which warfighters in the field are exposed.