Dysregulated host responses to infection can often result in sepsis – presenting a challenge for practitioners and causing many thousands of deaths and across the globe. Now, researchers at the University of Oxford, UK, have developed a new method to identify immune dysfunction and predict clinical outcomes in patients with acute infection – generating a personalized risk score for each patient in the process (1).
The authors’ previous work used patient subphenotypes to determine risk, which led to the development of two sepsis response signature (SRS) groups: SRS1 and SRS2. The first group consisted of an immunocompromised profile with increased risk of mortality. The second group had a immunocompetent profile with reduced mortality; however, compared with those who received a placebo, individuals in this latter group showed poorer survival when treated with corticosteroids (2).
Having completed this initial work, the authors recognized that there was no risk assessment tool for patients who do not qualify for a sepsis diagnosis. Using whole blood transcriptomics and a machine learning framework – called SepstratifieR, the team developed a new system to fill the gap.
SepstratifieR was trained on data from sepsis patients and healthy individuals and was constructed from three gene expression assay platforms. Rather than use strict SRS categorization, the machine learning model scored patients on a continuum, taking into account a new category – SRS3 – for participants who fall into a low severity cohort and are transcriptionally closer to good health. In short, the lower the resulting quantitative SRS (SRSq) score a patient received, the lower the risk of sepsis.
Notably, the SRS groups identified in the earlier research were already known to be dynamic – exhibiting change even during a single hospital stay; the newer findings demonstrate that SRSq also decreases over time along recovery, with larger decreases associated with better outcomes. The authors conclude that their method offers a new angle on immune dysfunction, “bringing us closer to precision medicine in infection.”
Credit: Images sourced from Unsplash.com
Newsletters
Receive the latest pathology news, personalities, education, and career development – weekly to your inbox.
References
- E Cano-Game et al., Sci Transl Med, 14 (2022). PMID: 36322631.
- D Antcliffe et al., Am J Respir Crit Care Med, 119, 990 (2019). PMID: 30365341.
