Community Acquired Pneumonia - Disease Course Prediction with a 5-Gene Signature

Kirsten H, Weis S, Ahnert P, Witzenrath M, Scicluna BP, Krohn K, Rade M, Horn F, Bertram C, Reiche K, Löffler D, Blumert C, Jenner S, Sohn K, Nouailles G, Kiehntopf M, Creutz P, Rosolowski M, Loeffler M, Suttorp N, PROGRESS Study Group, Scholz M, Bauer M 2026 Community Acquired Pneumonia - Disease Course Prediction with a 5-Gene Signature CHEST Pulmonary , 100256.

Abstract

Background

Clinical decision-making for patients with community-acquired pneumonia (CAP) at risk of organ dysfunction and death is currently guided by clinical evaluation and scores. Every fifth CAP patient requires admission to intensive care (ICU) with a subsequent high mortality; delayed admission to ICU increases this risk.

Research Question

Can a transcriptomic signature improve identification of at-risk patients compared to conventional scores and metrics?

Study Design and Methods

Time-course transcriptomic data were obtained from blood samples taken from 455 participants in 41 centers enrolled into the PROGRESS trial, a prospective observational cohort study of hospitalized patients with CAP who did not initially require organ support. Discovery (n=240) and Validation (n=215) cohorts were randomly assigned. Transcriptome data were analyzed for association with a severe CAP course, defined as a composite of requirement for ICU admission or 28-day mortality. Predictive performance of the gene expression profiles was compared against clinical scores and serum markers, and validated in publicly-available transcriptomic datasets.

Results

A 5-gene signature consisting of SIGLEC14, TNFSF14, YOD1, CLEC4A, and KLRB1 (acronymically termed “STYCK”) was identified and validated to predict clinical deterioration with subsequent ICU admission or 28-day mortality (AUCDiscovery 0.82, 95%CI 0.71–0.90; p = 6.5x10-7; AUCValidation 0.81, 95%CI 0.70–0.90; p = 1.3x10-6). The signature outperformed clinical scores in predicting severe CAP and improved prediction when added to the SOFA score (AUCSOFA: 0.70 vs. AUCSOFA+STYCK: 0.83; p=0.0002). Prognostic value was confirmed in 6/17 publicly available sepsis cohorts, particularly those with low case fatality.

Interpretation

We identified a 5-gene transcriptomic signature that, taken soon after hospital admission predicts disease course of hospitalized CAP patients. “STYCK” was superior to conventional, currently used scores and clinical metrics in predicting this deterioration and improved, if added to SOFA, its performance.