A 5-transcript signature for discriminating viral and bacterial etiology in pediatric pneumonia

Pneumonia stands as the primary cause of death among children under five, with its onset attributed to a broad spectrum of microorganisms. Diagnosis poses an ongoing challenge, relying on clinical or microbiological criteria, often resulting in delayed and inaccurate treatment and unnecessary therapy. Our research focuses on identifying host transcriptomic biomarkers in the blood of children affected by viral and bacterial pneumonia, alongside healthy controls. The main goal is to establish a gene-expression signature enhancing disease diagnosis and management. We conducted an analysis of a total of 192 whole blood samples, comprising 38 controls and 154 viral and bacterial pneumonia patients recruited through the EUCLIDS clinical network. Our investigation identified 5,486 differentially expressed genes (DEGs) when comparing blood RNA from pneumonia patients with healthy controls. Functional enrichment analysis highlighted pathways related to the immune system response, encompassing neutrophil degranulation, humoral immune response, and various inflammatory pathways. In the comparative analysis of gene expression between viral and bacterial pneumonia patients, we identified 272 DEGs. Gene set analysis revealed a significant difference in pathway enrichment for the immune response, contingent on the over-regulation of gene sets in viral or bacterial pneumonias. Furthermore, we identified a 5-transcript host signature specifically designed to distinguish between viral and bacterial pediatric pneumonia (FAM20A, BAG3, TDRD9, MXRA7 and KLF14; AUC: 0.95 [0.88–1.00]), pseudo-validated in a cohort including probable bacterial and viral patients (AUC: 0.87 [0.77–0.97]). This signature holds the potential to enhance the accuracy of previously described general transcript-based signatures for viral and bacterial infections. Children with pneumonia (n = 154) and age and sex matched healthy controls (n = 38) were selected from the EUCLIDS database for transcriptome analysis in whole blood samples collected in PAXgeneTM and Tempus™ RNA tubes. In a subsequent analysis, pneumonia samples were further classified using the algorithm developed by Herberg et al. C-reactive protein (CRP) and neutrophil threshold criteria were employed where available in definitive bacterial (DB) and definitive viral (DV), respectively (see 30 for details on clinical characteristics of the cohort and definitions on DB and DV).Thus, DB (n = 40) and DV (n = 9) pneumonias were selected to study the differences in transcriptome depending on the etiology of the infection