Back

Proteomic Biomarkers Differentiate Bacterial Infections in Febrile Infants: A Multicentre Prospective Study

Mills, C.; Drummond, H.; Karuna, N.; Mitchell, H.; McFetridge, L.; Rodgers, O.; Umana, E.; Groves, H. E.; Waterfield, T.

2026-02-05 infectious diseases
10.64898/2026.02.04.26345561 medRxiv
Show abstract

ObjectivesTo identify and validate plasma host-response protein biomarkers that improve discrimination of bacterial infection in febrile infants [&le;]90 days of age, and to assess whether novel biomarkers add value beyond established markers. MethodsSub-study of the prospective multicentre Febrile Infant Diagnostic Assessment and Outcome (FIDO) cohort. Novel biomarkers were identified through plasma proteomic profiling (Olink(R)) and combined with biomarkers and signatures from the literature for verification using Luminex and ELISA platforms. Diagnostic performance of novel biomarkers, established markers (CRP, PCT), and multi-protein signatures was evaluated. ResultsProteomic profiling of 110 samples identified 174 proteins differentially expressed between bacterial and viral infections, revealing distinct pathogen-specific immune signatures. Verification in the full cohort (n=445) demonstrated PCT had the highest individual accuracy for invasive bacterial infection (IBI) (AUC 0.89). Combining PCT with novel biomarkers, particularly lipocalin-2 (LCN2), improved discrimination (AUC 0.96). Diagnostic performance for the combined IBI/urinary tract infection (UTI) outcome was consistently lower (AUC <0.8). ConclusionsFebrile infants demonstrate biologically coherent host-response signatures that can be leveraged diagnostically. A PCT-LCN2 combination showed excellent accuracy for identifying IBI and may support future biomarker-guided diagnostic strategies, while reliable discrimination of UTI remains challenging.

Published in Journal of Infection (predicted rank #1) · training set

Matching journals

The top 10 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.