AUTHOR=Zhang Huipeng , Xu Nannan , Duishanbai Ahemala , Huang Gang , Zhang Jing , Bi Guanwei , Li Manyu , Wang Gang , Yu Yanbo 

TITLE=A two-transcript classifier model of host genes for discrimination of bacterial from viral infection in ulcerative colitis with opportunistic infections: a discovery and validation study

JOURNAL=Frontiers in Immunology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1642923

DOI=10.3389/fimmu.2025.1642923

ISSN=1664-3224

ABSTRACT=AimsWe aimed to develop and validate a classifier model to discriminate bacterial from viral infection in ulcerative colitis with opportunistic infections (UC-OI) by evaluating potential transcript signature in peripheral blood.MethodsThe study comprised UC patients with bacterial or viral infection or without opportunistic infections. We screened for differentially expressed genes associated with bacterial or viral infections (IFI44L, PI3 and ITGB2) and compared the expression levels of the genes in different infection subgroups. Subsequently, UC patients were randomly assigned (1:1) to either the discovery or validation groups. We developed a binary logistic regression model integrating the expression of candidate genes using discovery group and evaluated its discriminatory performance in validation group.ResultsThe expression levels of candidate genes differed significantly among infection subgroups. The IFI44L and PI3 combination was the most discriminatory and was used to construct the model. The two-transcript classifier model had an AUC of 0.867 (95% CI 0.794-0.941) to discriminate bacterial and viral infections in the validation group. Its performance was better than that of PCT, CRP and ESR and was less affected by pathogen type.ConclusionsIFI44L and PI3 transcript levels are robust classifiers to discriminate bacterial from viral infection in UC-OI, and measuring its levels appears to be predictive infection progression and treatment outcome in UC patients over time.