A host-based two-gene model for the identification of bacterial infection in general clinical settings.

OBJECTIVES: In this study, we aimed to develop a simple gene model to identify bacterial infection, which can be implemented in general clinical settings. METHODS: We used a clinically availablereal-time quantitative polymerase chain reaction platform to conduct focused gene expression assays on clinical blood samples. Samples were collected from 2 tertiary hospitals. RESULTS: We found that the 8 candidate genes for bacterial infection were significantly dysregulated in bacterial infection and displayed good performance in group classification, whereas the 2 genes for viral infection displayed poor performance. A two-gene model (S100A12 and CD177) displayed 93.0% sensitivity and 93.7% specificity in the modeling stage. In the independent validation stage, 87.8% sensitivity and 96.6% specificity were achieved in one set of case-control groups, and 93.6% sensitivity and 97.1% specificity in another set. CONCLUSIONS: We have validated the signature genes for bacterial infection and developed a two-gene model to identify bacterial infection in general clinical settings.