Transcriptomic Biomarkers to Discriminate Bacterial from Nonbacterial Infection in Adults Hospitalized with Respiratory Illness
Abstract Lower respiratory tract infection (LRTI) commonly causes hospitalization in adults. Because bacterial diagnostic tests are not accurate, antibiotics are frequently prescribed. Peripheral blood gene expression to identify subjects with bacterial infection is a promising strategy. We evaluate...
Guardado en:
| Autores principales: | , , , , , , , , , , , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
Nature Portfolio
2017
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/ba1469c18a8948bdb99d2592622d6433 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | Abstract Lower respiratory tract infection (LRTI) commonly causes hospitalization in adults. Because bacterial diagnostic tests are not accurate, antibiotics are frequently prescribed. Peripheral blood gene expression to identify subjects with bacterial infection is a promising strategy. We evaluated whole blood profiling using RNASeq to discriminate infectious agents in adults with microbiologically defined LRTI. Hospitalized adults with LRTI symptoms were recruited. Clinical data and blood was collected, and comprehensive microbiologic testing performed. Gene expression was measured using RNASeq and qPCR. Genes discriminatory for bacterial infection were identified using the Bonferroni-corrected Wilcoxon test. Constrained logistic models to predict bacterial infection were fit using screened LASSO. We enrolled 94 subjects who were microbiologically classified; 53 as “non-bacterial” and 41 as “bacterial”. RNAseq and qPCR confirmed significant differences in mean expression for 10 genes previously identified as discriminatory for bacterial LRTI. A novel dimension reduction strategy selected three pathways (lymphocyte, α-linoleic acid metabolism, IGF regulation) including eleven genes as optimal markers for discriminating bacterial infection (naïve AUC = 0.94; nested CV-AUC = 0.86). Using these genes, we constructed a classifier for bacterial LRTI with 90% (79% CV) sensitivity and 83% (76% CV) specificity. This novel, pathway-based gene set displays promise as a method to distinguish bacterial from nonbacterial LRTI. |
|---|