Host gene expression classifiers diagnose acute respiratory illness etiology

Area of Science:

• Biomarkers and Diagnostics
• Genomics and Infectious Disease
• Translational Medicine

Background:

• Acute respiratory infections (ARIs) are common, but current diagnostics often lead to inappropriate antibiotic prescriptions.
• Host response biomarkers present a novel diagnostic strategy to guide antimicrobial therapy.
• Distinguishing between bacterial, viral, and noninfectious causes of ARI remains a clinical challenge.

Purpose of the Study:

• To investigate if host gene expression patterns can differentiate noninfectious illnesses from infectious causes of ARI.
• To determine if gene expression can distinguish bacterial from viral etiologies of ARI.
• To develop and validate gene expression-based classifiers for ARI diagnosis.

Main Methods:

• An observational cohort study involving 273 patients with community-onset ARI or noninfectious illness and 44 healthy controls.
• Peripheral whole blood gene expression was measured using microarrays.
• Sparse logistic regression was employed to create classifiers for bacterial ARI, viral ARI, and noninfectious illness.

Main Results:

• Classifiers achieved 87% accuracy in distinguishing ARI causes, outperforming procalcitonin and existing classifiers.
• External validation in five datasets demonstrated high performance (AUC 0.90–0.99).
• A novel classification identified four distinct host response groups: bacterial ARI, viral ARI, coinfection, and non-infectious/non-microbial response.

Conclusions:

• Host gene expression classifiers demonstrate high accuracy in diagnosing the cause of acute respiratory illness.
• These classifiers offer a potential diagnostic platform to reduce inappropriate antibiotic use and mitigate antibiotic resistance.
• The findings support the clinical utility of host response gene expression profiling for infectious disease management.