Host response biomarkers offer a promising alternative diagnostic solution for identifying acute respiratory infection (ARI) cases involving influenza infection. However, most of the published panels involve multiple genes, which is problematic in clinical settings because polymerase chain reaction (PCR)-based technology is the most widely used genomic technology in these settings, and it can only be used to measure a small number of targets. This study aimed to identify a single-gene biomarker with a high diagnostic accuracy by using integrated bioinformatics analysis with XGBoost. The gene expression profiles in dataset GSE68310 were used to construct a co-expression network using weighted correlation network analysis (WGCNA). Fourteen hub genes related to influenza infection (blue module) that were common to both the co-expression network and the protein-protein interaction network were identified. Thereafter, a single hub gene was selected using XGBoost, with feature selection conducted using recursive feature elimination with cross-validation (RFECV). The identified biomarker was oligoadenylate synthetases-like (OASL). The robustness of this biomarker was further examined using three external datasets. OASL expression profiling triggered by various infections was different enough to discriminate between influenza and non-influenza ARI infections. Thus, this study presented a workflow to identify a single-gene classifier across multiple datasets. Moreover, OASL was revealed as a biomarker that could identify influenza patients from among those with flu-like ARI. OASL has great potential for improving influenza diagnosis accuracy in ARI patients in the clinical setting.
Keywords: OASL; WGCNA; XGBoost; host response; influenza infection.
Copyright © 2020 Li, Liu, Xu, Wu, Zhang, Li, He, Yang, Liang and Ma.