An integrative viro-bacterial signature based on viral load and high-resolution microbiome profiling predicts COVID-19 mortality.
Journal
International Journal of Infectious Diseases
Journal Volume
161
Start Page
Article Number : 108153
ISSN
1878-3511
Date Issued
2025-12
Author(s)
Lai, Zi-Lun
Su, Yang-Di
Hsu, Yi-Yao
Hung, Yuan-Hua
Liu, Yuan-Yu
Tai, Yi-An
Lin, Hsiu-Hsien
Su, Hung-Chieh
Shih, Hong-Mo
Ho, Mao-Wang
Cho, Der-Yang
Abstract
Background The nasopharyngeal (NP) microbiota may play a critical role in modulating host immune responses during SARS-CoV-2 infection, yet its utility for predicting clinical outcomes is not fully defined. We aimed to determine if an integrative approach, combining NP microbial profiles with virological and clinical data, could improve mortality prediction in COVID-19 patients. Methods We analyzed nasopharyngeal swabs from 81 COVID-19 patients and 70 non-infected controls. Full-length 16S rRNA sequencing was used for microbiome profiling. Predictive models were developed using machine learning to integrate microbial taxa with viral load and other clinical metadata. Results High viral load was identified as the strongest independent predictor of mortality by multivariate logistic regression (OR: 3.80). SARS-CoV-2 infection and its associated viral load were linked to significant reductions in microbial community evenness. Critically, a machine-learning model that integrated viral load, patient age, and specific microbial taxa achieved a high predictive accuracy for mortality (AUC = 0.9046), significantly outperforming models based on clinical data alone. Conclusions Our findings demonstrate that an integrative approach, combining nasopharyngeal microbiota profiles with viral load, provides a robust framework for predicting mortality in COVID-19 patients. This strategy offers a promising, non-invasive tool for improving clinical risk stratification.
Subjects
COVID-19
Full-length 16S rRNA sequencing
Mortality prediction
Nasopharyngeal microbiota
Predictive modeling
Viral load
Publisher
Elsevier B.V.
Type
journal article