By examining the host response to SARS-CoV-2 infection we gain valuable insights into viral pathogenesis and COVID-19 progression. MicroRNAs (miRNAs), a class of small (18-22nt), non-coding RNAs, often play central roles in the host-pathogen interface and have been recognised as promising biomarkers of infectious disease. In this study, we profiled the circulating miRNAs from 10 longitudinally sampled COVID-19 patients and their age and gender matched controls. We found 55 differentially expressed miRNAs in early-stage disease, including miRNAs with known pro- and anti-inflammatory roles. Machine learning also identified a three-miRNA signature of COVID-19 that predicted infection with 99.9% accuracy. This signature faded away as the patients recovered. When this three-miRNA signature was applied to ferrets (a common model of respiratory infections, including COVID-19), the signature predicted SARS-CoV-2 infection with 99.8% accuracy and could distinguish between SARS-CoV-2, influenza (H1N1), and uninfected controls with >95% accuracy.
This study demonstrates that SARS-CoV-2 infection results in a significant host miRNA response that aligns with our current knowledge of COVID-19 induced inflammation. By using a multivariate, machine learning approach, we have developed a robust miRNA biomarker signature of COVID-19. This signature could complement existing diagnostic tests by providing a new approach to detecting cases that might otherwise be missed.