摘要：SummaryAsymptomatic infection is a big challenge in curbing the spread of COVID-19. However, its identification and pathogenesis elucidation remain issues. Here, by performing comprehensive lipidomic characterization of serum samples from 89 asymptomatic COVID-19 patients and 178 healthy controls, we screened out a panel of 15 key lipids that could accurately identify asymptomatic patients using a new ensemble learning model based on stacking strategy with a voting algorithm. This strategy provided a high accuracy of 96.0% with only 3.6% false positive rate and 4.8% false negative rate. More importantly, the unique lipid metabolic dysregulation was revealed, especially the enhanced synthesis of membrane phospholipids, altered sphingolipids homeostasis, and differential fatty acids metabolic pattern, implicating the specific host immune, inflammatory, and antiviral responses in asymptomatic COVID-19. This study provides a potential prediagnostic method for asymptomatic COVID-19 and molecular clues for the pathogenesis and therapy of this disease.Graphical abstractDisplay OmittedHighlights•Lipidomic profiling of asymptomatic COVID-19 serum was carried out•A panel of 15 serum lipids distinguished asymptomatic patients from healthy controls•A new ensemble learning model reduced the false negative rate for clinical diagnosis•Unique lipidomic dysregulation was identified for asymptomatic SARS-CoV-2 infectionBiological sciences; Immunology; Lipidomics
