Metabolic signature of COVID-19 progression: potential prognostic markers for severity and outcome.

Abstract

Introduction: There are significant challenges remain in accurately categorizing the risk of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) patients.

Objectives: We used an untargeted 1H NMR-based metabolomics to assess the metabolomic changes in serum samples from a Danish cohort of 106 COVID-19-infected patients with mild to fatal disease courses and from patients with fatal outcomes from other diseases.

Methods: In total, 240 serum samples were used for this study. We used the data for multiple analyses (1) to construct a predictive model for disease severity and outcome, (2) to identify prognostic markers for subsequent disease severity and outcome, and (3) to understand the disease consequences in the metabolome and how recovery or death is reflected in the altered biological pathways.

Results: Our results revealed distinct alterations in the serum metabolome that could differentiate patients with COVID-19 by severity (mild or severe) or outcome (death or survival). Using receiver operating characteristic (ROC) curve analysis and four machine learning algorithms (random forest, linear support vector machine, PLS-DA, and logistic regression), we identified two biomarker sets with relevant biological functions that predict subsequent disease severity and patient outcome. The range of these severity-associated biomarkers was equally broad and included inflammatory markers, amino acids, fluid balance, ketone bodies, glycolysis-related metabolites, lipoprotein particles, and fatty acid levels.

Conclusions: Our data suggest the potential benefits of broader testing of these metabolites from newly diagnosed patients to predict which COVID-19 patients will progress to severe disease and which patients will manifest severe symptoms to minimize mortality.