Multi-platform omics analysis reveals molecular signature for COVID-19 pathogenesis, prognosis and drug target discovery

Abstract

Disease progression prediction and therapeutic drug target discovery for Coronavirus disease 2019 (COVID-19) are particularly important, as there is still no effective strategy for severe COVID-19 patient treatment. Herein, we performed multi-platform omics analysis of serial plasma and urine samples collected from patients during the course of COVID-19. Integrative analyses of these omics data revealed several potential therapeutic targets, such as ANXA1 and CLEC3B. Molecular changes in plasma indicated dysregulation of macrophage and suppression of T cell functions in severe patients compared to those in non-severe patients. Further, we chose 25 important molecular signatures as potential biomarkers for the prediction of disease severity. The prediction power was validated using corresponding urine samples and plasma samples from new COVID-19 patient cohort, with AUC reached to 0.904 and 0.988, respectively. In conclusion, our omics data proposed not only potential therapeutic targets, but also biomarkers for understanding the pathogenesis of severe COVID-19.

Fig. 1

Overview of samples for multi-omics study. a Multi-omics analysis design with three datasets. The training dataset combined with severe, non-severe, and healthy controls, proteins, lipids, and amino acids were quantified in plasma and used for biomarker discovery, using random forest. The validation cohort 1 contained ten plasma samples from from non-severe and five severe patients, 25 molecules were targeted quantified for prediction evaluation. The validation cohort 2 contained urine samples corresponding to plasma samples in the training dataset, and prediction precision was further evaluated using targeted quantification. b Sample information of COVID-19 patients in the training dataset with time annotation from onset of disease to admission or from admission to discharge

Fig. 2

Proteome profiling of COVID-19 patients. a Volcano plot of quantified proteins in COVID-19 vs healthy group, non-severe vs healthy group, severe vs healthy group, and severe vs non-severe group. b Heatmap of selected differential proteins expression levels and associated P values for COVID-19 patients annotated with functions and drug targets information. FC fold change

Fig. 3

Heatmap of lipids and amino acids related with COVID-19. a Heatmap of lipids expression levels and associated P values for COVID-19 patients. FC fold change. b Heatmap of amino acids expression levels and associated P values for COVID-19 patients. FC fold change

Fig. 4

Biomarker analysis based on multi-omics signatures. a ROC curve analysis for the predictive power of combined multiple omics signatures selected by random forest for distinguishing non-severe from severe group. b Principle component analysis for the non-severe and severe groups based on selected 25 signatures. c Normalized selected signatures expression values for each sample from individual non-severe patients or severe COVID-19 patients

Fig. 5

Validation performance in validation cohort 1 and validation cohort 2. a ROC curve analysis for the predictive power of validated lipid signatures in new plasma samples. b Performance of the model in new plasma cohort of ten COVID-19 patients. Samples classified into wrong group were labeled. c ROC curve analysis for the predictive power of validated lipid signatures in urine samples. d Performance of the model in urine cohort of ten COVID-19 patients. Samples classified into wrong group were labeled

Fig. 6

Drug target analysis by interaction among molecules. a Interaction between target protein-ANXA1 and other molecules include proteins, lipids, and amino acids. b Interaction between target protein-CLEC3B and other molecules include proteins, lipids, and amino acids