Identification of transcriptomics biomarker for the early prediction of critically ill COVID-19 patients

Abstract

Abstract Objective Identifying the biological subsets of severe COVID-19 could provide a basis for finding biomarkers for the early prediction of the prognosis of severe COVID-19 and poor prognosis, and may facilitate specific treatment for COVID-19. Methods In this study we downloaded microarray dataset GSE172114 from the Gene Expression Omnibus (GEO) database in NCBI, and screened differentially-expressed genes (DEGs) by using the limma package in R software. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted, and the results were presented by volcano, Venn, heat, and enrichment pathway bubble maps in the R language package. Gene set enrichment analysis (GSEA) was used to explore and demonstrate the signal pathways related to severe COVID-19. Protein-Protein Interaction (PPI) Network analysis and visualization were performed by using STRING and Cytoscape. Seven key protein expression molecules were screened by the MOCDE plug-in. Then, the cytoHubba plug-in was used to screen 10 candidate genes with maximal clique centrality (MCC) algorithm as the standard, and the intersection with the Venn diagram was used to obtain seven Hub genes. Receiver operating characteristic (ROC) curves were drawn to determine the area under the curve (AUC), and the predictive value of the key genes was evaluated. Results A total of 210 DEGs were identified, including 186 upregulated genes as well as downregulated ones. GO enrichment and KEGG pathway analysis were used, and the results were presented by volcano, Venn, heat, and enrichment pathway bubble maps in the R language package. Gene set enrichment analysis (GSEA) was used to explore and demonstrate the signal pathways related to severe COVID-19. Protein interaction network (PPI) analysis and visualization were performed by using STRING and Cytoscape. Seven key protein expression molecules were screened by the MOCDE plug-in. Then, the cytoHubba plug-in was used to screen 10 candidate genes with maximal clique centrality (MCC) algorithm as the standard, and the intersection with the Venn diagram was used to obtain seven Hub genes. Receiver operating characteristic (ROC) curves were drawn to determine the area under the curve (AUC), and the predictive value of the key genes was evaluated. The AUC of the PLSCR1 gene was 0.879, which was the most significantly upregulated key gene in critically ill COVID-19 patients. Conclusions Based on bioinformatics analysis, we found that the screened candidate gene, PLSCR1, may be closely related to the occurrence of severe COVID-19, and can thus be used for the early prediction of patients with severe COVID-19, and may provide meaningful research direction for their treatment.