Bioinformatics combined with machine learning to identify early warning markers for severe dengue
Objective The goals of this study were to identify early warning markers of severe dengue based on bioinformatics com-bined with machine learning,and explore the evaluation system of the risk of occurrence of severe dengue.Methods Based on the Gene Expression Omnibus database,the differentially expressed genes between dengue and severe dengue were analyzed,and Gene Ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses were conducted.Early warning genes of severe dengue were screened using a random forest model,and the accuracy of the genes was verified using receiver operating characteristic(ROC)curves.Finally,nomograms were constructed to quantify the warning genes and predict the risk of progression from dengue to severe dengue based on the expression level of these genes.Results A total of 817 differentially expressed genes were identified,along with the associated biolo-gical processes that may be closely related to the occurrence and development of severe dengue,namely,antimicrobial humoral response,humoral immune response,serine hydrolase activity,and arachidonic acid metabolism.Based on this analysis,five early warning genes were isolated:AZU1,PDCD4,COL4A3BP,TRPM4,and ATP4A.Among these,ATP4A,COL4A3BP,and TRPM4 showed low expression levels,whereas AZU1and PDCD4were highly expressed.The ROC curves indicated that these genes were accurate pre-dictors of severe dengue.The nomograms indicated good predictive accuracy,clinical benefit rate,and clinical effectiveness of the model.Conclusion Measuring the expression levels of five warning genes(AZU1,PDCD4,COL4A3BP,TRPM4,and ATP4A)may help to evaluate the risk of severe dengue.
severe denguewarninggenebiological processrisk assessment
国家科技期刊平台
NSTL
万方数据
