Machine Learning Identify Ferroptosis-Related Genes as Potential Diagnostic Biomarkers for Gastric Intestinal Metaplasia

Abstract

Background Gastric intestinal metaplasia(GIM) is an independent risk factor for GC, however, its pathogenesis is still unclear. Ferroptosis is a new type of programmed cell death, which may be involved in the process of GIM. The purpose of this study was to analyze the expression of ferroptosis-related genes (FRGs) in GIM tissues and to explore the relationship between ferroptosis and GIM. Method The results of GIM tissue full transcriptome sequencing were downloaded from Gene Expression Omnibus(GEO) database. R software (V4.2.0) and R packages were used for screening and enrichment analysis of differentially expressed genes(DEGs). The key genes were screened by least absolute shrinkage and selection operator(LASSO) and support vector machine-recursive feature elimination(SVM-RFE) algorithm. Receiver operating characteristic(ROC) curve was used to evaluate the diagnostic efficacy of key genes in GIM. Clinical samples were used to further validate hub genes. Results A total of 12 differentially expressed ferroptosis-related genes (DEFRGs) were identified. Using two machine learning algorithms, GOT1, ALDH3A2, ACSF2 and SESN2 were identified as key genes. The area under ROC curve (AUC) of GOT1, ALDH3A2, ACSF2 and SESN2 in the training set were 0.906, 0.955, 0.899 and 0.962 respectively, and the AUC in the verification set were 0.776, 0.676, 0.773 and 0.880, respectively. Clinical samples verified the differential expression of GOT1, ACSF2, and SESN2 in GIM. Conclusion We found that there was a significant correlation between ferroptosis and GIM. GOT1, ACSF2 and SESN2 can be used as diagnostic markers to effectively identify GIM.