Construction of a Novel Diagnostic Model Based on Ferroptosis-Related Genes for Hepatocellular Carcinoma Using Machine and Deep Learning Methods

Hepatocellular carcinoma (HCC) is one of the most general malignant tumors. Ferroptosis, a type of necrotic cell death that is oxidative and iron-dependent, has a strong correlation with the development of tumors and the progression of cancer. The present study was designed to identify potential diagnostic Ferroptosis-related genes (FRGs) using machine learning. From GEO datasets, two publicly available gene expression profiles (GSE65372 and GSE84402) from HCC and nontumor tissues were retrieved. The GSE65372 database was used to screen for FRGs with differential expression between HCC cases and nontumor specimens. Following this, a pathway enrichment analysis of FRGs was carried out. In order to locate potential biomarkers, an analysis using the support vector machine recursive feature elimination (SVM-RFE) model and the LASSO regression model were carried out. The levels of the novel biomarkers were validated further using data from the GSE84402 dataset and the TCGA datasets. In this study, 40 of 237 FRGs exhibited a dysregulated level between HCC specimens and nontumor specimens from GSE65372, including 27 increased and 13 decreased genes. The results of KEGG assays indicated that the 40 differential expressed FRGs were mainly enriched in the longevity regulating pathway, AMPK signaling pathway, the mTOR signaling pathway, and hepatocellular carcinoma. Subsequently, HSPB1, CDKN2A, LPIN1, MTDH, DCAF7, TRIM26, PIR, BCAT2, EZH2, and ADAMTS13 were identified as potential diagnostic biomarkers. ROC assays confirmed the diagnostic value of the new model. The expression of some FRGs among 11 FRGs was further confirmed by the GSE84402 dataset and TCGA datasets. Overall, our findings provided a novel diagnostic model using FRGs. Prior to its application in a clinical context, there is a need for additional research to evaluate the diagnostic value for HCC.