A Novel Ferroptosis-Related Gene Signature to Predict Prognosis of Esophageal Carcinoma

Abstract

Objective: This study aimed to develop a novel ferroptosis-related gene-based prognostic signature for esophageal carcinoma (ESCA).

Methods: The TCGA-ESCA gene expression profiles and corresponding clinical data were downloaded from the TCGA database. Ferroptosis-related genes were identified from the literature and public databases, which were intersected with the differentially expressed genes between ESCA and normal samples. After univariate Cox regression and random forest analyses, several ferroptosis-related feature genes were identified and used to construct a prognostic signature. Then, the prognostic value of the complex value and the correlation of the complex value with immune cell infiltration were analyzed. Moreover, function analysis, mutation analysis, and molecular docking on the ferroptosis-related feature genes were performed.

Results: Based on the TCGA dataset and ferroptosis pathway genes, 1929 ferroptosis-related genes were preliminarily selected. Following univariate Cox regression analysis and survival analysis, 14 genes were obtained. Then, random forest analysis identified 10 ferroptosis key genes. These 10 genes were used to construct a prognostic complex value. It was found that low complex value indicated better prognosis compared with high complex value. In different ESCA datasets, there were similar differences in the proportion of immune cell distribution between the high and low complex value groups. Furthermore, TNKS1BP1, AC019100.7, KRI1, BCAP31, and RP11-408E5.5 were significantly correlated with ESCA tumor location, lymph node metastasis, and age of patients. KRI1 had the highest mutation frequency. BCAP31 had the strongest binding ability with small molecules DB12830, DB05812, and DB07307.

Conclusion: We constructed a novel ferroptosis-related gene signature, which has the potential to predict patient survival and tumor-infiltrating immune cells of ESCA.

Figure 1

(a) Principal component analysis of the TCGA-ESCA dataset. (b) The Volcano plot of differentially expressed genes between ESCA and normal samples in TCGA. (c) Venn diagram of the intersection of differential genes and ferroptosis-related genes.

Figure 2

(a) The forest map of 44 prognostic genes. (b) The survival curves of 14 ferroptosis-related genes.

Figure 3

(a) MeanDecreaseGini scores of 14 potential ferroptosis factors. The blue node indicates key genes. (b) The expression level heatmap of ferroptosis key genes. (c) The ROC tumor prognosis curve of ferroptosis key genes.

Figure 4

(a) There were significant differences in PFS between groups with high- and low-complex values. (b) The risk distribution of each group. (c) The correlation heatmap between the key gene expression levels and high/low complex values. (d) The survival prognosis time distribution of each group.

Figure 5

(a, b) Univariate (a) and multivariate (b) Cox regression analysis results. (c) Nomogram and calibration curves for one-and three-year survival.

Figure 6

(a) The immune cell distribution differences in TCGA-ESCA, GSE161533, and GSE44021 datasets. (b) The correlation analysis between the complex value and the top 5 high proportions of immune cells.

Figure 7

(a) Bubble maps of GO BP and KEGG pathways enriched by ferroptosis key genes. (b) The top five pathways of TCGA-ESCA sample enrichment analyzed by GSEA. (c) The same BP obtained from clusterProfiler and GSEA were mainly classified into three categories.

Figure 8

TNKS1BP1, AC019100.7, KRI1, BCAP31, and RP11-408E5.5 were found to be significantly correlated with esophageal tumor central location, lymph node metastasis radiographic evidence, and age.

Figure 9

Conformation and interaction force analysis of BCAP31 with DB12830, DB05812, and DB07307. Cyan compounds are small molecules, yellow solid lines are hydrogen bonds, gray dotted lines are hydrophobic interactions, and blue amino acid residues that form hydrogen bonds with small molecules.