Identification of Transcription Factor-Related Gene Signature and Risk Score Model for Colon Adenocarcinoma

Abstract

The prognosis of colon adenocarcinoma (COAD) remains poor. However, the specific and sensitive biomarkers for diagnosis and prognosis of COAD are absent. Transcription factors (TFs) are involved in many biological processes in cells. As the molecule of the signal pathway of the terminal effectors, TFs play important roles in tumorigenesis and development. A growing body of research suggests that aberrant TFs contribute to the development of COAD, as well as to its clinicopathological features and prognosis. In consequence, a few studies have investigated the relationship between the TF-related risk model and the prognosis of COAD. Therefore, in this article, we hope to develop a prognostic risk model based on TFs to predict the prognosis of patients with COAD. The mRNA transcription data and corresponding clinical data were downloaded from TCGA and GEO. Then, 141 differentially expressed genes, validated by the GEPIA2 database, were identified by differential expression analysis between normal and tumor samples. Univariate, multivariate and Lasso Cox regression analysis were performed to identify seven prognostic genes (E2F3, ETS2, HLF, HSF4, KLF4, MEIS2, and TCF7L1). The Kaplan-Meier curve and the receiver operating characteristic curve (ROC, 1-year AUC: 0.723, 3-year AUC: 0.775, 5-year AUC: 0.786) showed that our model could be used to predict the prognosis of patients with COAD. Multivariate Cox analysis also reported that the risk model is an independent prognostic factor of COAD. The external cohort (GSE17536 and GSE39582) was used to validate our risk model, which indicated that our risk model may be a reliable predictive model for COAD patients. Finally, based on the model and the clinicopathological factors, we constructed a nomogram with a C-index of 0.802. In conclusion, we emphasize the clinical significance of TFs in COAD and construct a prognostic model of TFs, which could provide a novel and reliable model for the prognosis of COAD.

Keywords: bioinformatics; colon adenocarcinoma; nomogram; risk score; transcription factors.

FIGURE 1

The workflow to construct the transcription factor (TF)-related risk model in COAD patients. TCGA, The Cancer Genome Atlas; COAD, colon adenocarcinoma; DEG, differentially expressed gene; GEPIA2, Gene Expression Profiling Interactive Analysis 2.

FIGURE 2

The results of differential gene analysis. (A) The volcano plot of differentially expressed TF-related genes based on TCGA. The sky blue points are the downregulated genes, the orange points are the upregulated genes, representing p < 0.05, |log₂FC| > 1. (B) The Venn plot to show the same differentially expressed TF-related genes between TCGA and GEPIA2 database. (C) The heatmap of differentially expressed TF-related genes. The vertical axis refers to genes, the horizontal axis refers to differences in gene expression between tissues, the orange means high expression, and the sky blue means low expression. (D) Gene Ontology (GO) circle graph of the top 10 GO terms with the most enriched genes. (E) Bar graph of the top 24 Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway with the most enriched genes; the vertical axis refers to names of pathways, and the horizontal axis refers to the number of genes.

FIGURE 3

Construction of prognostic model for COAD. (A) Hazard ratio of univariate Cox analysis for DEGs. (B) Survival analysis to verify the prognostic model. (C) ROC curve to evaluate the predictive efficacy of the risk model. (D) Distribution of risk scores of each COAD patient. (E) Correlation between survival time and survival status of each patient. (F) The expression pattern of seven TF-related genes. DEGs, differentially expressed genes. ROC, receiver operating characteristic.

FIGURE 4

Analysis of clinical independence for riskScore. (A) Univariable Cox regression analysis for clinical characters and risk score. (B) Multivariable Cox regression analysis for clinical characters and risk score. T, T stage; N, N stage; M, M stage; stage, TNM stage; riskScore, risk score model.

FIGURE 5

Validation of the risk model in the GEO dataset and nomogram for predicting survival rate in COAD patients. (A) GSE17536 and (B) GSE39582 dataset. (C) The nomogram for predicting the 1, 3, and 5-year survival rate by age, stage, riskScore. (D–F) The 1, 3, and 5-year calibration curves of TCGA dataset. (G–I) The 1, 3, and 5-year calibration curves of Gene Expression Omnibus (GEO) dataset.

FIGURE 6

Analysis of clinical and immunological relevance for riskScore. (A–D) Analysis of relationship between clinicopathological factors and riskScore. (E,F) Analysis of relationship between immune cells and riskScore.