Exploring the prognostic function of TMB-related prognostic signature in patients with colon cancer

Abstract

Tumor mutation burden (TMB) level is identified as a useful predictor in multiple tumors including colon adenocarcinoma (COAD). However, the function of TMB related genes has not been explored previously. In this study, we obtained patients' expression and clinical data from The Cancer Genome Atlas (TCGA) and the National Center for Biotechnology Information (NCBI). TMB genes were screened and subjected to differential expression analysis. Univariate Cox and LASSO analyses were utilized to construct the prognostic signature. The efficiency of the signature was tested by using a receiver operating characteristic (ROC) curve. A nomogram was further plotted to assess the overall survival (OS) time of patients with COAD. In addition, we compared the predictive performance of our signature with other four published signatures. Functional analyses indicated that patients in the low-risk group have obviously different enrichment of tumor related pathways and tumor infiltrating immune cells from that of high-risk patients. Our findings suggested that the ten genes' prognostic signature could exert undeniable prognostic functions in patients with COAD, which might provide significant clues for the development of personalized management of these patients.

Keywords: Colon cancer; Prognosis; Risk signature; Tumor mutation burden.

Fig. 1

Establishment of the risk signature. (A and B) 410 differently expressed genes were visualized by using a heatmap and a volcano map. (C and D) LASSO regression analysis was utilized for the construction of risk signature. (E) The correlation between genes expression value and risk score value was visualized, *P < 0.05

Fig. 2

Prognostic function of evaluation and inner validation. (A) Survival difference between low-risk and high-risk patients, ***P < 0.001. (B) Prediction accuracy of the risk signature. (C) Expression difference of ten genes between low-risk group and high-risk group. (D) Survival status difference between low-risk and high-risk patients. (E) Risk score value difference between alive and dead patients, ***P < 0.001. (F and G) Survival difference and its corresponding AUC value in inner validation set 1, **P < 0.01. (H and I) Survival difference and its corresponding AUC value in inner validation set 2, ***P < 0.001

Fig. 3

Clinical correlation and clinical characteristic specific survival analysis. (A-F) The correlation between risk score and patients’ clinical characteristics, *P < 0.05, **P < 0.01. (G-L) Survival difference between low-risk and high-risk group in patients with diverse clinical characteristics, *P < 0.05, **P < 0.01

Fig. 4

Independent prognostic function identification and survival time prediction. (A and B) Univariable Cox regression and multivariable Cox analyses were performed to screen indicator with prognostic function. (C) A nomogram was generated to predict patients’ overall survival time. (D and E) A calibration curve and a ROC curve were generated to evaluated the accuracy of the nomogram. (F) C-index was used to compared the prediction accuracy difference between TMB-gene signature and other four signature in COAD.

Fig. 5

Comparing of signature’ prognostic value and discovering of molecular functions and pathways. (A-E) The survival difference and its’ corresponding accuracy value were characterized by using Kaplan-Meier analysis and ROC curve, **P < 0.01, ***P < 0.001. (F and G) GSEA analysis was used to obtain the enrichment of pathways in low-risk group and high-risk group (www.kegg.jp/kegg/kegg1.html)

Fig. 6

Correlation between the signature and tumor dryness, genomic instability and drug sensitivity. (A) The signature is nor correlated with the DNAss of COAD patients. (B) The risk score of the signature is negatively correlated with the RNAss of COAD patients, ***P < 0.001. (C) The risk score of the signature is not correlated with the MSI of COAD patients. (D-I) Patients with low risk are more sensitive to Alpelisib, AZ960, BMS-754,807, Doramapimod, JAK_8517 and Taselisib, ***P < 0.001

Fig. 7

Immune cell infiltration difference characterized by the risk signature. (A) Immune cell infiltration was assessed and visualized with a heatmap. (B) The correlation between risk score value and immune cells infiltration status. (C-F) The infiltration difference of ‘Tell CD4 + Th1’, ‘cancer associated fibroblast’ ‘macrophage M1’ and ‘Monocyte’ between two groups were visualized, *P < 0.05, ***P < 0.001

Fig. 8

Protein expression validation of ten genes in clinical samples from online website dataset