Mining TCGA database for genes of prognostic value in gastric cancer microenvironment

Abstract

Gastric cancer (GC) is the sixth most common malignancy and the third leading cause of cancer-related death worldwide. Emerging evidence suggests that tumour microenvironment cells play a vital role in the development and prognosis of GC. To investigate the possible effect of stromal scores and immune scores on the overall survival (OS) on the GC patients, we divided GC patients into 'high' and 'low' groups based on their stromal and immune scores, and found differentially expressed genes related to prognosis of GC patients. Functional enrichment analysis and GSVA further revealed that focal adhesion and ECM-receptor interaction are associated with GC patients' survival. Finally, we analysed the effects of genes commonly involved in focal adhesion and ECM-receptor interaction on GC patients' survival and validated our results in another GC cohort from GEO data sets. In conclusion, we obtained a list of tumour microenvironment-related genes that predict poor prognosis in GC patients.

Keywords: estimate; gastric cancer; prognostic marker; tumour microenvironment.

FIGURE 1

The flow diagram of this study

FIGURE 2

Stromal scores are associated with GC stages and their overall survival. A, STAD cases were divided into two groups based on their stromal scores: the top 2/3 of 258 cases with higher stromal scores and the bottom 1/3 of 128 cases with lower stromal scores. As shown in the Kaplan‐Meier survival curve, median survival of the low score group is longer than the high score group (2100 d vs 782 d), as indicated by the log‐rank test; P‐value is .0119. B, Similarly, STAD cases were divided into two groups based on their immune scores: the 2/3 of 258 cases and the 1/3 half of 128 cases. The median survival of the low score group is longer than the high score group (1043 d vs 869 d); however, it is not statistically different as indicated by the log‐rank test; P = .4504. C, Similarly, STAD cases were divided into two groups based on their estimate scores: the 2/3 of 258 cases and the 1/3 half of 128 cases. The median survival of the low score group is longer than the high score group (1043 d vs 869 d); however, it is not statistically different as indicated by the log‐rank test; P = .1688. D, Correlation analysis of stromal scores and immune scores. E‐G, Distribution of stromal scores, immune scores and estimate scores in the four different GC stages. Dot‐plot shows that there is a significant association between GC stages and the level of stromal scores, immune scores and estimate scores, respectively (n = 406, P < .001)

FIGURE 3

Comparison of gene expression profile with stromal scores in GC. A‐C, GO analysis to explore the 3000 most different genes participate in molecular function (MF) (A), biological process (BP) (B) and cellular component (CC) (C). D, To explore the 3000 most different genes involved in signalling according to the Kyoto Encyclopedia of Genes and Genomes (KEGG) data sets. E, GSEA of the data of RNA‐sequencing, which show these KEGG signalling pathways are up‐regulated in the ‘stromal_high’ group

FIGURE 4

Correlation of expression of individual signal pathway in overall survival in TCGA. A, Correlation of the GSVA value of the 10 signalling pathways, which may take part in the poor clinical performance of the ‘stromal‐high’ group. B, Correlation between stromal scores and the GSVA values of the 10 signalling pathways, most of which are positive correlation, apart from TNF‐mediated signalling pathway. C, Survival analysis was performed on N = 350 patients obtained from the TCGA cohort of gastric cancer patients that had long‐term clinical follow‐up data. Displayed gene sets are downloaded from http://www.gsea‐msigdb.org, most of which are downloaded from KEGG and GO data sets; GSVA scores of each signalling pathway are performed using R package GSVA; for each signalling pathway, the top 1/2 of 175 cases with higher GSVA scores are ‘high’ group, and the bottom ½ of 175 cases with lower stromal scores are ‘low’ group

FIGURE 5

Correlation of expression of individual DEGs in overall survival in TCGA. Kaplan‐Meier survival curves were generated for selected DEGs extracted from the comparison of groups of high (red line) and low (blue line) gene expression. P < .05(A) or P < .1(B) in log‐rank test. OS, overall survival in days

FIGURE 6

Different expression levels of the marker genes in TCGA cohort. Expression levels of 18 genes in Figure 4 in 27 pairs of tumorous samples and patient‐matched normal samples in TCGA cohort

FIGURE 7

Validation of different expression levels of the marker genes in public data sets. Expression levels of 17 genes in 98 pairs of tumorous samples and patient‐matched normal samples in GEO cohort

FIGURE 8

Validation of correlation of DEGs extracted from TCGA database with overall survival in public data sets. Kaplan‐Meier survival curves were generated for selected DEGs extracted from the comparison of groups of high (red line) and low (blue line) gene expression