Development of a Signature Based on Eight Metastatic-Related Genes for Prognosis of GC Patients

Abstract

Gastric cancer (GC) has been a common tumor type with high mortality. Distal metastasis is one of the main causes of death in GC patients, which is also related to poor prognosis. The mRNA profiles and clinical information of GC patients were downloaded from The Cancer Genome Atlas and Gene Expression Omnibus databases. Univariate Cox and LASSO Cox analyses were used to screen the optimal metastasis-related genes (MRGs) to establish a prognostic Risk Score model for GC patients. The nomogram was used to visualize the Risk Score and predict the 1-, 3-, 5-year survival rate. The immune cell infiltration was analyzed by CIBERSORT and the ratio of immune-stromal component was calculated by the ESTIMATE algorithm. A total of 142 differentially expressed genes were identified between metastatic and non-metastatic GC samples. The optimal 8 genes, comprising GAMT (guanidinoacetate N-methyltransferase), ABCB5 (ATP-binding cassette subfamily B member 5), ITIH3 (inter-alpha-trypsin inhibitor heavy chain 3), GDF3 (growth differentiation factor 3), VSTM2L (V-set and transmembrane domain-containing 2 like), CIDEA (cell death inducing DFFA like effector a), NPTX1 (neuronal pentraxin-1), and UMOD (uromodulin), were further screened to establish a prognostic Risk Score, which proved to be an independent prognostic factor. Patients in high-risk group had a poor prognosis. There were significant differences in the proportion of 11 tumor-infiltrating immune cells between high-risk and low-risk subgroups. In addition, the StromalScore, ImmuneScore, and ESTIMATEScore in high-risk group were higher than those in low-risk group, indicating that the tumor microenvironment of the high-risk group was more complex. A Risk Score model based on eight metastasis-related genes could clearly distinguish the prognosis of GC patients. The poor prognosis of patients with high-Risk Score might be associated with the complex tumor microenvironments.

Keywords: Gastric cancer; Immune; Metastasis-related genes; Prognosis; Risk Score.

Fig. 1

GC metastasis-related genes. A Volcano plot of DEGs. The horizontal axis represented the difference multiples after logarithmic conversion (Log₂FC) and the vertical axis represented -log₁₀ (FDR). Blue: down-regulated genes; red: up-regulated genes; B The top 20 enriched GO terms (Color figure online)

Fig. 2

The construction of the GC prognosis model. A Forest map of 20 genes associated with prognosis of GC by univariate analysis. The HR was presented with its 95% confidence intervals (95% CI). B The LASSO regression model determined the tuning parameter (Lambda). The smallest value of partial likelihood deviance was the optimal Lambda value. C Kaplan–Meier curve of GC patients in TCGA database. D Kaplan–Meier curve of GC patients in GEO database

Fig. 3

Stratified validation of the prognostic Risk Score model. The Risk Score was an independent prognostic indicator for GC patients. A Multivariate Cox regression analysis forest map. Compared with the reference sample, samples with HR > 1 represented a higher risk of death and HR < 1 represented a lower risk of death. B-G Kaplan–Meier survival curves of different groups. Different colors represented different groups. p value was calculated by log-rank test

Fig. 3

Stratified validation of the prognostic Risk Score model. The Risk Score was an independent prognostic indicator for GC patients. A Multivariate Cox regression analysis forest map. Compared with the reference sample, samples with HR > 1 represented a higher risk of death and HR < 1 represented a lower risk of death. B-G Kaplan–Meier survival curves of different groups. Different colors represented different groups. p value was calculated by log-rank test

Fig. 4

Nomogram could predict the survival probability of GC patients. A Nomogram for predicting the 1-, 3-, or 5-year overall survival time in GC patient. B-D Calibration curves of nomogram for predicting the 1-, 3-, or 5-year overall survival time in GC patients

Fig. 5

Immune cell infiltration in high- and low-risk GC patients. A The relative proportion of immune-infiltrating cells in all patients. B-L 11 types of immune cells with significantly different proportions of infiltration in the high- and low-risk groups. p-value was calculated by the Wilcoxon method

Fig. 5

Immune cell infiltration in high- and low-risk GC patients. A The relative proportion of immune-infiltrating cells in all patients. B-L 11 types of immune cells with significantly different proportions of infiltration in the high- and low-risk groups. p-value was calculated by the Wilcoxon method

Fig. 6

The StromalScore, ImmuneScore, and ESTIMATEScore of GC patients in high-risk and low-risk groups. A StromalScore. B ImmuneScore. C.ESTIMATEScore