Single-Cell Sequencing Analysis Based on Public Databases for Constructing a Metastasis-Related Prognostic Model for Gastric Cancer

Abstract

Background: Although incidences of gastric cancer have decreased in recent years, the disease remains a significant danger to human health. Lack of early symptoms often leads to delayed diagnosis of gastric cancer, so that many patients miss the opportunity for surgery. Treatment for advanced gastric cancer is often limited. Immunotherapy, targeted therapy, and the mRNA vaccine have all emerged as potentially viable treatments for advanced gastric cancer. However, our understanding of the immune microenvironment of gastric cancer is far from sufficient; now is the time to explore this microenvironment.

Methods: In our study, using TCGA dataset and the GEO dataset GSE62254, we performed in-depth transcriptome and single-cell sequencing analyses based on public databases. We analyzed differential gene expressions of immune cells in metastatic and nonmetastatic gastric cancer and constructed a prognostic model of gastric cancer patients based on these differential gene expressions. We also screened candidate vaccine genes for gastric cancer.

Results: This prognostic model can accurately predict the prognosis of gastric cancer patients by dividing them into high-risk and low-risk groups. In addition to this, we identified a candidate vaccine gene for gastric cancer: PTPN6.

Conclusions: Our study could provide new ideas for the treatment of gastric cancer.

Figure 1

Single-cell sequencing data quality control. (a) The number of genes, percentage of mitochondrial genes, percentage of erythrocyte genes, and enrichment fraction of G2M and S phase of the cell cycle in metastatic and nonmetastatic gastric cancer cells. (b) Screening of 3000 highly variable genes. (c) Dimensionally reduced distribution of cell cycle conditions in cells.

Figure 2

Dimensionality reduction clustering and pseudotime series analysis. (a) Spatial distribution of metastatic and nonmetastatic gastric cancer cells. It can be seen that gastric cancer cells are well divided into metastatic and nonmetastatic clusters. (b) Expression of immune cell markers in 2 clusters. (c) Spatial distribution of cluster1 and cluster0. (d) Spatial distribution of immune cells and nonimmune cells. (e) Heat map of the distribution of the top 10 differential genes in the two clusters. (f) Pseudotime series analysis. Gastric cancer-immune cells differentiate from the deeper blue to the lighter blue. (g) There are 5 states of gastric cancer-immune cell differentiation. (h) The cells analyzed are all gastric cancer-immune cells. (i) The differentiation track of metastatic gastric cancer-immune cells and nonmetastatic gastric cancer-immune cells.

Figure 3

Construction of the prognostic model. (a) Univariate Cox analysis. (b). Multivariate Cox analysis. (c, d) The Lasso regression and tenfold cross-validation. Finally, a risk scoring formula composed of 8 genes was obtained. The risk value = APOD × 0.125 + MAP3K9 × (−0.245) + CD59 × 0.240 + TAP1 × (−0.135) + BNC2 × (−0.268) + CRAMP1L × (−0.327) + SMAD5 + COL6A3 × 0.123 × 0.449.

Figure 4

Evaluation of the model. (a, c, e) In TCGA dataset, the relationship between the model and the prognostic overall survival, the accuracy and stability of the model in predicting the prognosis of gastric cancer patients, and the comparison of the risk value and clinical characteristics of the model and the combination of the two indicators for prognosis evaluation. Similarly, (b, d, f) are the results in the GSE62254 dataset.

Figure 5

Risk curve and t-SNE dimension reduction analysis. (a, c, e) In TCGA dataset, we explored the distribution of model genes in the high- and low-risk groups and the survival of gastric cancer patients with increasing risk values. Similarly, (b, d, f) are the above results in the GSE62254 dataset. (g, h) The t-SNE dimension reduction analysis in TCGA and GSE62254 datasets, respectively.

Figure 6

Independent prognostic value analysis. (a, c) Univariate and multivariate Cox analyses in TCGA dataset. Risk score is an independent prognostic factor for gastric cancer patients (P < 0.01). (b, d) Univariate and multivariate COX analyses in the GSE62254 dataset. Risk score is an independent prognostic factor for gastric cancer patients (P < 0.01).

Figure 7

Gene enrichment analysis. (a, b) The GO functional enrichment analysis. Differentially expressed genes were mainly related to transmembrane transport across the cell membrane and receptor exchange. (c, d) The KEGG pathway enrichment analysis. They were mainly enriched in the calcium signaling, cAMP signaling, ECM-receptor interaction, focal adhesion, and Wnt signaling pathways.

Figure 8

Exploration of immune microenvironment. (a) Distribution of immune cells with significantly different infiltration levels in the high- and low-risk groups in TCGA dataset (P < 0.05). (b, c) The expression of immune checkpoint genes in TCGA and GSE62254 datasets, respectively. (d, e) The expression of immunogenic cell death modulators (ICDs) genes in two datasets (^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001).

Figure 9

Mutation correlation analysis. (a, b) Mutation landscape of the top 30 genes in the high-risk group and mutation landscape of the 8 genes in the model. (c, d) The map of mutation landscape in the low-risk group and gene mutation in the model. (e, f) The relationship between gene TTN and overall survival in high- and low-risk groups. (g, h) The relationship between gene FAT3 and overall survival in the high- and low-risk groups. (i, j) The mutation symbiosis diagram of genes in the model in the high- and low-risk group.

Figure 10

Candidate vaccine genes and therapeutic drugs for gastric cancer. (a) Distribution of genes with copy number variation in gastric cancer. (b) The distribution of different genes on chromosomes between gastric cancer and normal tissues, where red is the upregulated gene in gastric cancer and green is the downregulated gene. (c) The survival curve of PTPN6 in gastric cancer. (d) The correlation analysis between PTPN6 and tumor purity, antigen-presenting cells, and B cells. (e–j) The IC50 value of different chemotherapy drugs in the high- and low-risk groups in TCGA dataset (^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001).

Figure 11

The construction of a Nomogram. (a) We constructed the Nomogram based on the risk value and clinical characteristics of the gastric cancer patient model, and predicted the 1, 3 and 5 year mortality of the patient TCGA-HU-A4Gy. (b) calibration curves of nomogram for prediction of prognosis at 1, 3 and 5 years in patients with gastric cancer