Identification and enrichment analysis of differentially expressed IRGs
The RNA-Seq data from TCGA consisted of 263 cases, including 241 male GC cases and 22 male non-tumour cases. Compared to male non-tumour tissues, a total of 276 differentially expressed IRGs including 87 downregulated and 189 upregulated were screened out, with the cut-off criteria of |log2 fold change (FC)| > 1 and false discovery rate (FDR) < 0.05 (Fig. 1a).
Additionally, to further the investigate potential functions of the differentially expressed IRGs in male GC patients, GO and KEGG analyses were performed in R software. As shown in Fig 1b, the results of GO enrichment analysis indicated that these IRGs can be significantly enriched in several important immune responses, including cell chemotaxis, leukocyte chemotaxis, myeloid leukocyte migration, leukocyte migration, granulocyte chemotaxis, neutrophil chemotaxis, granulocyte migration and neutrophil migration. KEGG analysis highlighted that the differentially expressed IRGs were mainly enriched in immune responses and tumour-related signalling pathways (Fig. 1c).
Identification of OS-related IRGs
A total of 20 differentially expressed IRGs were identified to be significantly associated with the OS of male GC patients based on the results of univariate Cox regression analysis (Fig. 2a). Then, by performing multivariate Cox regression analysis, a prognostic signature consisting of seven IRGs (LCN12, CCL21, RNASE2, CGB5, NRG4, AGTR1 and NPR3) was selected to construct a prediction model (Table. 1). All the seven IRGs were associated with high risk with hazard ratios (HRs) > 1. Among these IRGs, three genes (LCN12, RNASE2, and CGB5) were upregulated and four genes (CCL21, NRG4, AGTR1, and NPR3) were downregulated in male GC tissues compared to the normal tissues based on the TCGA dataset (Fig. 2b).
Establishment of the seven-IRGs risk signature
Subsequently, we constructed a prognostic model on the basis of the seven-IRGs. The risk score of each male GC patient was calculated as follows: risk score = expression level of LCN12 × 0.13933 + expression level of CCL21 × 0.00181 + expression level of RNASE2 × 0.01554 + expression level of CGB5 × 0.01036 + expression level of NRG4 × 0.34521 + expression level of AGTR1 × 0.14512 + expression level of NPR3 × 0.23201. The male GC patients were classified into high- and low-risk groups according to the median risk score. The distribution of risk scores and the survival status of male GC patients were displayed in Fig. 3a. In addition, the heatmap revealed the differentially expressed levels of the seven-IRGs in the high- and low-risk groups (Fig. 3b).
Next, the prognostic value of the risk score was evaluated. Univariate Cox regression analysis showed that the risk score (P < 0.001, HR = 1.268, 95%CI 1.151-1.397) was significantly correlated with the OS of male GC patients (Fig. 4a). Notably, as shown in Fig. 4b, the risk score could be an independent prognostic indicator (P < 0.001, HR = 1.288, 95%CI 1.167-1.422). The Kaplan–Meier curve demonstrated that male GC patients with high risk scores had a shorter survival time than those with low risk scores (log-rank P < 0.001, Fig. 4c). The areas under the curve (AUCs) for the risk score at 1-, 3- and 5-year in predicting OS were 0.73, 0.633 and 0.745, respectively (Fig. 4d). Moreover, compared to other clinical parameters, the risk score had the highest performance in the survival prediction of male GC patients (AUC = 0.712, Fig. 4e). Taken together, the above results indicated that the risk score performed well at predicting OS in the TCGA dataset.
Construction of the prognostic nomogram
To better predict the prognosis of male GC patients, we established a nomogram to predict the OS probability at 1-, 3- and 5-years (Fig. 5a). The variables of age, grade, stage and risk score were included in the prediction model. The C-index of the nomogram was 0.695 (95% CI: 0.632–0.759). As shown in Fig. 5b, the calibration plots demonstrated favourable agreement between predicted probabilities from the nomogram and the observed outcomes. Collectively, these results implied that the nomogram had good reliability in predicting survival for male GC patients.
External validation of the seven-IRGs prognostic signature
To further examine the performance of the seven IRGs-based model, the gene expression data and survival outcomes from GSE15460 were used for external validation. We calculated the risk score with the same formula for each male GC patient, and then divided them into high‐ and low‐risk groups according to the median risk score. Compared with patients in the low‐risk group, male GC patients with high risk scores suffered significantly more survival risks (Fig. 6a). Similar to the abovementioned findings, male GC patients in the high-risk group were associated with poor survival outcomes (log-rank P = 0.001, Fig. 6b). As presented in Fig. 6c, the AUCs of 1-, 3- and 5-year in predicting OS were 0.595, 0.621 and 0.657, respectively. Furthermore, the calibration plots indicated quite good agreement between prediction and observation for the 3- and 5-year OS probabilities of the patients (Fig. 6d).
Associations of the risk signature with tumour-infiltrating immune cells
To investigate distinct patterns of immune infiltration, we used CIBERSORT algorithm to estimate the composition of 22 infiltrating immune cells types in each male GC sample. As presented by the radar plot in Fig. 7a, the abundance of the 22 infiltrative immune cells was significantly different between high- and low-risk groups. Specifically, the infiltration levels of resting memory CD4 T cells (P = 0.034), activated NK cells (P = 0.003), regulatory T cells (Tregs) (P = 0.002), monocytes (P = 0.004) and resting mast cells (P = 0.008) were significantly higher in the high-risk group than in the low-risk group, whereas the infiltration levels of activated memory CD4 T cells (P < 0.001), resting NK cells (P < 0.001), follicular helper T cells (P = 0.009) and M1 macrophages (P = 0.006) were the opposite (Fig. 7b).