The DEG labeling method applies to normal and cancerous tissues. Differential expression of the 17 pyroptosis-related genes (DEGs) was examined in the GTEx and TCGA data from 375 normal and 32 tumor tissues. (all P < 0.01). There are 16 different genes(GSMDA, IL1B, BLRP7, NLRP2, SCAF11, IL18, CASP8, CASP5, CASP1, NLRP6, CASP6, CASP3, GSDMD, GSDMB, PYCARD, PRKACA) were down-regulated while. Sixteen other genes (NOD2, GSDMC, IL6, TNF, NLRP1, AIM2, NLRP3, NLRC4, GPX4, ELANE, CASP9, PJVK, GSDME, PLCG1, NOD1, TIRAP)(Fig1A). The gene expression heatmaps show the RNA levels of these genes; Green is low in expression level while red is high. The protein-protein interaction studies revealed that these pyroptosis-related genes are involved in another process. (Fig1B). We found that GSDMC, NLRC4, SCAF11, CASP8, ELANE, PLCG1, NOD1, An input parameter for the PPI analysis was used to determine the lowest allowable interaction score, which for this study was set at 0.9(Fig1C).shows the association network including all Pyroptosis related genes (red: positive correlations; blue: negative correlations).
Tumor categorization using the DEGs
the pyroptosis-related DEGs and their numerous GC subtypes were studied to examine the links between expression and the resulting GC types; We analyzed the comparative data of all 375 participants in the TCGA dataset using consensus clustering. We observed that when k = 2, intragroup correlations were most vital, and intergroup correlations were lowest, GC patients, which include 375 individuals, could be well divided into two groups according to the 31 different genes identified(Fig2A). A heatmap displays the gene expression profile as well as clinical characteristics such as tumor differentiation (G1-G3), age (67 or >67 years), and survival status (alive or dead). However, we observed no variation in clinical aspects between the two groups (Fig2B). Overall survival (OS) time was not significantly different between the two clusters(P = 0.41, Fig2C).
In the TCGA cohort, a prognostic gene model was developed.
One hundred seventy-five genotyped GC samples and all patient survival data were used to find a match. For the initial screening of survival-related genes, univariate Cox regression analysis was utilized. The six genes (IL6, ELANE, GSDME, TIRAP, PYCARD, and CASP3) that satisfied the P< 0.01 criterion were kept for additional investigation, and Three of them( IL6, ELANE, and GSDME) were linked to an elevated risk with HRs greater than one, but the other three genes (TIRAP, PYCARD, and CASP3) were linked to a lower risk with HRs less than one (Fig3A). LASSO Cox regression analysis and the optimum value were used to construct a 6-gene signature (Fig3B, C). The following formula was used to get the risk score: (0.060*IL6 exp.) + + (0.018*ELANE exp.) + (0.122* GSDME exp.) + (0.015*TIRAP exp.) + (0.175*PYCARD exp.) + (-0.126*CASP3 exp.) Based on the risk score algorithm's median score, A total of 375 patients were separated into two risk groups: low-risk and high-risk(Fig3D). Patients with various levels of risk were successfully divided into two groups by the principal component analysis (PCA)(Fig3E). The OS time for the low-risk and high-risk groups was significantly different(P 0.0063, Fig3F). The prognostic model's sensitivity and specificity were assessed using time-dependent receiver operating characteristic (ROC) analysis. The AUC for 2-year survival was 0.58, 4-year survival was 0.61, and for 6-year survival, it was 0.63. (Fig 3G).
In the TCGA cohort, a prognostic gene model was developed.
In total, 375 genome-wide cancer sample sets had matched patients with relevant survival data, who had all provided the requested samples. Survival-related genes were investigated with Cox regression analysis during the preliminary stage. The six genes (IL6, ELANE, GSDME, TIRAP, PYCARD, and CASP3) that satisfied the P< 0.02 criterion were kept for additional investigation, and four of them( IL6, ELANE, and GSDME) were linked to an elevated risk with HRs greater than one, but the other three genes (TIRAP, PYCARD, and CASP3) were linked to a lower risk with HRs less than one (Fig. 3A). The Cox regression approach employed the LASSO shrinkage and selection operator to develop a 6-gene signature (Fig 3B, C). The following formula was used to get the risk score: (0.060*IL6 exp.) + + (0.018*ELANE exp.) + (0.122* GSDME exp.) + (0.015*TIRAP exp.) + (0.175*PYCARD exp.) + (-0.126*CASP3 exp.) A median score generated by the risk score method was used to identify low- and high-risk groups for 375 individuals(Fig3D). Principal component analysis (PCA) organized patients into two distinct groups with differing risks(Fig3E). The OS time difference between the low-risk and high-risk groups was substantial(P 0.0063, Fig3F). The prognostic model's sensitivity and specificity were assessed using time-dependent ROC analysis. For two-year survival, the ROC curve (AUC) was 0.58, while for four-year survival, it was 0.61, and for six-year survival, it was 0.63. (Fig3G).
The risk signature is validated externally.
GC patients that are a part of the GEO (Genes and Expression Omnibus) cohort were used in the validation process (GSE62254). Before the further study, the "Scale" tool was used to standardize the gene expression data. The mean risk score of the TCGA cohort based on its median was Those who had a 273 person threshold were deemed low risk, while the 77 people who exceeded it were designated as high risk. Those in the low-risk group (on the left side of the dotted line) had more extended life periods and lower mortality rates than patients in the high-risk cohort (Fig. 4A). The PCA demonstrated enough separation between the two categories (Fig. 4B). Furthermore, Kaplan–Meier analysis revealed a statistically significant difference in survival rates between the low and high-risk groups (P = 0.0001, Fig. 4C). Our model has a high degree of predictive efficacy (AUC = 0.62 for 2-year survival, 0.63 for 4-year survival, and 0.63 for 6-year survival).
The risk model's independent prognostic value
We ran univariate and multivariable Cox regression analyses to determine if the gene signature model's risk score is likely to be used as a prognosis factor. The risk score was an independent factor in the univariate Cox regression analysis. In both the TCGA and GEO cohorts, poor survival was anticipated. (HR = 4.520, 95 % CI: 1.7873–11.432 and HR:6.000, 95 % CI: 2.316–15.544, Fig. 5A,B). After adjusting for additional confounding factors, the multivariate analysis revealed that the risk score was a prognostic predictor (HR = 2.213, 95% CI: 1.589–3.083 and HR: 1.947, 95 % CI: 1.391–2.726, Fig. 5C, D) for GC patients in both cohorts In addition, for the TCGA cohort, we developed a heatmap of clinical features(Fig5E.) They found that patients in the low- and high-risk categories differed greatly concerning age and survival status(P< 0.05).
The functional assessments are based on the risk model.
To investigate the variations in gene functions and pathways across the risk model subgroups Using the "limma" R package, we performed DEGs of differential expression using the FDR of 0.05 and the log2FC of 1 criterion. GO enrichment analysis and KEGG pathway analysis was conducted using these defined DEGs. The findings revealed that DEGs were mainly associated with immunological response, Inflammatory cell chemotaxis, and chemokine-mediated signaling pathways(Fig 6A-D).
Immune activity levels in different subgroups are compared.
Using the single-sample gene set enrichment approach, using both the TCGA and GEO cohorts, we evaluated the range of immune cell array scores. In the TCGA cohort, there was also activation in 13 immune-related pathways in low- and high-risk groups (ssGSEA) (Fig. 7A). The high-risk subgroup exhibited reduced levels of immune cell infiltration, particularly of CD8+ T cells, neutrophils, natural killer (NK) cells, T helper (Th) cells (Tfh, Th1, and Th2 cells), tumor-infiltrating lymphocytes (TILs), and regulatory T (Treg) cells. The other 12 immune pathways in the research were less active in the high-risk group than in the low-risk group, except the type-2 IFN response pathway—the TCGA cohort (Fig. 7B). Similar findings were reached when examining the immunological state of the GEO cohort. Furthermore, We discovered that DCs are enriched with iDCs and macrophages in the low-risk group, but type-2 IFN responses are significantly reduced(Fig7C, D).