The datasets on STAD, which contain 32 normal and 375 tumor samples, were queried in the Cancer Genome Atlas (TCGA) for the purpose of extracting information on gene expression and clinical annotations. The GSE84437 cohort of the Gene Expression Omnibus (GEO) database had 433 tumor samples. Batch normalization was implemented using the R package “sva” (20). We also collected data from the TCGA platform on single nucleotide variation (SNV), transcriptome profiles, copy number variation (CNV), methylation, and clinical characteristics of pan-cancer transcriptomes. In addition, 4594 nucleotide metabolism-related genes and 271 neuroendocrine-related genes (relevance score>2.5) were obtained from the GeneCard Database (https://www.genecards.org/). After taking the intersection of the two groups of genes, 196 NMNGs were obtained and displayed by Venn diagram. The TCGA and GEO cohorts were screened for NMNGs with prognostic values using univariate cox regression for subsequent analysis. The prognostic significance of them was also validated using Kaplan-Meier (KM) analysis based on 808 STAD samples.

There has been minimal investigation on the relationship between nucleotide metabolism, neuroendocrine, and cancers yet. Consequently, the distinctions between NMNGs in diverse cancers are inadequately documented. SNV, CNV, methylation, and mRNA expression data were analyzed and visually displayed as heatmaps to offer an overview of NMNGs across all cancer types. A univariate Cox regression analysis between mRNA expression and cumulative survival was carried out to further assess the involvement of NMNGs in the prognosis of various cancers. The potential impact of NMNG on the traditional oncogenic pathway of pan-cancer was then investigated using the single sample gene set enrichment analysis (ssGSEA) (21). All of these analyses were carried out using R as in earlier published studies (22, 23).

Unsupervised clustering analysis was done to identify distinctive NMNG modification patterns and divided STAD samples into different categories based on the expression of prognostic NMNGs. A consensus clustering procedure was conducted to determine the number and stability of each individual cluster, and the “Consensus Cluster Plus” software package provided additional support (24). The clustering effect indicated that the clustering stability was improved with k=2.

The “GSVA” R package was used to conduct GSVA enrichment analysis so as to investigate differences in biological processes across various NMNG modification patterns. We used ssGSEA to evaluate a large number of immune cells (25), such as activated dendritic cells, activated CD4 T cells, activated CD8 T cells, and activated B cells, that have infiltrated the TME of STAD. According to the results of the ssGSEA analysis, the relative abundance of each TME-invading cell for every sample was represented by the enrichment scores.

We filtered out DEGs using the R package “limma” in accordance with two distinct NMNG modification patterns that we have identified. An adjusted P-value < 0.001 was used for the rank criterion. Additionally, we fundamentally annotated DEGs using the R package “clusterProfiler”. Similar to this, the univariate Cox model chose the survival-related DEGs of the two distinct NMNG modification patterns, and cluster analysis based on the DEGs with prognostic significance was carried out with the aid of the R package “Consensus Cluster Plus.” Furthermore, differences (variations) in NMNG expression and patient survival between clusters were characterized.

On the basis of the results, we subsequently created a scoring system for the NMNG. According to relevant research (26), we conducted PCA of the principal components and chose principal components 1 and 2 as feature scores. Then, the NMNG score was computed using the following equation: score = (PCli+PC2i), where ‘i’ stands for the genes related to the NMNG phenotype.

We used the Wilcoxon test to compare the differential expression of immunological checkpoints such as PDCD1, PDCD1LG2, CD40, CD80, and CD276 in groups with low and high NMNG scores. From the Cancer Immunome Atlas Database, we concurrently gathered the Immune Checkpoint Inhibitor (ICI) Immunophenoscore (IPS) file. The immunotherapeutic relevance of the NMNG gene signature was evaluated using IPS, a reliable tool for assessing tumor immunogenicity.

For all statistical analyses, R was utilized. The one-way ANOVA and Kruskal-Wallis test were used to compare the outcomes across multiple groups. The best cutoff score was divided into groups with low- and high-NMNG scores using the “surv-cutpoint” function. The prognostic survival curve was produced using the KM approach. We evaluated the mutation status of people with low- and high-NMNGscore subtypes using the waterfall function in the maftools package (25). For all analyses, the significance of the correlation is determined by the criterion of p < 0.05.

Figure 1 displays a flowchart outlining the steps in the research process. First, we did the intersection of genes related to neuroendocrine function and genes related to nucleotide metabolism to obtain 196 NMNGs ( Supplementary Figure 1 ). The 808 STAD samples were then subjected to a univariate cox regression analysis to yield 46 NMNGs that were prognostically significant for this study ( Supplementary Table S1 ). The prognostic significance of 46 NMNGs was validated by KM analysis based on 808 STAD samples ( Supplementary Figure 2 ).

Data on CNV, SNV, methylation, mRNA expression profiles, and survival information for 46 NMNGs in various malignancies were provided by TCGA for the pan-cancer study. The percentage of CNV was examined in order to investigate the genetic aberrations of NMNGs in cancer. Different CNV profiles of NMNGs were detected in varied cancer types ( Figure 2A ). For instance, in nearly all tumors, GNAS, IGFBP3, EGFR, STX1A, and KCNB1 were more likely to experience copy number gain than copy number loss, whereas PGR and CDH23 displayed the opposite profile. Moreover, we examined SNV data related to NMNGs to determine the frequency and types of variants in each cancer subtype. Figure 2B demonstrates the remarkable SNV of NMNGs in SKCM, STAD, UCEC, LUSC, and LUAD. In addition, 91.71% of the NMNGs had SNV frequency (5343 of 5826 tumors). Missense mutations were discovered to be the primary SNP type as per the variant-type analysis. According to SNV percentage analysis, the top 4 mutated genes were TP53, BRCA2, CDH23, and EGFR, with respective mutation percentages of 67%, 8%, 7%, and 7%. ( Figure 2C ). In addition to CNV, promoter methylation can control gene expression, and abnormal promoter DNA methylation is linked to cancer (27). We noted that most NMNG consistently exhibited hypomethylation in 20 cancer types, with the exception of CDH23, ADCYAP1, and UCHL1 ( Figure 2D ).

For every cancer type, differential expression analysis was done to look into changes in the gene expression patterns of NMNGs in addition to genetic changes between the tumor and nearby normal tissues. Most of the gene expression levels in cancer tissues were different from those in healthy tissues, we discovered, except in pancreatic cancer tissues. In cancer tissue, most NMNGs are expressed at high levels ( Figure 2E ). Univariate Cox regression of mRNA expression and OS was then used to identify risky NMNGs with HR > 1 and pValue < 0.05 as well as protective NMNGs with HR < 1 and pValue < 0.05, as shown in Figure 2F . We found that most genes are risk factors for multiple cancer types. Since it is currently unknown how nucleotide metabolism and neuroendocrine regulates pathways connected to cancer, it is imperative to examine any possible connections between these pathways and NMNGs. This will lay the basis for future investigation into how NMNGs control pan-cancer-related pathway regulation. According to our findings, NMNGs in pan-cancer were significantly correlated with a number of signaling pathways, including TNFA signaling via NFKB, KRAS signaling, interferon-gamma response, inflammatory response, and epithelial-mesenchymal transition (EMT) ( Figure 2G ).

The function of 46 NMNGs in STAD is the subject of our upcoming discussion. In the beginning, 46 NMNGs in STAD were examined for CNV and somatic mutations ( Figure 3A ). In 278 out of 433 samples (64.2%), NMNG mutations were found. The TP53 gene had the highest rate of mutation (44%), followed by CDH23 (9%). However, CNV deletions are more likely in EZH2, CDKN2B, and GRP, CNV amplification is more common in IGF1R, KCNB1, and GNAS ( Figure 3B ). The chromosomal positions of NMNG with different copy numbers are shown in Figure 3C . In order to ascertain the links between STAD and the expression of NMNG, we also evaluated the mRNA expression levels of 46 NMNGs in both tumor and normal samples. The results showed that 46 NMNGs had different expression patterns. LEPR, AR, PGR, KIT, and VIP were down-regulated in tumor tissues, whereas TP53, GNAS, UCP3, BRCA2, and BRCA1 were up-regulated in tumor tissues ( Figure 3D ).

The NMNG network outlines the co-expressed relationship among NMNGs with prognostic significances in STAD patients. ( Figure 4A ). The Consensus Cluster Plus R program was utilized to classify patients with various NMNG patterns by quantifying the expression of the 46 NMNGs. Unsupervised clustering was then used to identify two distinct modification patterns ( Figure 4B and Supplementary Figure 3 ). These two gene clusters were obviously divided ( Figure 4C ). In a subsequent prognostic analysis, it was found that cluster B modification pattern was found to have a better prognosis ( Figure 4D ). Figure 4E illustrated the clinical characteristics of the two subgroups of patients. Two NMNG modification patterns were examined using a KEGG enrichment assay to determine their biological roles. It was found that cluster A was significantly enriched in pathways that promote carcinogenesis, including the MAPK signaling pathway, the TGF BETA signaling pathway, the calcium signaling pathway, and the interaction between ECM receptors. DNA replication, mismatch repair, spliceosomes, and base excision repair pathways were enriched in cluster B ( Figure 4F ). We also examined the TME cell invasion of various clusters. According to the analysis, immune cells such as activated B cells, CD8+ T cells, activated dendritic cells, eosinophils, immature B cells, immature dendritic cells, macrophages, mast cells, NK cells, plasmacytoid dendritic cells, regulatory T (Treg) cells, follicular helper T (Tfh) cells, and type 1 helper T (Th1) cells were significantly enriched in cluster A ( Figure 4G ).

In order to further analyze the molecular mechanisms behind the two NMNG modification patterns, we used the limma package to find 6922 DEGs related to the NMNG phenotype. Figures 5A, B display the findings of the analysis of DEG-related enriched GO and KEGG pathways. We used univariate Cox regression analysis on 6922 DEGs to identify 2117 DEGs that affected prognosis for the purpose of further verifying the regulatory mechanism of NMNG modification in STAD. Patients were then divided into three gene clusters based on the results of the unsupervised clustering analysis of these genes ( Figure 6A and Supplementary Figure 4 ). This implies that there are different NMNG patterns in STAD. The prognosis for cluster C (141 patients) and cluster B (301 patients) patients among the 803 STAD patients was worse ( Figure 6B ). In Figure 6C , the various clinicopathological traits of these subgroups are depicted. We found significant variations in NMNG expression across these three clusters ( Figure 6D ).

These procedures are unable to anticipate the NMNG alteration patterns of specific patients because they only consider STAD patient populations. To assess the pattern of NMNG alterations in each STAD patient, we developed a scoring system, the NMNG score. To see the alterations in several patient parameters, alluvial plots were employed ( Figure 7A ). Then, using the NMNGscore as the parameter, we investigated patient survival results. High NMNGscore patients have greater survival rates ( Figure 7B ). Activated CD4+ T cells, neutrophilia, and Th17 cells were positively correlated with NMNGscore, whereas activated B cells, Eosinophilia, T cells, immature B and dendritic cells, MDSCs, macrophages, mast cells, NKT cells, NK cells, plasmacytoid dendritic cells, Treg cells, Tfh cells, and Th1 cells were negatively correlated ( Figure 7C ). The effect of different ICG expression levels on the TME was subsequently examined. In the group with low NMNGscores, we discovered increased expression levels of ICG, which might have been linked to their worse prognosis ( Figures 7D–H ). Then, we conducted research on the variation in NMNGscore between NMNGclusters and gene clusters. NMNGcluster B and gene cluster B both had significantly higher NMNGscores than NMNFcluster A and gene clusters A and C, accordingly ( Figures 7I, J ). These findings suggested that the NMNGscore can be used to evaluate the pattern of NMNG alteration as well as the tumor’s TME immune cell infiltration features in a particular patient.

Somatic mutations in tumor genomes are related to patients’ response to immunotherapy, according to our studies. As a result, we focused on the distributions of tumor mutation burden (TMB) across distinct NMNGscore categories. The low NMNGscore group had a lower TMB in comparison to the group with a high NMNGscore, and the NMNGscore was positively linked to TMB ( Figures 8A, B ). Furthermore, there was a markedly higher chance of survival in the group of patients with high mutational loads, and this benefit was much more obvious in those patients who also had high NMNGscores ( Figures 8C, D ). The distribution landscape of somatic mutations was then compared between the groups with high and low NMNGscores using the R maftools package. The results showed that the group with a high NMNGscore had a greater TMB than the group with a low NMNGscore ( Figures 8E, F ). Overall, NMNG changes interact with somatic mutations, and NMNGscore categorization may be influenced by chromosomal variance.

We assessed whether NMNG alterations could predict patient responses to ICIs in the context of the recent approval of drugs that target PD-1 and CTLA-4 for the treatment of various types of cancers. We observed that patients with a high NMNGscore had significant clinical benefits and considerably improved survival status ( Figures 9A, B ). In order to anticipate how efficiently ICIs would work, we also investigated the link between STAD patients’ Immunophenoscore (IPS) and NMNGscore signatures. The variations in treatment outcomes between the groups with high and low NMNGscores were displayed in Figures 9C–F . The higher IPS scores in the high NMNGscore group suggest that they are more immunogenic on ICIs. These studies reveal that NMNG scores can be used to estimate how an immunotherapy treatment will affect a patient.