Depiction of Aging-Based Molecular Phenotypes With Diverse Clinical Prognosis and Immunological Features in Gastric Cancer

Objective Aging acts as a dominating risk factor for human cancers. Herein, we systematically dissected the features of transcriptional aging-relevant genes in gastric cancer from multiple perspectives. Methods Based on the transcriptome profiling of prognostic aging-relevant genes, patients with gastric cancer in The Cancer Genome Atlas (TCGA) stomach adenocarcinoma (TCGA-STAD) cohort were clustered with a consensus clustering algorithm. Mutational landscape and chemotherapeutic responses were analyzed and immunological features (immunomodulators, immune checkpoint molecules, cancer immunity cycle, and tumor-infiltrating immune cells) were systematically evaluated across gastric cancer. Weighted gene co-expression network (WGCNA) was conducted for screening aging molecular phenotype-relevant genes, and key genes were identified with Molecular Complex Detection (MCODE) analyses. Expressions of key genes were examined in 20 paired tumors and controls with RT-qPCR and Western blotting. Proliferation and apoptosis were investigated in two gastric cancer cells under MYL9 deficiency. Results Three aging-based molecular phenotypes (namely, C1, C2, and C3) were conducted in gastric cancer. Phenotype C1 presented the most prominent survival advantage and highest mutational frequencies. Phenotype C2 indicated low responses to sorafenib and gefitinib, while C3 indicated low responses to vinorelbine and gemcitabine. Additionally, phenotype C2 was characterized by enhanced immune and stromal activation and an inflamed tumor microenvironment. Seven aging molecular phenotype-relevant key genes (ACTA2, CALD1, LMOD1, MYH11, MYL9, MYLK, and TAGLN) were identified, which were specifically upregulated in tumors and in relation to dismal prognosis. Among them, MYL9 deficiency reduced proliferation and enhanced apoptosis in gastric cancer cells. Conclusion Collectively, aging-based molecular subtypes may offer more individualized therapy recommendations and prognosis assessment for patients in distinct subtypes.


INTRODUCTION
Gastric cancer ranks the sixth most frequent malignancy as well as the fifth major cause of cancer death across the globe (1). When diagnosed at an advanced stage, patients' 5-year overall survival rate is merely 5% (1). As a heterogeneous disease, it has the features of diverse histological and molecular subtypes (2). At present, according to the morphology, differentiation, and cohesion of gland cells, gastric cancer is histopathologically classified as intestinal and diffuse (3). Genomic analyses have become the major methodology applied in international efforts for discovering novel biological targets in gastric cancer (4). It is fundamental to unravel the complicated biology underlying gastric cancer etiology and development for overcoming the highly heterogeneous malignancy.
Accumulated pieces of evidence have uncovered the implication of tumor-associated structures and activated signaling pathways both in tumor cells and in the tumor microenvironment (5). Aging is a complicated process primarily categorized by a reduction in tissues, cells, and organ functions as well as an elevated risk of mortality, which acts as a dominant risk factor of diverse fatal malignancies, especially cancers (6). This process presents prominent correlations to telomere attrition, mitochondrial dysfunction, DNA injury, impaired immune system, and the like (7). Nevertheless, the specific mechanisms involved in aging are still indistinct. Transcriptomic studies have identified abundant human aging-relevant genes (8). The human aging genome resource (HAGR) project offers a powerful set of aging-specific network features, which reveals aging-relevant gene signatures as network hubs through comprehensive analyses of biology and genetics of the human aging process (8). Cellular senescence is a permanent state of stagnant replication of proliferating cells as well as a sign of aging (9). Senescent tumor cells triggered by tumorigenesis may lead to cell cycle arrest, as an antitumor mechanism (10). Nevertheless, senescent cells surrounding tumor cells generate opposite results as well as present prominent correlations with senescence-associated secretory phenotype factor secretions (11). Moreover, senescence displays two-tier influences upon cancer immunity (12, 13). Aging-relevant gene signatures exert critical functions in modulating cellular senescence, not only inhibiting tumor progression through modulating senescence of cancer cells but also promoting malignant progression of cancers and dismal clinical outcomes (14). Nevertheless, there is still a lack of systematic analyses of aging-relevant genes during gastric carcinogenesis. Herein, we identified three aging-based molecular phenotypes that offered more individualized therapy options and prognosis prediction for gastric cancer patients.  (17). Somatic mutation data [Mutation Annotation Format (MAF) format] of 433 patients with gastric cancer on the basis of the whole-exome sequencing platforms were curated from the TCGA project. Mutational types and frequencies of genes were analyzed as well as visualized utilizing the maftools package (18). Tumor mutation burden (TMB) was defined as the entire number of non-synonymous variations within the coding regions per megabase (19). In addition, copy number alteration (CNA) data were retrieved from GDAC Firehose (https://gdac. broadinstitute.org), and prominent amplification and deletion across the whole genome were identified with GISTIC2.0 (20). Somatic copy-number alterations (SCNAs) and homologous recombination deficiency (HRD) across gastric cancer specimens were also curated from Davoli et al. (21).

Molecular Characterization for Subtypes
Tumors with qualitatively diverse aging-relevant gene expressions were clustered utilizing hierarchical agglomerative clustering on the basis of Euclidean distance as well as Ward's linkage. Unsupervised clustering method (K-means) was utilized for identifying aging-related molecular phenotypes as well as classifying samples for subsequent analyses. Through consensus clustering algorithm, the number of clusters was determined using TCGA-STAD and GSE84437 cohorts for assessing the stability of the identified molecular phenotypes. This procedure was presented through adopting the ConsensuClusterPlus package as well as repeated 50 times for ensuring the accuracy regarding this classification (22).

Gene Set Variation Analysis (GSVA)
Gene set variation analysis, a non-parametric and unsupervised gene set enrichment algorithm, may infer the enrichment scores of specific pathways or signatures on the basis of transcriptomic profiling (23). The 50 hallmarks of gene signatures were collected from the Molecular Signatures Database (MSigDB) project (24). Moreover, the gene sets of other relevant biological processes were curated from Mariathasan et al. containing CD8 T effectors, DNA damage repair, pan-fibroblast TGF-β response signature (Pan-F-TBRS), antigen-processing machinery, immune checkpoint, epithelial-mesenchymal transition (EMT) markers, FGFR3-related genes, angiogenesis, Fanconi anemia, WNT targets, cell cycle regulators, and the like (25). Utilizing single sample gene set enrichment analysis (ssGSEA) from the GSVA package, gene sets of hallmarks and other relevant biological processes were chosen for presenting quantifications of pathway activity.

Estimation of Chemotherapeutic Response
Chemotherapeutic sensitivity in cancer cells, as well as molecular markers of chemotherapeutic response profiles, were curated from the largest publicly available pharmacogenomics project: the Genomics of Drug Sensitivity in Cancer (GDSC; https:// www.cancerrxgene.org/) (26). Four commonly applied chemotherapeutic agents, sorafenib, gefitinib, vinorelbine, and gemcitabine, were chosen. The prediction procedure was implemented via the pRRophetic package (27). The half-maximal inhibitory concentration (IC50) values were estimated with the ridge regression method, and the prediction accuracy was assessed through 10-fold cross-verification.

Evaluation of Tumor Immune Microenvironment and Immunogenomic Characteristics
Immunological features of the tumor immune microenvironment contained the expression profiles of immunomodulatory factors and immune checkpoint molecules, the activity of the cancer immunity cycle, and infiltrations of immune cells. In total, 122 immunomodulatory factors comprising MHC, receptor, chemokine, and immune stimulator were curated from Sokolov et al. (28). Immune checkpoint molecules with therapeutic potential were collected from Auslander et al. (29). The cancer immunity cycle uncovers antitumor immune response, and the activity of each step determines the fate of tumor cells (30). Here, the activity of each step was quantified with ssGSEA on the basis of the expression profiling of individual specimens. Thereafter, the ssGSEA algorithm was developed for quantifying the abundance of lymphocytes within the tumor immune microenvironment utilizing bulk RNA-seq profiles. Through Estimation of Stromal and Immune cells in Malignant Tumor tissues using Expression data (ESTIMATE), immune and stromal contents (immune and stromal scores and tumor purity) were inferred across gastric cancer specimens (31). Tumor tissues with abundant immune cell infiltration represented an increased immune score and a decreased level of tumor purity.

Quantification of Gene Expression-Based Stemness Index (MRNAsi)
Through the one-class logistic regression (OCLR) method, the stemness index was calculated on the basis of transcriptome profiling of normal PSCs (32). The stemness signatures were generated with the OCLR algorithm (28). Thereafter, this study estimated Spearman's correlation between the weight vector of the stemness signatures and mRNA expression across gastric cancer. Eventually, the stemness index was mapped onto the range of 0 to 1 utilizing a linear conversion, which subtracted the minimum as well as separated through the maximal correlation coefficient. The stemness index produced from transcriptome profiling was defined as mRNAsi.

Weighted Gene Co-expression Network Analysis (WGCNA)
Weighted gene co-expression network analysis was presented for identifying underlying co-expression modules that were prominently correlated with aging-associated molecular phenotypes. The soft-thresholding for the scale-free network was identified. The topological overlap matrix similarity was adopted for the evaluation of the distance between gene pairs. Furthermore, hierarchical clustering analyses with mean and dynamic methods were utilized for building the clustering tree as well as classifying the gene signatures into diverse modules. Following merging the initial modules in line with their similarity, functional modules were eventually conducted. Spearman's correlation coefficient, as well as matched p-value between aging-associated molecular phenotypes and functional modules, was determined through cor function. For each module, gene significance (GS) and module membership (MM) were calculated. Genes with GS > 0.5 and MM > 0.8 were utilized as aging phenotype-relevant genes. Through the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) tool (33), protein-protein interaction (PPI) analysis of aging phenotype-relevant genes was carried out. Molecular Complex Detection (MCODE) (34), a plugin in Cytoscape software (35), was used for screening the significant modules of the PPI network in line with the filtrating criteria of degree cutoff = 2, node score cutoff = 0.2, k-core = 2, and depth from depth = 100.

Patients and Specimens
In total, 20 patients with gastric cancer were recruited at the General Hospital of Ningxia Medical University. Adjacent gastric tissues (3-6 cm from the tumor). Tumor tissues and adjacent non-cancerous gastric tissues (>5 cm from the edge of tumor tissues) were harvested during surgical resection. The inclusion criteria included: (1) patients pathologically diagnosed with gastric cancer and (2) patients who did not experience radioand/or adjuvant chemotherapy prior to surgery. The exclusion criteria included: (1) patients previously diagnosed with other malignancies; (2) patients who were previously treated with radio-or adjuvant chemotherapy; and (3) patients who died within 4 weeks of this surgery. All specimens were frozen in liquid nitrogen at once following collection and were stored at −80 • C before usage. This study was conducted in accordance with the

Real-Time Quantitative Reverse Transcription PCR
Tissues or cells were lysed with RNAiso plus (Takara, Japan). Thereafter, RNA extraction was presented with the phenolchloroform/isopropanol method. The cDNA was prepared through PrimeScript RT reagent kits as well as a gDNA eraser. About 20 µl qPCR system was prepared, followed by analysis with GoTaq qPCR Master Mix. The relative mRNA expressions were quantified with the 2 − Ct method, with GAPDH as a control. The primer sequences are listed in Table 1.

Western Blotting
Tissues or cells were lysed with RIPA buffer plus protease inhibitor cocktail. Following separation via electrophoresis in SDS/PAGE gel, the protein was transferred onto PVDF membranes. The membrane was blocked in PBS-T buffer in supplements of 5% milk/BSA lasting 2 h and presented the incubation with primary antibody targeting MYL9

Aging-Genomic Profiles Identify Three Diverse Molecular Phenotypes of Gastric Cancer
This study analyzed the expression patterns of aging-relevant genes across gastric cancer specimens in the TCGA cohort. Through univariate-cox regression analyses, abnormal expression of 24 aging-relevant genes was in relation to gastric cancer prognosis ( Table 2). With the consensus clustering method, patients with gastric cancer were clustered into three aging-relevant molecular phenotypes (C1, 143 samples; C2, 117 samples; C3, 91 samples) in accordance with the transcriptome profiling of prognostic aging-relevant genes ( Figure 1A). PCA uncovered the dissimilarity between agingrelevant molecular phenotypes ( Figure 1B). The prominent discrepancy in expressions of prognostic aging-relevant genes was investigated among phenotypes ( Figure 1C). Survival analyses demonstrated that three aging-relevant molecular phenotypes presented prominent survival outcomes. C1 phenotype possessed more favorable OS (Figure 1D), DFS (Figure 1E), and DSS ( Figure 1F) outcomes than C2 and C3 phenotypes. The classification accuracy was confirmed in the GSE84437 cohort (Supplementary Figures 1A-D). Figure 1G showed the heterogeneity in the distribution of three agingrelevant molecular phenotypes (C1, C2, and C3) among the most known molecular subtypes (CIN, EBV, GS, and MSI). C1 subtype occupied the highest percentage in EBV and MSI, while C2 occupied the highest percentage in GS. Thus, the aging-relevant molecular phenotypes presented remarkable associations with the most known molecular subtypes of gastric cancer.

Aging-Associated Molecular Phenotypes With Distinct Activations of Functional Pathways and Chemotherapeutic Responses
We further investigated the mechanisms underlying distinct aging-associated molecular phenotypes. In Figure 3A, we observed that immune activation pathways (complement, IL2-STAT5 signaling, inflammatory response, IL6-JAK-STAT3 signaling, allograft rejection, and interferon gamma response) and stromal activation pathways (epithelialmesenchymal transition, angiogenesis, and WNT β-catenin signaling) were prominently upregulated in aging-associated molecular phenotype C2. Several tumorigenic pathways (mTORC1 signaling, MYC targets, DNA repair, E2F targets, and G2M checkpoint) presented a significant activation in molecular phenotypes C1 and C3. Consistently, pan-F-TBRS, immune checkpoint, EMT1-3, angiogenesis, and WNT target were prominently upregulated in molecular phenotype C2 ( Figure 3B). The above data demonstrated the immune and stromal activation in molecular phenotype C2. The chemotherapeutic responses to sorafenib, gefitinib, vinorelbine, and gemcitabine were compared among three molecular phenotypes. Our results showed that molecular phenotype C2 presented the lowest therapeutic responses to sorafenib and gefitinib (Figures 3C,D) while phenotype C3 had the lowest therapeutic responses to vinorelbine and gemcitabine (Figures 3E,F).

Aging-Associated Molecular Phenotypes Display Diverse Tumor Immune Microenvironment and Immunological Status
In Figure   CCL4, and CCL8) and their receptors (XCR1, CCR1, CCR10, CCR2, CCR4, CCR5, CCR6, CCR7, CCR8, CCR9, CXCR1, CXCR3, CXCR4, and CXCR5), and the lowest expression of above molecules was found in molecular phenotype C1 (Figures 4B,C). Above chemokines and receptors facilitate the recruitment of effector lymphocytes like CD8+ T cell, TH17 cell, as well as antigen-presenting cells. In Figure 4D, molecular phenotype C2 was characterized by the highest expressions of most immune checkpoint molecules (IL2RA, IL6, IL6R, KLRK1,  LTA, NT5E, RAET1E, TNFRSF13B, TNFRSF13C, TNFRSF17,  TNFRSF4, TNFRSF8, TNFSF13B, TNFSF14, TNFSF18, TNFSF4,  ENTPD1, BTNL2, CD27, CD276, CD28, CD40, CD40LG, CD48, and CD86). These data reflected the activated immunological status in molecular phenotype C2. Cancer immunity cycle activity is the overall manifestation of the chemokine system as well as immunomodulatory factors. Most steps in the cancer immunity cycle presented the highest activities in molecular phenotype C2, like cancer cell antigen release and presentation, priming and activation, recruitment of B cell, CD4 T cell, dendritic cell, eosinophil, macrophage, monocyte, T cell, Th17 cell, and Treg ( Figure 4E). Thereafter, we calculated the infiltration levels of immune cells utilizing the ssGSEA algorithm. The infiltration levels of most immune cells were upregulated in molecular phenotype C2, like activated B cell, activated CD8 T cell, central memory CD4 T cell, central memory CD8 T cell, effector memory CD4 T cell, effector memory CD8 T cell, gamma delta T cell, immature B cell, memory B cell, regulatory T cell, T follicular helper cell, type 1 T helper cell, type 2 T helper cell, activated dendritic cell, eosinophil, immature dendritic cell, macrophage, mast cell, MDSC, natural killer cell, natural killer T cell, and plasmacytoid dendritic cell ( Figure 4F). Collectively, molecular phenotype C2 had an inflamed tumor microenvironment.

Aging-Associated Molecular Phenotypes Associated With Immunotherapeutic Response Predictors in Gastric Cancer
We investigated the difference in immunotherapeutic responses among three aging-associated molecular phenotypes through comparisons of multiple immunotherapeutic predictors. Agingassociated molecular phenotype C2 presented higher stromal and immune scores as well as reduced tumor purity compared with C1 and C3, indicating that samples in phenotype C2 had increased infiltrations of stromal and immune cells (Figures 5A-C). The mRNAsi was quantified for reflecting the levels of cancer stem cells across gastric cancer. There was the lowest mRNAsi in phenotype C2, while the highest mRNAsi in phenotype C1 ( Figure 5D). Also, we investigated the lowest SCNA in phenotype C2 but the highest SCNA in phenotype C3 ( Figure 5E). In Figure 5F, phenotype C2 displayed the lowest MSI, while phenotype C1 possessed the highest MSI. Phenotype C2 presented the lowest TMB score but C1 had the highest TMB score (Figure 5G). We also evaluated the differences in cancer testis antigen (CAT) and HRD score among three aging-associated molecular phenotypes. We investigated that phenotype C3 had the highest CAT score, followed by C2 and C1 (Figure 5H). Additionally, the highest HRD score was found in phenotype C3 (Figure 5I). The above data suggested that aging-associated molecular phenotypes presented distinct immunotherapeutic responses in gastric cancer.

Identification of Aging Molecular Phenotype-Relevant Key Genes
The WGCNA method was adopted for the construction of a co-expression network as well as finding genes highly associated with aging molecular phenotypes. We first detected outliers among gastric cancer specimens on the basis of gene expression profiling. As a result, there was no outlier sample ( Figure 6A). Thereafter, soft thresholding power β was calculated, and β was set at 5 for ensuring a scale-free network (Figures 6B,C). In total, 11 co-expression modules were merged, as depicted in Figure 6D. Among them, the brown module presented the strongest association with aging molecular phenotype C2 ( Figure 6E). Thereafter, we evaluated intramodular analyses of genes in each module. Especially, genes in the brown module had high correlations with aging molecular phenotype C2 (Figure 6F). Eventually, 312 genes in this module were selected as aging molecular phenotype-relevant genes in accordance with the criteria of module membership >0.8 and gene significance >0.5 (Supplementary Table 3). We further observed the interactions between aging molecular phenotyperelevant genes through the STRING database. With MCODE analyses, seven aging molecular phenotype-relevant hub genes were identified, namely, ACTA2, CALD1, LMOD1, MYH11, MYL9, MYLK, and TAGLN ( Figure 6G). In Figure 6H, we noted that the hub genes displayed remarkable associations with the infiltration levels of immune cells. All of them were negatively correlated to the infiltration levels of activated CD4 T cell, CD56dim natural killer cell, neutrophil, and type 17 T helper cell but were positively associated with the other immune cells.

Verification of Prognostic Implication and Deregulated Expression of Aging Molecular Phenotype-Relevant Key Genes
Survival analyses were conducted for investigations of the prognostic implications of aging molecular phenotype-relevant key genes across patients with gastric cancer. Our data demonstrated that the upregulations of ACTA2, CALD1, LMOD1, MYH11, MYL9, MYLK, and TAGLN were in relation to more dismal survival outcomes in comparisons with their downregulations (Figures 7A-G). We further verified their expressions in 20 paired tumors and controls.
In Figure 7H, compared with controls, their prominent upregulations were investigated in tumors in line with RT-qPCR. Additionally, Western blotting results confirmed their abnormal expressions of these key genes in gastric cancer (Figures 7I-P).

In Gastric Cancer Cells, MYL9 Loss Weakens Proliferation and Triggers Apoptosis
Among aging molecular phenotype-relevant key genes, only the role of MYL9 in gastric cancer remains unknown. Thus, we investigated the function of MYL9 in gastric carcinogenesis. Herein, MYL9 expressions were reduced in MGC-803 and BGC-823 cells under two shRNAs against MYL9 transfections (Figures 8A-C). In accordance with CCK-8 results, MYL9 loss reduced the cell viability of MGC-803 and BGC-823 cells (Figures 8D,E). Additionally, apoptosis of MGC-803 and BGC-823 cells was enhanced when MYL9 expressions were defective (Figures 8F,G). The above data indicated the gastric tumorigenic roles of MYL9.

DISCUSSION
In our study, we conducted three aging-based molecular phenotypes with a consensus clustering algorithm. Further, aging-based molecular phenotypes were characterized by diverse clinical prognoses, mutational status as well as the immunological status of tumor microenvironment across gastric cancer. In this aspect, our findings offered individualized treatment options and prognosis evaluation for distinct subpopulations based on the aging-related molecular phenotypes. The tumor microenvironment is comprised of a heterogeneous cellular milieu that influences cancer cell behaviors (3). The feature produces a far-reaching impact on treatment responses like immunotherapy. An inflamed tumor microenvironment combined with preexisting antitumor immunity is necessary for immunotherapy that suppresses tumor growth through tumor-cytotoxic T-cell re-invigoration. In theory, molecules and signals contribute to an inflamed tumor microenvironment that may trigger sensitivity to immunotherapy. Herein, in accordance with immunological features (immunomodulators, immune checkpoint molecules, cancer immunity cycle, and tumor-infiltrating immune cells), aging-based molecular phenotype C2 presented an inflamed tumor microenvironment. This indicated that the subpopulations in this phenotype possessed greater chances of responding to immunotherapy. Cancer stem cells contribute to chemotherapeutic resistance as well as distant metastases due to the self-renewal and tumorigenic capacities (37). Through mRNAsi, we quantified the levels of cancer stem cells across gastric cancer. There was the lowest mRNAsi in phenotype C2, while the highest mRNAsi in phenotype C1. TMB and MSI are capable of predicting the clinical responses to immunotherapy. Nevertheless, the predictors are examined utilizing complex molecular tools that are slow and expensive. Thus, it is an urgent medical requirement for developing faster and economical predictors. Our data indicated that phenotype C2 displayed the lowest MSI and TMB scores, while phenotype C1 possessed the highest MSI and TMB scores. HRD leads to impaired double-strand break repair, which is a common driving factor of carcinogenesis (38). Herein, phenotype C2 presented the features of reduced HRD score, while phenotype C3 was characterized by elevated HRD score.
Through WGCNA combined with MCODE methods, we identified seven aging molecular phenotype-relevant key genes, namely, ACTA2, CALD1, LMOD1, MYH11, MYL9, MYLK, and TAGLN. The above genes displayed the specific upregulations in gastric cancer and contributed to a dismal clinical prognosis. Previously, CALD1 acts as a prognostic indicator and also is in relation to immune infiltrates in gastric carcinoma (39). MYH11 expression is downregulated in gastric carcinoma and is indicative of a dismal clinical prognosis (40). Hypermethylation of MYLK serves as a circulating diagnostic marker of gastric carcinoma (41). Stromal fibroblasts in the microenvironment trigger gastric carcinoma metastases through the upregulation of TAGLN (42). Among them, our experimental pieces of evidence demonstrated that MYL9 deficiency reduced proliferation as well as enhanced apoptosis in gastric carcinoma cells, confirming the tumorigenic function of MYL9. Nevertheless, there were a few limitations in this study. The aging-based molecular phenotypes should be further verified in large patients from multicenter cohorts for identifying the characteristics of clinical prognosis and drug responses. Additionally, we identified aging molecular phenotype-relevant key genes, especially MYL9. Nevertheless, the specific experimental verifications should be designed for the assessment of the biological implications.

CONCLUSION
Herein, our comprehensive assessment of the cellular, molecular, and genetic features correlated with agingbased molecular phenotypes generated novel insights on how gastric tumors responded to immunotherapy and guided the development of more effective combination therapeutic regimens.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

ETHICS STATEMENT
The study was approved by the Ethics Committee of General Hospital of Ningxia Medical University (approval no. 2020-031). The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
YZhu and SY conceived and designed the study. FH, HD, and YZhou conducted most of the experiments and data analysis and wrote the manuscript. YW and JX participated in collecting data and helped to draft the manuscript. All authors reviewed and approved the manuscript.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmed. 2021.792740/full#supplementary-material confirmed the classification accuracy of aging-relevant molecular phenotypes. (C) Heatmap showed the expression patterns of prognostic aging-relevant genes in three aging-relevant molecular phenotypes. (D) Kaplan-Meier survival curves were conducted for gastric cancer patients with diverse molecular phenotypes.