Abstract Background Prostate cancer (PCa) is a prevalent malignancy among men, primarily originating from the prostate epithelium. It ranks first in global cancer incidence and second in mortality rates, with a rising trend in China. PCa's subtle initial symptoms, such as urinary issues, necessitate diagnostic measures like digital rectal examination, prostate-specific antigen (PSA) testing, and tissue biopsy. Advanced PCa management typically involves a multifaceted approach encompassing surgery, radiation, chemotherapy, and hormonal therapy. The involvement of aging genes in PCa development and progression, particularly through the mTOR pathway, has garnered increasing attention. Methods This study aimed to explore the association between aging genes and biochemical PCa recurrence and construct predictive models. Utilizing public gene expression datasets ([30]GSE70768, [31]GSE116918, and TCGA), we conducted extensive analyses, including Cox regression, functional enrichment, immune cell infiltration estimation, and drug sensitivity assessments. The constructed risk score model, based on aging-related genes (ARGs), demonstrated superior predictive capability for PCa prognosis compared to conventional clinical features. High-risk genes positively correlated with risk, while low-risk genes displayed a negative correlation. Results An ARGs-based risk score model was developed and validated for predicting prognosis in prostate adenocarcinoma (PRAD) patients. LASSO regression analysis and cross-validation plots were employed to select ARGs with prognostic significance. The risk score outperformed traditional clinicopathological features in predicting PRAD prognosis, as evidenced by its high AUC (0.787). The model demonstrated good sensitivity and specificity, with AUC values of 0.67, 0.675, 0.696, and 0.696 at 1, 3, 5, and 8 years, respectively, in the GEO cohort. Similar AUC values were observed in the TCGA cohort at 1, 3, and 5 years (0.67, 0.659, 0.667, and 0.743). The model included 12 genes, with high-risk genes positively correlated with risk and low-risk genes negatively correlated. Conclusions This study presents a robust ARGs-based risk score model for predicting biochemical recurrence in PCa patients, highlighting the potential significance of aging genes in PCa prognosis and offering enhanced predictive accuracy compared to traditional clinical parameters. These findings open new avenues for research on PCa recurrence prediction and therapeutic strategies. Keywords: Aging-related Genes, Machine Learning, Immune microenvironment, Prognosis, Prostate Cancer, Single-Cell Analysis, Biochemical recurrence Introduction PCa is one of the most common malignant tumors in men, mostly caused by malignant tumors of the prostate epithelium [[32]1]. From the latest global cancer statistics in 2019, we know that the incidence of PCa is ranked first and the death rate is also ranked second, and the incidence of PCa in China has been rising in recent years[[33]2, [34]3]. Early symptoms are more insidious, such as urinary frequency, urgency, and reduced urinary flow, comparable to those of prostate enlargement[[35]4]. PCa is not diagnosed accurately through symptoms but requires digital rectal examination, prostate-specific antigen (PSA) test to assist in the diagnosis, and tissue biopsy to confirm the diagnosis [[36]5, [37]6]. Patients with advanced disease usually require a combination of therapies including surgery, radiation, chemotherapy, and hormonal therapy [[38]7, [39]8]. The mechanisms and significance of aging are becoming increasingly important as the population ages. Aging genes are a group of genes involved in the aging process whose main role is to control a number of molecular biological processes of organismal aging (DNA damage response and cell cycle regulatory pathways, etc.) [[40]9, [41]10]. Aging genes may be involved in PCa development and progression by regulating cancer cell proliferation, cell cycle, and apoptosis [[42]11]. Its involvement in the encoding of the mTOR protein and over-activation of the mTOR pathway, which activates Akt, can inhibit apoptosis in cancer cells28283069 [[43]12]. These findings provide valuable insights into the relationship between the Aging gene and PCa. The aim of this study was to construct a predictive model by exploring the association between senescence genes and biochemical recurrence in PCa. PCa remains a significant clinical challenge due to its high incidence and potential for recurrence after initial treatment. Despite advances in therapeutic strategies, predicting BCR remains difficult, which underscores the urgent need for reliable biomarkers. However, the specific contribution of these genes to PCa recurrence has not been fully elucidated. By utilizing two public gene expression profile datasets, [44]GSE70768 and [45]GSE116918, as training sets, and validating the findings with the TCGA dataset, this study aims to identify key genes that are strongly associated with the biochemical recurrence of PCa. Through rigorous statistical analyses, including analysis of variance, univariate analysis, and the application of lasso and stepwise multifactorial screening, a prognostic model was developed. The innovative aspect of this study lies in its focus on the integration of senescence-related gene expression profiles to predict BCR in PCa, an area that has been underexplored. This study includes data from a total of 323 patients, offering new theoretical insights that may inform future research on predicting and treating biochemical recurrence in PCa, thereby filling a critical gap in existing research. Methods Data sources We downloaded the PCa expression profile microarray from the public database Gene Expression Omnibus (GEO), which includes peripheral blood samples from PRAD patients and controls, containing [46]GSE70769 [[47]13] and [48]GSE116918 [[49]14] totaling 268 samples, and the TCGA database to do the validation group, containing 55 columns. Prognostic analysis and column line plot construction We conducted univariate and multivariate Cox regression analyses to determine if risk scores could serve as independent prognostic indicators. Utilizing the "rms" R package, we integrated these risk scores with clinicopathologic features to create histograms that predict patient survival at 1, 3, and 5 years within the TCGA-PRAD cohort. Estimation of infiltrating immune cells using CIBERSORT analysis CIBERSORT analysis ([50]https://cibersort.stanford.edu/) is a robust tool that leverages gene expression data to estimate the relative proportions of various immune cell types within complex tissue samples. To assess immune cell infiltration in cancer tissues, we utilized this method, allowing us to precisely quantify the relative abundance of different immune cells. This approach provides a detailed understanding of the immune landscape in heterogeneous samples, offering valuable insights into the tumor microenvironment.[[51]15]. Drug sensitivity To evaluate the efficacy of treatments across different risk categories, we utilized the "pRRophetic" R package. This tool allowed us to analyze treatment responses in patients classified as either high-risk or low-risk based on their prognostic scores. We drew on data from the Genomics of Drug Sensitivity in Cancer (GDSC) database, which provides comprehensive information on drug responses. Specifically, we used the dataset to obtain half-maximal inhibitory concentration (IC50) values, which measure the concentration of a drug required to inhibit a biological process by 50%. This approach enabled us to assess how effectively various treatments could suppress cancer growth in PRAD patients, providing insights into potential therapeutic strategies [[52]16]. GeneSet cancer analysis database GSCALite ([53]http://bioinfo.life.hust.edu.cn/web/GSCALite/) is a comprehensive online tool that integrates genomic data from 33 types of cancer available in The Cancer Genome Atlas (TCGA) with normal tissue data from the Genotype-Tissue Expression (GTEx) project. In our study, we utilized GSCALite to perform a detailed analysis of various genomic alterations, including copy number variations, DNA methylation patterns, and pathway activities related to ARGs in PRAD. This platform facilitated a thorough examination of how these genomic features influence the biological behaviors of ARGs in PRAD. Tumor immune single cell hub database TISCH ([54]http://tisch.comp-genomics.org) is a comprehensive database dedicated to single-cell RNA sequencing data specifically focused on the tumor microenvironment (TME) [[55]17]. Utilizing this resource, we systematically investigated the heterogeneity of the tumor microenvironment across various cell types and datasets. Statistical analysis All statistical analyses were conducted using R software (V. 4.2.0). To evaluate the reliability of the diagnostic model, ROC curves were generated, with the area under the curve (AUC) used to determine predictive accuracy, applying a significance threshold of P < 0.05. Furthermore, the goodness-of-fit of the constructed nomograms was assessed using the Hosmer–Lemeshow test. This rigorous analysis ensured a thorough evaluation of the model's performance and fit. Result Construction and validation of ARGs signatures. A risk score model based on ARGs was developed to identify prognostic biomarkers for PRAD patients. LASSO regression analysis (Fig. [56]1A) was utilized on DE-ARGs with prognostic significance, and cross-validation plots (Fig. [57]1B) identified 12 key genes: HDAC3, IRS2, HIF1A, PRKCA, MSRA, APOE, HSPA9, CDKN2A, TP53BP1, CNR1, CDKN2B, and SERPINE1. High-risk genes were found to have a positive correlation with risk, whereas low-risk genes were negatively associated. The risk score model demonstrated strong predictive power for PRAD prognosis, with an AUC of 0.787, outperforming traditional clinicopathologic features (Fig. [58]1C). In the GEO cohort, the model's predictive performance was further validated, showing high sensitivity and specificity with AUC values of 0.67, 0.675, 0.696, and 0.696 at 1, 3, 5, and 8 years, respectively (Fig. [59]1D). Similarly, in the TCGA cohort, the AUC values were 0.67, 0.659, 0.667, and 0.743 at 1, 3, and 5 years (Fig. [60]1E). Survival analysis in the GEO cohort revealed that patients with higher risk scores had significantly increased mortality, while those in the low-risk group demonstrated a better prognosis (P < 0.001) (Fig. [61]1F). This trend was consistent in the TCGA cohort, where a better prognosis was also observed in the low-risk group (P = 0.003) (Fig. [62]1G). Overall, this ARG-based risk score model offers superior predictive power and may serve as a valuable tool in identifying prognostic biomarkers and guiding clinical decisions for PRAD patients. Fig. 1. [63]Fig. 1 [64]Open in a new tab Construction and validation of ARGs signatures. A Ten-fold cross-validation used to adjust parameter selection in the LASSO model. B Lasso coefficient profiles. C Multi-exponential ROC analysis. D ROC profile analysis of the GEO cohort over time. E ROC profile analysis of the TCGA cohort over time. F KM curves comparing overall PRAD patients between low and high risk groups in the GEO cohort. G KM curves comparing overall PRAD patients between low and high risk groups in the TCGA cohort Exploring the relationship between the prognostic model's scoring profile of prostate patients and biochemical recurrence (BCR) in the clinic. We collected BCR profiles and time to recurrence (bcr time) of prostate patients and plotted a scatterplot with the prognostic model's scoring profiles (dataset), and from the plot, we clearly observed that with the rise in patients' RISK SCREEN, not only the number of recurrences in BCR of recurrence, the time to recurrence (bcr time) also increased from 0–6 years to a wide range of 0–8 years (Fig. [65]2A, [66]B), which also suggests that our prognostic model can, to some extent, also predict the circumstances and timing of biochemical recurrence in prostate patients. We then categorized the patients into high and low risk groups based on the median value of the risk score (Fig. [67]2C, [68]D), and viewed the expression of the modeled genes between the high and low risk groups through the limma package, and we could see that all modeled genes, except HSPA9, were actively expressed in the high risk group (Fig. [69]2E, [70]F). By principal component analysis (PCA) of the patients we can clearly see that the high and low risk labels can classify and summarize the patients well and there is very little overlap of patients (Fig. [71]2G, [72]H). Fig. 2. [73]Fig. 2 [74]Open in a new tab Exploring the relationship between the prognostic model's scoring profile of prostate patients and biochemical recurrence (BCR) in the clinic. A, B Scatterplot of the scoring profile of the prognostic model for prostate patients versus the time to clinical biochemical recurrence (BCR) recurrence (bcr time). C, D Scatterplot of the distribution of risk scores for prostate patients. E, F Heatmap of differential expression of modeled genes between high, and low risk groups. G, H Principal component analysis of prostate patients The construction of the nomogram. To explore the relationship between risk factors such as prognostic model scores and clinical characteristics and BCR outcomes in prostate patients. We plotted forest plots by the results of univariate and multivariate COX regression analyses as well as Log Rank tests (Fig. [75]3A, [76]B) in which the P value of T-stage and risk scores were < 0.05, the Hazard ratio (HR) of T-stage in one-way, multifactor COX regression was 0.569 as well as 0.585, and HR > 1 was a risk factor. Hazard ratio (HR) for multifactorial COX regression was 0.569 as well as 0.585, in addition to pvalue < 0.001 for both risk scores, and HR > 1 was a risk factor. Subsequently, we constructed a multi-indicator clinical prognostic model based on the results of the multifactorial COX analysis, and plotted the nomogram and the corrected curves (Fig. [77]3C, [78]D). Based on the nomogram, each sample was scored by the COX analysis, and based on the Points, we predicted the patient's OS in the clinical diagnosis and treatment, and the results of the corrected curves showed the discrepancy between the predicted and actual results of COX model. The results of the correction curves show that the COX model has excellent predictive performance, as the difference between the predicted results and the actual situation is very small. Fig. 3. [79]Fig. 3 [80]Open in a new tab The construction of the nomogram. A Forest plot for a one-factor COX analysis. B Forest plot for multifactor COX analysis. C Nomogram of a multifactor COX analysis model. D Correction curve plot for a multifactor COX model Correlation between clinical characteristics and biochemical recurrence of ARGs in PRAD patients. Given that BCR in the high-risk (HR) and low-risk (LR) populations differed dramatically in terms of individual clinical attributes. In order to study this difference in depth and compare it more precisely, we categorized patients diagnosed with PRAD into five different subgroups based on clinical parameters. These stratifications included age (≤ 65 and > 65 years), gleason (6–7 and 8–9), pca (10 > and <  = 10), and T-stage (T1-2 and T3-4). Notably, in all subgroups, the survival of LR patients was significantly better than that of HR patients, which was characterized by a longer survival period (Fig. [81]4A-H). Based on the analysis of the results, we further strengthened our confidence in the reliability of the ARGs profile as a clinical predictive tool. Fig. 4. [82]Fig. 4 [83]Open in a new tab Correlation between clinical characteristics and biochemical recurrence of ARGs in PRAD patients. A–H KM curves between different clinical features Analysis of GSCALite and cBioPortal Data Figure [84]5 highlights genomic alterations involving ARGs and hub genes across three domains: single-nucleotide variation (SNV) (Fig. [85]5A-B, [86]I), copy-number variation (CNV) (Fig. [87]5D, [88]H), and methylation (Fig. [89]5F-G). ARGs did not show significant mutations in KIRP (Fig. [90]5A). Fig. 5. [91]Fig. 5 [92]Open in a new tab Analysis of GSCALite and cBioPortal Data. A SNV of all mutated genes in the gene set in PRAD. B SNV classes of hub-gene set in PRAD, C Survival difference between high and low methylation in each cancer. D Survival difference between CNV groups. E Correlations between methylation and mRNA expression of ARGs in PRAD. F Correlation between methylation and mRNA expression; G Methylation differences among tumor and normal samples of SKA3 and top ten hub genes in PRAD; H Pie plot summarizing CNV of ARGs. I SNV of ARGs and hub genes in PRAD. J Correlation between CTRP drug sensitivity and mRNA expression CNV in PRAD patients encompassed heterozygous, homozygous, amplifications, and deletions (Fig. [93]5H). Importantly, no apparent correlation was observed between heterozygous or homozygous CNV. However, CNV in genes such as MSRA, TP53BP1, IRS2, HSPA9, and HDAC3 displayed significant associations with mRNA expression, with MSRA exhibiting a particularly strong correlation (F[94]ig. [95]5I). In PRAD patients, CNV groups, including CNR1, HSPA9, and TP53BP1, demonstrated a negative correlation with overall survival (OS) and progression-free survival (PFS), while others showed varying degrees of positive correlations (Fig. [96]5D). The analysis also unveiled differential methylation patterns in PRAD genes between tumor and normal samples (Fig. [97]5G). Specifically, low methylation of APOE and CDKNA2 was associated with poorer overall survival (OS) in KIRP (Fig. [98]5C). Furthermore, methylation of ARGs displayed a negative correlation with mRNA expression (Fig. [99]5F). In the majority of genes, there was a positive correlation between CTRP drug sensitivity and mRNA expression. SERPINE1 exhibited a significant positive correlation, while a few genes, including CNR1, displayed negative correlations. This suggests a degree of specificity in the drug sensitivity experiment, providing valuable insights for future clinical research in developing treatment strategies. Correlation Analysis of ARGs with Clinicopathological Characteristics In Fig. [100]6A, our analysis revealed a notable influence of the ARGs on the distribution of specific four clinicopathological features within both high-risk and low-risk groups. It was obvious that patients aged 65 and older, those with pca more than 10, and individuals at gleason7 comprised a larger proportion of patients in the high-risk (HR) group. Furthermore, the heatmap illustrates various clinicopathological features, including T stage, age, gleason, pca, and risk scores across the entire cohort of TCGA-PRAD patients (Fig. [101]6B). We extended our analysis to explore the relationship between risk scores and various clinicopathological factors, including tumor grade, disease stage, T stage, patient age, and gender. These correlations were systematically evaluated to understand how each factor interacts with the risk scores, as illustrated in Fig. [102]6C to 6F. The analysis revealed significant variations in risk scores among patients with differing age, pca, gleason, and T stages, with patients in more advanced stages showing higher risk scores. Based on our findings, we concluded that a significant positive correlation exists between risk scores and various clinicopathological factors. Fig. 6. [103]Fig. 6 [104]Open in a new tab Distribution of risk scores in different clinical subtypes. A The proportion of patients with different clinical subtypes (Age, Pca, Gleason, T stage) in the HR group and LR group. B Heatmap of clinicopathological variables in HR group and LR group. The proportion of patients with different clinical subtypes (Age, Pca, Gleason, T stage) in the HR group and LR group. C-F Risk score distribution of different clinical subtypes Immunoassays in patients with PRAD Immune cell infiltration represents a fundamental aspect of the TME. Utilizing the CIBERSORT algorithm for Spearman correlation analysis, we observed a notable association between risk scores and the abundance of immune cells in the PRAD TME. Specifically, CD8 + T cells were predominantly correlated with CD4 + T cells (Fig. [105]7A). In the combined analysis of the 12 ARGs with immune cells, HSPA9 was found to be highly correlated with M1 macrophages (Fig. [106]7B). To assess the distribution and correlation of the 22 tumor-infiltrating immune cells (TICs) in the GEO cohort, we utilized CIBERSORT as the immune analysis tool. The results indicated that BRC samples exhibited significantly higher levels of immune infiltration compared to non-BRC samples, particularly in B cells, plasma cells, and macrophages (Fig. [107]7C). The ARG-based risk score model effectively differentiated between various immune subtypes, thereby influencing the response to immunotherapy. To further investigate changes in immune function, we conducted a comparison of single-sample GSEA (ssGSEA) scores, revealing a significant increase in scores for the high-risk group (Fig. [108]7D). Additionally, we examined differences in the expression of immune checkpoint genes, which are critical for tumor immunotherapy. In the low-risk group, 13 immune checkpoint genes, including BTNL2, CD244, CD28, CD40LG, CTLA4, LAIR1, NRP1, PDCD1, TIGIT, TNFRSF25, TNFRSF8, TNFRSF9, and TNFSF9, were significantly upregulated. In contrast, the high-risk group showed upregulation of only TNFSF9 and TNFRSF25 (Fig. [109]7E). The upregulation of immune checkpoints suggests the presence of inflammation within the TME [[110]18], implying that low-risk patients may have an inflammatory microenvironment. Targeted therapies against these elevated immune checkpoints could potentially benefit this tumor subtype [[111]19]. Fig. 7. [112]Fig. 7 [113]Open in a new tab Immunoassays in patients with PRAD. A Histogram of immune cells. B Correlation of 12 genes with immune cells. C Differences in immune cell infiltration between high- and low-risk groups. D Immune function ssGSEA scores between high- and low-risk groups. E Differences in immune checkpoints between high- and low-risk groups Correlation study of ARGs with the immune microenvironment of PRAD We scrutinized the expression of 12 ARGs in the immune microenvironment using the PRAD_GSE143791 single-cell dataset retrieved from the TISCH database.There are 15 different immune cell types in [114]GSE143791 (Fig. [115]8A). We used pie charts to represent the proportional composition of different immune cells and their distribution in the samples (Fig. [116]8B). To deeply investigate the expression levels of individual ARGs in immune cells, we generated a downscaled distribution map of CCRGs in immune cells (Fig. [117]8C-N). Our analysis showed that HDAC3, HIF1A, HSPA9, and TP53BP1 were widely expressed in a wide range of AML immune cells, whereas the expression of CNR1 and CDKN2B in the immune microenvironment was almost negligible. These findings based on the PRAD dataset validate the correlation studies between ARGs and the immune microenvironment, thus complementing and refining the clinical targeting of PRAD induced by the immune microenvironment. Fig. 8. [118]Fig. 8 [119]Open in a new tab Correlation study of ARGs with the immune microenvironment of PRAD. A Downscaled distribution of various immune cell subpopulations of PRAD_GSE143791. B Pie chart showing percentage of immune cells. C–N Distribution of 12-ARGs in PRAD_GSE143791 Discussion PCa is the most common malignancy among men, emphasizing the importance of effective screening and detection methods. PSA testing has proven valuable in identifying localized PCa. However, PSA testing is limited by its lack of sensitivity and specificity. While PSA screening has raised the lifetime risk of a PCa diagnosis to 16%, the mortality rate remains relatively low at 3.4% [[120]20]. This discrepancy suggests that increased detection of slow-growing or relatively benign cancers, which do not necessarily require definitive treatment, has led to concerns about overdiagnosis and overtreatment, exposing patients to unnecessary risks and potential urinary and bowel dysfunction post-treatment [[121]21]. Recent reports indicate that a significant proportion of men with low PSA levels still develop PCa, many of which are high-grade malignancies. Thus, PSA is less effective as a screening tool for differentiating between high and low-risk cases. Research is ongoing to identify other markers that could more accurately pinpoint malignancies that are clinical threats while avoiding interventions for inert diseases. Preventive strategies tailored to genetic or other risks may help reduce the incidence of PCa [[122]22]. PCa incidence escalates markedly with age. Data from the US Surveillance, Epidemiology, and End Results Program (2000–2008) indicate that the rate of PCa is 9.2 per 100,000 men in the 40–44 age group. This incidence rises sharply to 984.8 per 100,000 men aged 70–74 years before experiencing a slight decline [[123]20]. PCa often develops gradually, typically preceded by dysplastic lesions that may remain undetected for many years or even decades. Autopsy studies have indicated that if most men lived to 100 years old, they would likely develop PCa [[124]23]. Macrophage-tumor cell interactions have been found to promote androgen resistance and increase PCa invasion through tissue factor expression. Studies by Parrinello et al. demonstrated that aged mice with increased macrophage infiltration in the prostate glands reflect the role of immune cells in aging and its association with PCa development [[125]24]. Thus, prognostic models based on senescence-related biomarkers can complement PSA screening for early diagnosis and predict genetic risk related to senescence [[126]25]. We merged two prostate tumor patient cohort transcript datasets, [127]GSE70768 and [128]GSE116918, collected from public databases, and extracted the aging-related differential genes that were differentially expressed in the patients after de-batching and normalization, and then the Aging-DEGs were analysed by one-way COX analysis, followed by a lasso machine learning approach and stepwise multifactorial COX analysis. analysis to screen the genes for constructing prognostic models (HDAC3,IRS2,HIF1A,PRKCA,MSRA, APOE, HSPA9, CDKN2A, TP53BP1, CNR1, CDKN2B, and SERPINE1), and finally a prognostic model was constructed by logistic regression algorithms in patients with 12-Agings prostate tumors, and additionally A nomogram of prostate tumor patients was depicted as well as the immune microenvironment, immune function, mutation load, pathway enrichment analysis, and clinical subgroup survival analysis of patients with high- and low-risk prostate tumors were explored using the CIBERSORT database. Histone deacetylase 3 (HDAC3) is an enzyme with histone deacetylase activity that plays a critical role in transcription regulation. By binding to the promoter region, HDAC3 inhibits transcription. Additionally, it modulates gene expression through interaction with the zinc finger transcription factor YY1 and suppresses p53 activity, which is essential for regulating cell growth and apoptosis. HDAC3 is recognized as a potential tumor suppressor gene. It has been proposed that the corepressor SMRT, together with N-CoR and HDAC3, forms a complex that inhibits AR activity and interacts with AR nuclear steroid receptors to suppress specific protein expression in PCa cell lines [[129]26]. Hypoxia-inducible factor 1 subunit alpha (HIF1A) is a crucial transcriptional regulator that enables cells to adapt to low oxygen environments. In hypoxic conditions, HIF1A drives the expression of over 40 genes that enhance oxygen delivery and support metabolic adaptation. These include genes for HILPDA, vascular endothelial growth factor, glycolytic enzymes, glucose transporters, and erythropoietin [[130]27, [131]28]. HIF1A plays a crucial role in embryonic and tumor angiogenesis, as well as in the pathophysiology of ischemic diseases, influencing both cell proliferation and survival [[132]29]. Early prostatic intraepithelial neoplasia (PIN) is hypoxic, and HIF1A signaling in luminal cells enhances malignant progression by suppressing immune surveillance and promoting luminal plasticity, leading to the emergence of cells that impair androgen signaling [[133]30]. Protein kinase C (PKC) encompasses a family of serine- and threonine-specific kinases that are activated by calcium and diacylglycerol. As key receptors for tumor-promoting phorbol esters, PKC family members display unique expression profiles and contribute to various cellular functions, including adhesion, transformation, cycle checkpoints, and volume regulation. Aberrant PKC expression is a well-recognized cancer hallmark, with elevated levels linked to enhanced cell proliferation and diminished apoptosis in several malignancies, such as bladder cancers [[134]31], gliomas, and PCas [[135]32]. Aggressive PCa cells with high PKCα expression require this for mitogenic activity [[136]33]. Apolipoprotein E (APOE) is a protein-coding gene.APOE is a core component of plasma lipoproteins and is involved in their production, transformation, and clearance [[137]34], Venanzoni MC et al. examined protein expression in 20 prostatectomy specimens by immunohistochemistry and determined the association between the Gleason score of each sample and ApoE expression. ApoE expression was positively associated with Gleason score, hormone independence, and both local and distant invasiveness in prostate tissue sections. In contrast, while ApoE was positive in prostate intraepithelial neoplasia (PIN) adjacent to clinically evident cancer, more distant PINs showed a negative expression for ApoE [[138]35]. Additionally ApoE gene scores were performed by blood samples from patients with prostate tumors. The E3/E3 genotype was found at a significantly higher frequency in patients compared to controls (P = 0.004). Carriers of the E3/E3 genotype had a 3.6-fold increased likelihood of being patients compared to controls (OR = 3.67, 95% CI = 1.451–9.155; p = 0.004). Moreover, patients with the E3/E3 genotype exhibited significantly higher Gleason scores (p = 0.017) and a greater prevalence of Gleason scores above 7 (P = 0.007). In contrast, the E4 allele was more prevalent in the control group (P = 0.006)(26,851,028). Heat shock protein family A (Hsp70) member 9 (HSPA9) is a chaperone protein crucial for mitochondrial iron-sulfur cluster (ISC) biogenesis. HSPA9 interacts with and stabilizes ISC cluster assembly proteins, including FXN, NFU1, NFS1, and ISCU [[139]36]. HSPA9 regulates erythropoiesis by stabilizing ISC assembly and may also play a role in controlling cell proliferation and cellular senescence [[140]36, [141]37]. It has been shown that JG-70, a variant inhibitor of HSP98, inhibits aerobic respiration by targeting mitochondrial HSP70 (HSPA9) and re-sensitizes desmoplasia-resistant PCas to androgen deprivation drugs in addition to Hirth CG et al. retrospectively reviewed the records of 636 patients who underwent radical prostatectomy and mounted paraffin embedded adenocarcinomatous and non-tumor tissues for microarrays. We evaluated the ability of HSPA9 to predict postoperative PSA outcome, response to adjuvant/rescue therapy and systemic disease. Results showed that HSPA9 was diffusely expressed in tumor cells and that diagnostic HSPA9 staining helped identify patients at increased risk of recurrence after salvage therapy [[142]38]. Serpin family E member 1 (SERPINE1), also known as plasminogen activator inhibitor 1 (PAI-1), inhibits tissue-type plasminogen activator (tPA) and urokinase (uPA). These enzymes convert plasminogen into plasmin, which in turn activates matrix metalloproteinases (MMPs) to degrade the extracellular matrix (ECM), thereby promoting invasion and metastasis. SERPINE1 blocks cancer cell invasion by inhibiting uPA protease activity. Additionally, knocking down six transmembrane epithelial antigens (STEAP2) in PCa cells upregulates SERPINE1, reducing their invasive potential. This indicates that SERPINE1 may serve as a downstream effector of certain oncogenes to regulate prostate cancer cell migration [[143]39]. Additionally epigenetic changes in the remaining biomarkers are thought to be associated with severe subtypes of prostate tumors and metastatic, invasive capacity. This includes mutations in CDKN2A [[144]40] as well as large amounts of unclipped TP53BP1 [[145]41] and hypermethylated CDKN2B [[146]42]. In order to more systematically analyze the mutations in the genes of patients with prostate tumors and to explore their correlation with the prognostic models predicted by the 12-Agings Prostate Tumor Patient's Prognostic Model for high and low risk patients. We obtained patient mutation data from the TCGA database and conducted analyses across various omics levels, including genomic and copy number levels. The analysis revealed that single nucleotide variants (SNVs) were the most common mutations in the cohort, with single nucleotide polymorphisms (SNPs) being the predominant type. Additionally, we identified the genes with the highest mutation frequencies. Subsequently, we analyzed the proportion and type of homozygous versus heterozygous mutations among the copy number variants (CNVs) in the sample. We conducted Spearman correlation analysis to explore the relationship between CNVs and gene expression. Moreover, significant correlations were found between the expression of senescence biomarkers and drug sensitivity in the Cancer Treatment Response Portal (CTRP) and the Genomics of Cancer Drug Sensitivity (GDSC) databases. These results suggest that our risk markers could potentially serve as predictors of chemotherapy drug sensitivity or be targeted in future drug development efforts. However, our prognostic model still has a lot of shortcomings, including the lack of real clinical cases and the lack of in vivo and ex vivo experiments to validate the expression and enrichment of the corresponding genes and pathways with the progression of the disease. These further experimental studies will be discussed in our subsequent papers. In conclusion, our proposed 12-Agings signature is a novel biomarker with significant potential for predicting patient prognosis and serving as a therapeutic target in prostate tumor patients. The 12-Agings signature is capable of predicting clinical outcomes, thereby assisting physicians in identifying cases that are at risk of deterioration and recurrence. Moreover, it can characterize the immune environment of prostate tumors, enabling a more precise stratification of patients and the development of individualized treatment plans. Additionally, the signature facilitates the early identification of patient subgroups that may benefit from immunotherapy and chemotherapy based on mRNA expression profiles. These capabilities underscore the potential of the 12-Agings signature to improve clinical decision-making and patient management in prostate tumors. Conclusion In conclusion, our study presents a robust prognostic tool for PRAD utilizing ARGs. Validated through LASSO regression and cross-validation, the risk score model identified 12 pivotal genes that illuminate the molecular mechanisms underlying PRAD. High-risk genes were positively correlated with increased risk, whereas low-risk genes were negatively correlated. These findings are anticipated to enhance PRAD treatment and facilitate the development of more targeted therapeutic strategies. Limitations Although the GEO and TCGA datasets have been meticulously curated in terms of scale and quality, the diversity of their origins may introduce sample heterogeneity, potentially affecting the generalizability of our findings. Additionally, since these datasets were generated by different research centers, variations in sample collection and processing methods might result in batch effects. Despite implementing appropriate bioinformatics techniques to mitigate these issues, some inherent variability may still influence the interpretation of the results. Contributions Our study makes a significant contribution to PRAD research by introducing a novel prognostic model rooted in ARGs. Validated by rigorous statistical methods, the model outperforms traditional clinicopathologic factors and provides greater accuracy for PRAD prognosis. The identification of 12 key genes provided valuable insights into the molecular mechanisms that drive PRAD prognosis. By demonstrating the robustness and clinical relevance of the model, we facilitate more informed therapeutic decisions for patients with PRAD, potentially enabling personalized treatment. Furthermore, our work highlights the importance of exploring ARGs in cancer prognosis, paving the way for future research in this critical area of oncology. Author contributions The study was conceptualized by YW, RX, and JW. ZL, YW and RX were responsible for drafting the manuscript. YW and ZL conducted the literature search and gathered the relevant data. Subsequently, YW and JW analyzed and presented the data in a visual format. The final version of the manuscript was reviewed by ZL and JW, and necessary revisions were made. All authors critically evaluated and provided their approval for the final version of the manuscript. Funding This study did not receive any specific grants from funding agencies in the public, commercial, or nonprofit sectors. Data availability The datasets employed in this study can be accessed through the GEO repository ([147]https://www.ncbi.nlm.nih.gov/geo/) and the TCGA portal ([148]https://portal.gdc.cancer.gov/). Additionally, the raw data files, code files, and images supporting this research are available for download via the following link: [149]https://www.jianguoyun.com/p/DY7CFbEQkeKyCxiQvaAFIAA. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Jing Wang, Email: raulkxlys@163.com. Zhibin Luo, Email: luozhibincq@126.com. References