Identification of programmed cell death associated key genes in benign prostatic hyperplasia and prostate cancer development by integrated bioinformatics analysis and machine learning

Introduction

Benign prostatic hyperplasia (BPH) and prostate cancer (PCa) are among the most prevalent urological disorders affecting aging men, posing significant challenges to global healthcare systems. While BPH is a common cause of lower urinary tract symptoms that impair quality of life, PCa remains a leading cause of cancer-related morbidity and mortality worldwide (1,2). Despite their distinct pathological entities, BPH and PCa share some common risk factors, such as aging (3) and hormonal imbalances (4). Notably, the dysregulation of programmed cell death (PCD), a crucial process for maintaining tissue homeostasis, has been implicated in the pathogenesis of both conditions (5). However, the underlying molecular mechanisms, especially the specific PCD-related genes driving the development and progression of BPH and PCa, are not fully elucidated.

The emergence of high-throughput technologies and public genomic databases, such as the Gene Expression Omnibus (GEO), has provided unprecedented opportunities for uncovering molecular features of complex diseases. Bioinformatics approaches have been widely applied to identify diagnostic biomarkers and therapeutic targets for BPH (6) and PCa (7).

Therefore, in this study, we employed an integrated bioinformatics approach combining multiple GEO datasets (GSE119195 and GSE55597) with diverse machine learning algorithms [including least absolute shrinkage and selection operator (LASSO), XGBoost, random forest, and Boruta] to identify robust PCD-related hub genes. We hypothesize that this multi-faceted strategy can uncover novel and reliable diagnostic biomarkers and therapeutic targets for BPH and PCa, overcoming the limitations of single-dataset or single-method analyses. We present this article in accordance with the TRIPOD reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-527/rc).

Methods

Data acquisition and processing

We accessed the Gene Expression Omnibus (GEO) database to retrieve GEO datasets GSE55597 and GSE119195. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Validation sets were generated using RNA sequencing data and matched prostate adenocarcinoma (PRAD) clinical data from the Cancer Genome Atlas (TCGA) database. Differential expression analysis was performed using the Limma package in R language. Genes with a P<0.05 and an absolute log2 fold change (|log2FC|) >0.5 were considered statistically significant differentially expressed genes (DEGs). We visualized DEGs in a volcano plot using ggvolcano package in R, intersected DEGs from both databases using ggvenn package, and created a heatmap of DEGs using pheatmap package in R. Finally, we cross-referenced the PCD gene with the resulting intersection. The specific version numbers for all R packages used can be found in Table S1.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology (GO) enrichment analysis

GO enrichment and KEGG pathway analysis were carried out using the clusterProfiler package in R to determine the DEGs enrichment differences between BPH and normal groups, as well as between BPH and PCa groups.

Immune infiltration analysis

For the 22 different types of immune cell composition of prostate tissue in the GSE119195 and GSE55597, immune infiltration analysis was performed in conjunction with signature matrix. Mixture expression file of the DEGs or the 15 co-significant genes was imported to conduct the immune infiltration analysis with the IOBR package.

Protein-protein interaction (PPI) network analysis

STRING established the PPI network. In short, the DEGs were uploaded to the STRING database. Then, the interaction file including the source nodes and target nodes was imported into Cytoscape software to eliminating the intersection of DEGs and PCD using the cytohubba plug-in unit.

Identification of hub genes using machine learning

To further refine the list of candidate genes, we employed four distinct machine learning algorithms on the 15 PCD-related DEGs, we applied four complementary machine-learning approaches in a unified framework. First, we fitted a 10-fold cross-validated LASSO with the glmnet package, selecting the penalty at lambda.min and retaining genes with non-zero coefficients at that value. We then trained a Random Forest using the randomForest package with ntree =500 and ranked variables by Mean Decrease Gini, carrying forward the highest-importance genes. In parallel, we implemented XGBoost with the xgboost package and prioritized features by gain-based importance. We also ran Boruta, a Random-Forest wrapper, and kept features flagged as Confirmed by the algorithm. Final hubs were defined by cross-method consensus, requiring appearance in at least three of the four models, which yielded BMP5 and CYP1B1 for downstream analyses.

Construction and validation of the nomogram

A nomogram was constructed based on the expression levels of the final hub genes (BMP5 and CYP1B1) using the rms package in R. The predictive performance of the nomogram was assessed by its discrimination and calibration abilities. Discrimination was evaluated by calculating the area under the curve (AUC) using the pROC package. Calibration was visually assessed using calibration curves that compared predicted probabilities with observed outcomes.

Predicted gene-miRNA analysis

The miRWalk 2.0 database was used to examine gene-miRNA interactions. The miRWalk databases were used to predict the genes that were chosen to target miRNAs. Additionally, Cytoscape was used to visualize the interaction network.

Survival analysis and analysis of The Human Protein Atlas (HPA)

We set disease-free survival (DFS) as prognostic indicators. The survival package and survminer package was used to analyze the prognosis of PCa based on the gene signature that was obtained above. The HPA was used to validate potential gene candidates at the protein level.

Statistical analysis

All statistical analyses were performed using R software (version 4.3.1). Survival analysis was performed using the Kaplan-Meier method, and differences in survival curves between high- and low-expression groups (stratified by the median expression level) were compared using the log-rank test. A two-sided P<0.05 was considered statistically significant in all analyses unless otherwise specified.

Result

Identification of DEGs and visualization of differential genes in heat map

We identified 542 up-regulated and 791 down-regulated DEGs in the GSE55597 dataset, and 288 up-regulated and 434 down-regulated DEGs in the GSE119195 dataset, based on the criteria of |log2FC| >0.5 and a P<0.05 (Figure S1A,S1B). In addition, we obtained 159 hub genes by taking the intersection of DEGs from two datasets GSE55597 and GSE119195 (Figure S1C). Moreover, the DEGs were visualized on the following heatmaps: all DEGs in GSE119195 (Figure S2A); the top 25 DEGs in GSE119195 (Figure S2B); all DEGs in GSE55597 (Figure S2C); and the top 25 DEGs in GSE55597 (Figure S2D).

DEGs functional annotation and KEGG pathway analysis

In the GSE119195 dataset, as depicted in Figure S3A,S3B, these findings suggest that this pathological pathway plays a crucial role in BPH pathogenesis. In GSE55597, as shown in Figure S3C,S3D, it is worth noting that KEGG pathway analysis reveals that the DEGs are primarily associated with Focal adhesion, cAMP signaling pathway, Tight junctions, Proteoglycans in cancer, and Wnt signaling pathway.

Expression landscape of the genes of PCD patterns among different samples

Fifteen hub genes were obtained by intersecting 159 DEGs with PCD genes (Figure 1A). The various genes were discovered to be highly closely connected and to typically operate as a complex (Figure 1B). Obtain the top ten nodes ranked by using three different assays (including MCC, MNC, DMNC) in cytohubba (Figure 1C-1E), which is a plugin in Cytoscape software.

Figure 1 Identification of 15 PCD-related hub genes from common DEGs. (A) Venn diagram illustrating the intersection between 159 common DEGs (from GSE55597 and GSE119195) and a curated list of PCD-associated genes, resulting in 15 candidate hub genes. (B) PPI network of the 15 hub genes, constructed using the STRING database and visualized in Cytoscape. Nodes represent proteins, and edges represent predicted functional associations. The top ten hub genes within the PPI network ranked by three different maximal clique centrality algorithms in CytoHubba: (C) MCC, (D) MNC, and (E) DMNC. The color intensity of the nodes corresponds to the ranking score, with red indicating the highest rank. DEGs, differentially expressed genes; DMNC, density of maximum neighborhood component; MCC, maximal clique centrality; MNC, maximum neighborhood component; PCD, programmed cell death; PPI, protein-protein interaction.

PCD patterns as potential biomarkers for BPH

LASSO regression method was used to construct a model for the GSE55597 dataset, and six genes were screened out (Figure 2A,2B). After that, the Random Forest method is used to build the model, and the top ten variables are screened out, among which BMP5 is the most important (Figure 2C). Then Boruta’s method was used to construct a model, which screened out 12 confirmed important variables, 2 confirmed unimportant variables and 1 tentative attribute (Figure 2D). Then, the model was constructed using XGboost method, and six genes were selected, among which QSOX1, RIPK2, NR4A2, TRIM22 and DEPTOR belonged to the same cluster, while BMP5 belonged to a cluster alone, and BMP5 was still the most important variable (Figure 2E). We constructed a nomogram incorporating the expression levels of BMP5 and CYP1B1 to assess their combined predictive potential for discriminating between PCa (Figure 2F). A similar method was used to analyze the GSE199195 dataset, and it was found that LASSO regression screened out 5 genes (Figure 2G,2H), and Random Forest screened out the top 10 variables (Figure 2I). Final hubs were defined by cross-method consensus, requiring appearance in at least three of the four models, which yielded BMP5 and CYP1B1 for downstream analyses.

Figure 2 Comprehensive analysis of machine learning-based feature selection and diagnostic model validation. (A,B) LASSO coefficient profiles of fifteen hub genes in GSE55597. (C) Utilizing the random forest algorithm, determine which top 10 important genes among the 15 hub genes in GSE55597. (D) The importance of the 15 hub genes was calculated using the XGboost algorithm. (E) The importance of the 15 hub genes was calculated using the Boruta algorithm. (F) Nomogram of the two key genes in GSE55597. The value of each gene was given a score on the point scale axis. Every single score could be easily added up to a total score, and we could estimate the probability of PCa by projecting the total score to the lower total point scale. (G,H) LASSO coefficient profiles of fifteen hub genes in GSE119195. (I) Utilizing the random forest algorithm, determine which top 10 important genes among the 15 hub genes in GSE119195. (J) Nomogram of the two key genes in GSE119195. The value of each gene was given a score on the point scale axis. Every single score could be easily added up to a total score, and we could estimate the probability of BPH by projecting the total score to the lower total point scale. (K) ROC curve analysis of nomogram in GSE55597. (L) ROC analysis of nomogram in GSE119195. AUC, area under the curve; BPH, benign prostatic hyperplasia; LASSO, least absolute shrinkage and selection operator; PCa, prostate cancer; ROC, receiver operating characteristic.

The nomogram for predicting the risk of BPH and PCa

We constructed a nomogram incorporating the expression levels of BMP5 and CYP1B1 to assess their combined predictive potential for discriminating BPH (Figure 2J). This model demonstrated high discriminatory power in the internal validation. The risk score derived from the nomogram achieved an AUC of 0.988 in the GSE55597 dataset and an AUC of 1.0 in the GSE119195 dataset (Figure 2K,2L). It is important to note that the perfect AUC (1.0) in the GSE119195 dataset, while indicating excellent performance on this specific dataset, may reflect model overfitting due to the relatively limited sample size and warrants further validation in larger, independent cohorts.

The immune infiltration for 22 types of immune cell composition of prostate tissue in GSE119195 and GSE55597

The immune infiltration analysis revealed distinct immune landscapes across the groups (Figure S3E). Notably, the proportion of T helper cells was highest in healthy controls and exhibited a progressive decrease in BPH and PCa samples. This observed decline suggests a potential weakening of adaptive immune surveillance during prostate disease progression. The correlation patterns between 22 immune cells also differed significantly between BPH and PCa (Figure S4A). Interestingly, BMP5 and CYP1B1 expression showed predominantly negative correlations with immune cells in PCa but positive correlations in BPH (Figure 3), hinting at complex, disease-specific roles for these genes in the tumor immune microenvironment. The calibration curves also showed satisfactory agreement between predicted and observed probabilities (Figure S4B,S4E).

Figure 3 Analysis of correlations between key gene expression and immune cell infiltration. (A) Correlation between the expression levels of BMP5 and CYP1B1 and immune cells in GSE55597. (B) Correlation between the expression levels of BMP5 and CYP1B1 and immune cells in GSE1119195. *, P<0.05; ***, P<0.001. MDSC, myeloid-derived suppressor cell.

Further miRNA mining and interaction network analysis

Figure S5 illustrates the relationship between two core genes and the miRNAs they target; CYP1B1 was linked to 38 miRNAs, while BMP5 was linked to 7 miRNAs. However, the miRNA of these two genes did not intersect.

The nomogram for predicting the prognosis of PCa

BMP5 expression in BPH was significantly higher than in PCa, and CYP1B1 expression in BPH was significantly lower than in PCa (Figure 4A). Survival analysis indicated that low expression levels of CYP1B1 and BMP5 were associated with a trend towards lower DFS in PCa patients (Figure 4B). Univariate Cox regression analysis identified a significant association between lower CYP1B1 expression and poorer DFS (Figure 4C). Importantly, in multivariate analysis adjusted for clinical covariates, CYP1B1 expression remained an independent risk factor (Figure 4D), suggesting its potential prognostic value. Figure 5A showed that CYP1B1, Gleason score and tumor (T) in tumor-node-metastasis (TNM) classification were significantly associated with the prognosis of PCa patients. In addition, although the calibration curve and the correction curve have a certain deviation, they are still relatively consistent (Figure 5B). Consistent with our mRNA findings, immunohistochemical data from the Human Protein Atlas (HPA) confirmed that CYP1B1 protein expression was elevated in PCa tissues compared to normal prostate tissues (Figure 5C,5D), providing supportive evidence at the protein level.

Figure 4 Validation of key gene expression and their prognostic value in prostate cancer. (A) The comparison of the expression of BMP5 and CYP1B1 in GSE55597 and in GSE119195. (B) Survival analysis of CYP1B1 and BMP5 associated with prostate cancer using Kaplan Meier method. (C) Univariate cox regression analysis of 5 key genes. (D) Multivariate cox regression analysis of 5 key genes. BPH, benign prostatic hyperplasia; CI, confidence interval; PCa, prostate cancer.

Figure 5 Construction of a prognostic nomogram for prostate cancer and validation of CYP1B1 protein expression. (A) A prognostic nomogram for predicting 1-, 3-, and 5-year DFS probability in prostate cancer patients. The nomogram integrates clinical factors (Gleason score, T stage) and the expression level of the key gene CYP1B1. Points are assigned for each variable, and the total points correspond to the probability of DFS. (B) Calibration curves for the prognostic nomogram. The 45-degree dotted line represents perfect prediction. The solid lines showed the agreement between the nomogram-predicted probability of DFS and the actual observed outcome for 1-, 3-, and 5-year survival. (C,D) Immunohistochemistry images obtained from The Human Protein Atlas (HPA) database showing CYP1B1 protein expression in (C) normal prostate tissue (image available from https://www.proteinatlas.org/ENSG00000138061-CYP1B1/tissue/prostate#img) and (D) prostate cancer tissue (image available from https://www.proteinatlas.org/ENSG00000138061-CYP1B1/cancer/prostate+cancer#img). The scale bar indicates 100 µm. *, P<0.05; **, P<0.01. DFS, disease-free survival.

Discussion

For middle-aged and older men, BPH is one of the most prevalent urinary system health issues. The lack of a comprehensive understanding of the pathophysiology and origin of BPH has hindered the development of novel and efficient treatments for the condition. Previous research suggests that BPH may be related to growth factors (such as insulin-like growth factor, fibroblast growth factor, and epidermal growth factor), androgens (8), estrogens (9), and insulin as well as specific hereditary risk factors. In the prostate, insulin-like growth factor 1 (IGF-1) primarily mediates the effects of insulin. The present investigation revealed a statistically significant increase in IGF-1 expression in BPH samples compared to normal prostate tissue. The study has demonstrated that IGF-1 can enhance BPH stromal cell growth (10). Nonetheless, there are few data on biomarkers for BPH, and the precise mechanisms and pathological connections remain unclear.

This study, through integrated bioinformatics and multiple machine learning algorithms, aimed to identify key PCD-associated genes relevant to BPH and PCa. We identified 159 common DEGs across two independent datasets. Intersection with a PCD gene set ultimately highlighted two of the most promising candidate genes: BMP5 and CYP1B1. A predictive model based on these genes demonstrated excellent discriminatory power in internal validation and was significantly associated with patient prognosis in PCa. The following sections will discuss these findings in greater depth.

In this study, GO and KEGG pathway enrichment analysis were performed for the two DEGs. According to Xu et al. (11), oxytocin can stimulate prostate proliferation by activating the MEK-ERK-RSK pathway. Furthermore, it has been demonstrated that the Wnt signaling pathway activates the EMT, which causes prostatic hyperplasia, and MicroRNA-340 can reverse EMT by impairing this pathway (12). Although the specific study has not yet been confirmed, suggest that Axon guidance may also be connected to the development of BPH. Conversely, focal adhesion, the cAMP signaling pathway, and tight junctions might be crucial in the initiation and advancement of PCa. Since the mTOR and MAPK signaling pathways are unique to GSE119195, they might also contribute to the onset and progression of BPH. Liu et al. demonstrated that overexpression of STEAP4 inhibited apoptosis of prostate cells through the AKT/mTOR signaling pathway, leading to the development of BPH (13). Similarly, Shi et al. (14) discovered that prostate cell proliferation could be boosted by TRAF6 expression upregulated in the prostatic stroma, an effect mediated by Akt/mTOR signaling. Furthermore, an imbalance between prostatic proliferation and apoptosis may be linked to aberrant coordination of the MAPK cascade. Currently, there is proof linking BPH to both the p38MAPK and ERK (1/2) cascades (15,16).

Our findings are partially consistent with previous research and provide new perspectives on the roles of BMP5 and CYP1B1 in prostate diseases. Luo et al. reported consistent upregulation of BMP5 in BPH patients via cDNA microarray analysis, which aligns with our results (17). Middleton et al. suggested that BMP5 might promote the proliferation of prostate epithelial cells potentially through the epithelial-mesenchymal transition (EMT) (18). Our enrichment analysis implicating pathways like MAPK and Wnt provides a broader signaling context for BMP5’s mechanism of action. Regarding CYP1B1, existing studies report significantly higher expression in PCa compared to BPH (19), and its polymorphism is associated with PCa risk (20); our analytical results strongly support these findings. More importantly, our survival analysis further revealed the potential clinical value of low CYP1B1 expression being associated with poorer prognosis. This seemingly paradoxical phenomenon (high expression in cancer tissue yet association with better survival) hints at the complex functionality of CYP1B1, warranting further investigation.

A core innovation of our study lies in linking the filtered genes to PCD. The process of PCD is crucial for maintaining homeostasis. One well-known example is apoptosis, which is involved in both cell development and the elimination of aberrant and senescent cells in order to preserve the proper ratio of cells in the body (21). Liu et al. found that NELL2 depletion induced mitochondria-dependent apoptosis, whereas overexpression of NELL2 promoted cell proliferation and inhibited cell apoptosis (22), and that NELL2 was highly up-regulated in patients with BPH (23). How might BMP5 and CYP1B1 interact with PCD pathways? First, as a member of the TGF-β superfamily, BMP5 has been widely implicated in regulating apoptosis. The imbalance between proliferation and apoptosis in stromal and epithelial cells is a recognized pathological basis of BPH (24,25). We speculate that aberrant overexpression of BMP5 may inhibit normal apoptosis in prostate cells by interfering with canonical Smad signaling or crosstalk with non-canonical pathways like MAPK, thereby promoting hyperplastic lesions (22). On the other hand, the mechanism of CYP1B1, as a metabolic enzyme, might be more indirect yet equally critical. It is involved in estrogen metabolism, generating genotoxic quinone metabolites that lead to oxidative stress and DNA damage—common triggers for PCD events such as apoptosis and pyroptosis (26). In PCa, persistently high CYP1B1 expression might create constant genotoxic stress, potentially selecting for cancer cell clones with anti-apoptotic capabilities. This could be a potential mechanism explaining its association with a trend towards poorer prognosis. Certainly, these speculations require validation through subsequent functional experiments.

The PPI interaction network was built using fifteen hub genes via the STRING platform. The visualization results showed that IGF-1, followed by CYP1B1, occupied the central position in the PPI interaction network. It is debatable whether IGF-1 can accelerate the onset of BPH, though. Relevant research has demonstrated that high levels of IGF-1 and insulin raise the risk of BPH and could even be used to predict prostate size (10). Paradoxically, Mantzoros et al. (27) found no association between IGF-1 and BPH, but increased IGF-1 expression was associated with an increased risk of PCa. Further studies may be needed to verify the relationship between IGF-1 and BPH in the future.

Our immune infiltration analysis revealed a potentially important immune dynamic: the infiltration proportion of T helper cells showed a progressive decrease from healthy prostate to BPH to PCa. This phenomenon may hold significant biological relevance. T helper cells play a central role in activating anti-tumor immune responses. Their early reduction in the BPH stage might indicate that the local immune microenvironment is already shifting towards an immunosuppressive state. This “immunoediting” process could create conditions favorable for the subsequent malignant transformation of epithelial cells. The opposite correlations of BMP5/CYP1B1 with immune cell infiltration in BPH versus PCa strongly suggest that these two genes may modulate the tumor immune microenvironment in a disease-specific manner, offering a new perspective for understanding their distinct roles in different disease contexts.

Despite the valuable insights provided by our study, its inherent limitations must be acknowledged. First, this is a purely bioinformatics analysis, and all conclusions are based on transcriptomic data from public databases. The lack of protein-level validation (e.g., by Western Blot or immunohistochemistry) and confirmation through in vitro or in vivo functional experiments is a major limitation. Second, the sample size for BPH analysis is relatively limited, and the absence of a completely independent external clinical cohort to validate our diagnostic model poses a risk of overfitting; the generalizability of the model needs further assessment. The AUC of 1.0 in the GSE119195 dataset, while encouraging, should be interpreted with particular caution. Finally, this study primarily focuses on mRNA expression, while post-transcriptional regulation and protein modifications were not considered. Therefore, future work should focus on: (I) validating the diagnostic and prognostic value of BMP5 and CYP1B1 in prospective, large cohorts incorporating more clinicopathological parameters; (II) elucidating the specific molecular mechanisms by which these genes regulate PCD and influence disease progression using prostate cell lines and animal models via gain-of-function and loss-of-function experiments; and (III) confirming their expression and function at the protein level.

Conclusions

This study utilized an integrated bioinformatics approach to analyze publicly available transcriptomic data (GSE119195) to investigate the mechanisms of BPH. The functional enrichment and pathways of DEGs in BPH were examined in this study using an integrated bioinformatics methodology. MAPK signaling pathway, Oxytocin signaling pathway, mTOR signaling pathway and Wnt signaling pathway may play a key role in the occurrence and development of BPH. Furthermore, BMP5 and CYP1B1 are strong indicators of BPH and the risk factors of PCa, thus they are promising biomarkers for BPH and PCa. It is important to note that these conclusions are derived from computational analysis, and further experimental validation in clinical settings is necessary to confirm their clinical applicability.

Acknowledgments

We would like to thank the TCGA and GEO databases for providing publicly accessible data.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-527/rc

Peer Review File: Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-527/prf

Funding: This study was funded by the National Natural Science Foundation of China (Grant Number 82000721) and Program from the Department of Science and Technology of Sichuan Province (Grant Number 22NSFSC1750).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-527/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet 2018;392:1789-858. [Crossref] [PubMed]
3. Nichols JH, Loeb S, Metter EJ, et al. The relationship between prostate volume and prostate-specific antigen variability: data from the Baltimore Longitudinal Study of Aging and the Johns Hopkins Active Surveillance Program. BJU Int 2012;109:1304-8. [Crossref] [PubMed]
4. Welén K, Damber JE. Androgens, aging, and prostate health. Rev Endocr Metab Disord 2022;23:1221-31. [Crossref] [PubMed]
5. Zhu M, Liu D, Liu G, et al. Caspase-Linked Programmed Cell Death in Prostate Cancer: From Apoptosis, Necroptosis, and Pyroptosis to PANoptosis. Biomolecules 2023;13:1715. [Crossref] [PubMed]
6. Liu D, Shoag JE, Poliak D, et al. Integrative multiplatform molecular profiling of benign prostatic hyperplasia identifies distinct subtypes. Nat Commun 2020;11:1987. [Crossref] [PubMed]
7. Jiang H, Liu M, Deng Y, et al. Identification of prostate cancer bone metastasis related genes and potential therapy targets by bioinformatics and in vitro experiments. J Cell Mol Med 2024;28:e18511. [Crossref] [PubMed]
8. Song L, Shen W, Zhang H, et al. Differential expression of androgen, estrogen, and progesterone receptors in benign prostatic hyperplasia. Bosn J Basic Med Sci 2016;16:201-8. [Crossref] [PubMed]
9. McPherson SJ, Hussain S, Balanathan P, et al. Estrogen receptor-beta activated apoptosis in benign hyperplasia and cancer of the prostate is androgen independent and TNFalpha mediated. Proc Natl Acad Sci U S A 2010;107:3123-8. [Crossref] [PubMed]
10. Sreenivasulu K, Nandeesha H, Dorairajan LN, et al. Elevated insulin and reduced insulin like growth factor binding protein-3/prostate specific antigen ratio with increase in prostate size in Benign Prostatic Hyperplasia. Clin Chim Acta 2017;469:37-41. [Crossref] [PubMed]
11. Xu H, Fu S, Chen Y, et al. Oxytocin: its role in benign prostatic hyperplasia via the ERK pathway. Clin Sci (Lond) 2017;131:595-607. [Crossref] [PubMed]
12. Chen SY, Du Y, Song J. MicroRNA-340 Inhibits Epithelial-Mesenchymal Transition by Impairing ROCK-1-Dependent Wnt/β-Catenin Signaling Pathway in Epithelial Cells from Human Benign Prostatic Hyperplasia. Chin Med J (Engl) 2018;131:2008-12. [Crossref] [PubMed]
13. Liu J, Zhou W, Yang L, et al. STEAP4 modulates cell proliferation and oxidative stress in benign prostatic hyperplasia. Cell Signal 2024;113:110933. [Crossref] [PubMed]
14. Shi YF, Yu DJ, Jiang CY, et al. TRAF6 regulates proliferation of stromal cells in the transition and peripheral zones of benign prostatic hyperplasia via Akt/mTOR signaling. Prostate 2018;78:193-201. [Crossref] [PubMed]
15. Zhou Z, Wu L, Zhao S, et al. Upregulation of FGF7 Induced Intravesical Prostatic Protrusion of Benign Prostatic Hyperplasia via the ERK1/2 Signaling Pathway. Gerontology 2023;69:615-27. [Crossref] [PubMed]
16. Li C, Hu WL, Lu MX, et al. Resveratrol induces apoptosis of benign prostatic hyperplasia epithelial cell line (BPH-1) through p38 MAPK-FOXO3a pathway. BMC Complement Altern Med 2019;19:233. [Crossref] [PubMed]
17. Luo J, Dunn T, Ewing C, et al. Gene expression signature of benign prostatic hyperplasia revealed by cDNA microarray analysis. Prostate 2002;51:189-200. [Crossref] [PubMed]
18. Middleton LW, Shen Z, Varma S, et al. Genomic analysis of benign prostatic hyperplasia implicates cellular re-landscaping in disease pathogenesis. JCI Insight 2019;5:e129749. [Crossref] [PubMed]
19. Tokizane T, Shiina H, Igawa M, et al. Cytochrome P450 1B1 is overexpressed and regulated by hypomethylation in prostate cancer. Clin Cancer Res 2005;11:5793-801. [Crossref] [PubMed]
20. Tanaka Y, Sasaki M, Kaneuchi M, et al. Polymorphisms of the CYP1B1 gene have higher risk for prostate cancer. Biochem Biophys Res Commun 2002;296:820-6. [Crossref] [PubMed]
21. Moyer A, Tanaka K, Cheng EH. Apoptosis in Cancer Biology and Therapy. Annu Rev Pathol 2025;20:303-28. [Crossref] [PubMed]
22. Liu J, Liu D, Zhang X, et al. NELL2 modulates cell proliferation and apoptosis via ERK pathway in the development of benign prostatic hyperplasia. Clin Sci (Lond) 2021;135:1591-608. [Crossref] [PubMed]
23. Ke ZB, Cai H, Wu YP, et al. Identification of key genes and pathways in benign prostatic hyperplasia. J Cell Physiol 2019;234:19942-50. [Crossref] [PubMed]
24. Minutoli L, Rinaldi M, Marini H, et al. Apoptotic Pathways Linked to Endocrine System as Potential Therapeutic Targets for Benign Prostatic Hyperplasia. Int J Mol Sci 2016;17:1311. [Crossref] [PubMed]
25. Zhang X, Zhang Q, Zhang Z, et al. Apoptosis profiles in benign prostatic hyperplasia: close associations of cell kinetics with percent area density of histologic composition. Urology 2006;68:905-10. [Crossref] [PubMed]
26. Mosli HA, Tolba MF, Al-Abd AM, et al. Catechol estrogens induce proliferation and malignant transformation in prostate epithelial cells. Toxicol Lett 2013;220:247-58. [Crossref] [PubMed]
27. Mantzoros CS, Tzonou A, Signorello LB, et al. Insulin-like growth factor 1 in relation to prostate cancer and benign prostatic hyperplasia. Br J Cancer 1997;76:1115-8. [Crossref] [PubMed]