Investigating key genes and molecular mechanisms of prostate cancer and coronary heart disease through transcriptomics and experimental validation

Introduction

Background

Prostate cancer (PC) is the most common malignant tumor affecting the male genitourinary tract, and its incidence rises markedly with age. Its development is multifactorial, involving genetic predisposition, hormonal changes, aging, dietary habits, lifestyle factors, chronic inflammation, and environmental exposures (1,2). PC is the second most common cancer among men worldwide, and its incidence is continuously increasing (3). The early symptoms of PC are often not obvious and PC is usually detected only at advanced stages, imposing a heavy financial burden on patients, severely affecting their quality of life and mental health, and placing immense pressure on healthcare resources (4,5). Currently, treatment options for PC include surgery, radiotherapy, endocrine therapy, and chemotherapy. However, diagnosis still faces many challenges, such as the limitations of prostate-specific antigen (PSA) testing, the invasiveness of diagnostic procedures, and tumor heterogeneity. These factors collectively increase the difficulty of early diagnosis and precision treatment (6,7).

Coronary heart disease (CHD) is a cardiovascular condition resulting from atherosclerotic narrowing or obstruction of the coronary arteries, which impairs blood flow and leads to myocardial ischemia, hypoxia, or even infarction. It represents the most prevalent form of heart disease (8). CHD ranks among the top causes of death globally, with its occurrence strongly linked to risk factors including age, gender, diabetes, hypertension, smoking, and hyperlipidemia (9,10). According to data from the World Health Organization, approximately 126 million people worldwide were affected by CHD in 2019, resulting in 8.9 million deaths, placing a tremendous burden on healthcare systems (11). Recent years have seen significant progress in understanding the interplay between cancer and cardiovascular disease. Cardio-oncology has emerged as a multidisciplinary field of increasing interest. It mainly investigates the bidirectional interactions between cancer and the cardiovascular system, as well as the prevention and management of cardiovascular toxicity related to cancer therapies (12).

Rationale and knowledge gap

In the study of comorbidity between PC and CHD, this interaction is particularly complex and has a significant impact on treatment and prognosis (13). Studies have demonstrated the critical role of the interaction between thyroid hormone and androgen signaling in the inflammation and progression of PC (14). However, androgen deprivation therapy (ADT), although a mainstay in the clinical management of PC, may increase the risk of cardiovascular disease (15). Gonadotropin-releasing hormone (GnRH) agonists are commonly used ADT drugs; however, their use has been associated with an elevated risk of cardiovascular events, including CHD, myocardial infarction, and sudden cardiac death (16). The underlying mechanisms may involve tumor volume increase triggered by testosterone surge, hypercoagulable states, and instability of atherosclerotic plaques (16). Additionally, CHD itself is considered a risk factor for PC, and the two may interact through shared mechanisms such as metabolic disorders and inflammatory responses (17). As major diseases that threaten human health, both PC and CHD present significant treatment challenges individually, and their interrelation further adds complexity to clinical management. However, most existing studies have concentrated on single diseases (18,19), and systematic investigations into comorbid conditions remain limited. Accordingly, enhanced multidisciplinary collaboration is urgently required to develop optimized comprehensive management strategies and thereby improve overall therapeutic outcomes.

Objective

This study conducted a cross-disease analysis using bioinformatics approaches. First, genes with differential expression were screened from the training dataset. Next, key genes were selected using machine learning algorithms, and utilizing the selected key genes, a nomogram model was established for predictive assessment. The study further analyzed the subcellular and chromosomal localization of the key genes, performed pathway enrichment and immune infiltration analyses, constructed a molecular regulatory network, predicted potential therapeutic drugs, and conducted molecular docking. Finally, reverse transcription quantitative polymerase chain reaction (RT-qPCR) experiments were conducted on clinical samples to confirm whether the expression patterns of the key genes aligned with the findings from the bioinformatics analysis. This study innovatively integrates transcriptomic data from two diseases, providing a scientific basis for uncovering the shared pathological mechanisms between PC and CHD, and for developing cross-disease diagnostic and therapeutic targets. We present this article in accordance with the TRIPOD reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-519/rc).

Methods

Data sources

PC and CHD gene expression data were acquired from the Gene Expression Omnibus (GEO) database. (https://www.ncbi.nlm.nih.gov/geo/). To be specific, PC training set 1 was the GSE70768 dataset (GPL10558 platform), which consisted of 113 prostate tissue samples from PC patients (PC group) and 73 control prostate tissue samples (control group). PC validation set 1 was the GSE88808 dataset (GPL22571 platform), holding 49 prostate tissue samples from PC patients (PC group) and 49 control prostate tissue samples (control group). Likewise, CHD training set 2 was the GSE113079 dataset (GPL20115 platform), made up of 93 peripheral blood mononuclear cell (PBMC) samples from CHD patients (CHD group) and 48 control PBMC samples (control group). CHD validation set 2 was the GSE66360 dataset (GPL570 platform), including 49 peripheral blood leukocyte samples from CHD patients (CHD group) and 50 control peripheral blood leukocyte samples (control group).

Analysis of differentially expressed genes (DEGs)

To identify the DEGs between the PC group and the control group, differential analysis (PC group versus control group) in training set 1 (GSE70768) was conducted using the “limma” package [version 3.54.0 (20)]. The screening criteria were set as P<0.05 and |log₂fold change (FC)| >0.5. For visual representation of these DEGs, the “ggplot2” package [version 3.4.1 (21)] was utilized to create a volcano plot, and the “pheatmap” package [version 1.0.12 (22)] was used to draw a heatmap. The top 10 most upregulated and top 10 most downregulated DEGs were selected for display based on |log₂FC|. For exploring the DEGs between the CHD group and the control group, differential analysis (CHD group versus control group) in training set 2 (GSE113079) was carried out with the “limma” package (version 3.54.0). The screening criteria were also P<0.05 and |log₂FC| >0.5. To visualize these DEGs, the “ggplot2” package (version 3.4.1) was employed to produce a volcano plot, and the “pheatmap” package (version 1.0.12) was used to generate a heatmap, presenting the top 10 most activated and top 10 most repressed DEGs2 selected for display based on |log₂FC|.

Acquisition and mechanistic exploration of candidate genes

To acquire candidate genes associated with PC and CHD, the “VennDiagram” package (version 1.7.1) (23) was utilized to conduct an intersection analysis of DEGs1 (PC) and DEGs2 (CHD). The activated genes of DEGs1 and DEGs2 were intersected, as were the repressed genes of DEGs1 and DEGs2. Subsequently, the union of these two sets of intersected genes was taken, and this gene union set was designated as the candidate genes.

To investigate the functions of the candidate genes and the biological processes they participate in, the “clusterProfiler” package (version 4.2.2) (24) was employed to carry out gene ontology (GO) functional enrichment analysis, encompassing three aspects: biological process (BP), molecular function (MF), and cellular component (CC) (P<0.05). Moreover, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was executed (P<0.05). Ultimately, the results of GO and KEGG were visualized by means of the “ggplot2” package (version 3.4.1).

To investigate protein-protein interaction (PPI) networks among the candidate genes,the online tool Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (https://string-db.org/) was utilized to build a PPI network for the candidate genes (with a confidence score >0.4). Subsequently, Cytoscape software (version 3.8.2) (25) was applied for visualization, and the maximal clique centrality (MCC) algorithm was used to pick out the candidate genes for subsequent analysis.

Identification of feature genes

To obtain the feature genes for PC and CHD, the following unified analysis workflow was adopted based on all samples from training set 1 (GSE70768) and training set 2 (GSE113079): the “Boruta” package (version 8.0.0) (26) was used to analyze the importance of the top 20 candidate genes screened by MCC. By comparing the importance of real features with randomly generated “shadow variables” and ranking them, genes with importance scores significantly higher than those of shadow variables were defined as Boruta1 (PC) and Boruta2 (CHD), respectively. Subsequently, the “Random Forest” package (version 4.7-1.2) (27) was employed for random forest (RF) analysis. Multiple training subsets were generated through repeated random sampling of the original dataset, and multiple decision trees were trained on each subset. Genes with importance scores exceeding 2 were ultimately selected and designated as RF1 (PC) and RF2 (CHD), respectively.

Finally, the “VennDiagram” package (version 1.7.1) was used to find the intersection of Boruta1, Boruta2, RF1 and RF2, so as to identify the feature genes.

Identification of key genes

For analyzing the expression levels of feature genes in training set 1, validation set 1, training set 2, and validation set 2, the “pROC” package (version 1.18.0) (28) was utilized to conduct receiver operating characteristic (ROC) analysis on these genes and compute the area under the curve (AUC). Genes that satisfied the condition of having an AUC >0.7 and AUC ≠1 across all four datasets were designated as candidate key genes. Subsequently, The Wilcoxon rank-sum test was used to compare the expression levels of candidate key genes across all samples in training set 1, validation set 1, training set 2, and validation set 2. Candidate key genes showing consistent expression patterns and statistically significant differences (P<0.05) across all four datasets were selected and identified as key genes.

Establishment of the nomogram of key genes

To analyze the predictive ability of key genes for the occurrence of PC and CHD, the following unified analysis workflow was adopted based on training set 1 (GSE70768) and training set 2 (GSE113079): The “rms” package (version 1.7-14) (29) was used to construct nomograms predicting the risk of developing PC and CHD, respectively. The “pROC” package (version 1.18.0) was employed to plot ROC curves and calculate the AUC, with criteria of AUC >0.7 and AUC ≠1 defining effective prediction. The “regplot” package (version 1.1) (30) was utilized to generate calibration curves for evaluating the predictive accuracy and clinical reliability of the key genes (P>0.05). Additionally, decision curve analysis (DCA) was performed using the “ggDCA” package (version 1.1) (31) to assess the clinical utility of the nomogram models.

Chromosomal and subcellular localization analysis

To ascertain the chromosomal locations of key genes, the “Circos” package (version 1.2.2) (32) was utilized to analyze the distribution of these genes across chromosomes.

For a more in-depth understanding of the subcellular distribution of these key genes, the nucleotide/protein sequences in FASTA format were retrieved from the National Center for Biotechnology Information (NCBI) database(https://www.ncbi.nlm.nih.gov/gene/). Subsequently, the subcellular localization of these key genes was predicted via the mRNALocater database (http://bio-bigdata.cn/mRNALocater/). The outcomes were then analyzed and visualized with the “ggplot2” package (version 3.4.1).

Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA)

To characterize the functional annotations and signaling pathways related to the identified key genes in PC and CHD. the following standardized analysis process was implemented: Spearman correlation (cor) analysis was carried out between the key genes and other genes in training set 1 (GSE70768) and training set 2 (GSE113079) by using the “psych” package (version 2.1.6) (https://CRAN.R-project.org/package=psych). The “c2.cp.kegg.v7.0.symbols.gmt” file was retrieved from the molecular signatures database (MSigDB) (https://www.gsea-msigdb.org) as the reference gene set. Genes were ranked in decreasing order of the cor. GSEA was then performed on these ranked genes using the “clusterProfiler” package (version 4.2.2). Key-gene-enriched pathways were filtered based on the following criteria: false discovery rate (FDR) <0.25, P value <0.05, and absolute normalized enrichment score (|NES|) >1. The top 5 enriched pathways for key genes in each disease were sorted according to the P value.

To compare the differences in enriched pathways between the PC group and the control group, as well as between the CHD group and the control group, a standardized analytical workflow was implemented as follows: using the “GSVA” package (version 1.42.0) (33), GSVA analysis was performed on the samples of the PC group and control group in training set 1 (GSE70768), and the CHD group and control group in training set 2 (GSE113079), with the “h.all.v2024.1.Hs.symbols.gmt” gene set from MSigDB as the reference to obtain GSVA scores. Subsequently, the GSVA scores between groups were compared using the “limma” package (version 3.54.0) (P<0.05, |t|>2). Finally, bar plots were generated for visualization using the “ggplot2” package (version 3.4.1).

Molecular regulatory network and gene multiple association network integration algorithm (GeneMANIA)

To examine the underlying regulatory pathways of the key genes., the microcosm database within the “multiMiR” package (version 1.24.0) (34) was utilized to predict the microRNAs (miRNAs) associated with key genes. Subsequently, the KnockTF database (https://bio.liclab.net/KnockTF/index.php) was employed for predicting the transcription factors (TFs) of key genes. Finally, the “ggplot2” (version 3.4.1) and “ggalluvial” packages (version 0.12.5) (35) were used to build a TFs-key gene-miRNA network. To explore the interaction relationships between key genes and related genes, the GeneMANIA tool (http://GeneMANIA.org/) was employed to predict functionally associated genes and the biological processes they are involved in.

Analysis of immune infiltration

To assess immune cell infiltration in the PC and control groups of training set 1 (GSE70768), as well as the CHD and control groups of training set 2 (GSE113079), the following standardized analysis procedure was employed: The CIBERSORT algorithm, which estimates relative subsets of RNA transcripts to identify cell types, was utilized to measure the infiltration abundances of 22 immune cell types in samples from both training sets (samples with P values >0.05 were removed). The proportions of these 22 immune cells in PC vs control and CHD vs control samples were visualized using the “ggplot2” package (version 3.4.1). Next, a Wilcoxon test (P<0.05) was carried out to filter immune cells with significant intergroup differences, defined as differentially immune cells, and the results were presented via “ggplot2” (version 3.4.1). Moreover, to explore the cor between key genes and differentially immune cells, and among differentially immune cells, a Spearman cor analysis was conducted on all samples from training set 1 and training set 2. The “cor” function in the R package (version 4.2.2) (36) was used to calculate cor coefficients between key genes and differentially immune cells, and among differentially immune cells (|cor| >0.3, P<0.05). Heatmaps were plotted using the “corrplot” package (version 0.92) (37) to visualize the results.

Drug prediction and molecular docking

To explore potential drugs associated with key genes, The Drug-Gene Interaction Database (DGIdb) (https://www.dgidb.org/) was applied to identify potential drug candidates targeting key genes, and Cytoscape software (version 3.8.2) was used to construct and visualize the gene-drug interaction network.

To investigate the possible therapeutic mechanisms of these drugs, a molecular docking analysis of drug-key gene interactions was further conducted: first, three-dimensional (3D) structure files of drugs were acquired from PubChem (https://pubchem.ncbi.nlm.nih.gov/), and PDB structure files of key genes were retrieved from the RCSB protein data bank (PDB) (https://www.rcsb.org/pdb). Next, pymol molecular graphics system (PyMOL) (version 4.6.0) (https://www.pymol.org/) was used to remove solvents and water molecules from the structures, and AutoDockTools (version 1.2.0) (https://autodock.scripps.edu/) was utilized to add hydrogen atoms, calculate charges, and convert proteins into PDBQT format. The 3D structure files of key genes obtained from PubChem were converted into PDB files using Open Babel GUI34 (http://openbabel.org/). The resulting PDB files of key genes were then imported into AutoDockTools (version 1.2.0) to add hydrogen atoms, calculate charges and affinities, designate them as docking ligands, set their torsion trees, and export them in PDBQT format. By analyzing the binding sites of key gene proteins, the corresponding docking active pockets were identified. AutoGrid (https://autodock.scripps.edu/) was used to set up the Grid Box and execute the docking process, with genetic algorithm parameters configured (GA run times =50) and default docking parameters applied for validation. Finally, PyMOL (version 4.6.0) was used to visualize the results. Docking results were sorted by score, and a docking score ≤−5 kcal/mol was considered to indicate good binding affinity between the drug and the target.

RT-qPCR

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the Second Hospital of Tianjin Medical University {approval number: Science Review [2025] No. (026)}. The patients provided their written informed consent to participate in this study. The experiment firstly conducted total RNA extraction. Five pairs of frozen human whole blood samples (1–5 for Control group and 6–10 for PC + CHD group) were used. For each sample, 600 µL of whole blood was mixed with 700 µL of Novozam TRizol reagent, homogenized thoroughly, and placed on ice for 10 min to lyse cells. Then 200 µL of chloroform from Chengdu Guerda Adhesive Co., Ltd. was added, shaken vigorously for 30 sec, and stood at room temperature for 10 min. After centrifugation at 12,000 g and 4 °C for 15 min, the upper aqueous phase was transferred to a new EP tube, mixed with equal-volume ice-cold isopropanol, and stood for 10 min (or overnight at −20 °C for small samples). Following centrifugation at 12,000 g for 10 min, the RNA precipitate was washed twice with 1 mL of 75% ethanol, dried, dissolved in 20–50 µL RNase-free water, and quantified by NanoPhotometer N50. Subsequently, reverse transcription was performed using Kunming Yungeng’s HP All-in-one qRT Master Mix II. The reaction system (5 µL 4x Master Mix, 2 µg RNA, RNase-free water to 20 µL) was incubated at 50 °C for 10 min, 85 °C for 5 sec, and held at 4 °C. The cDNA was diluted 5–20 times, and qPCR was conducted with 3 µL cDNA, 5 µL Servicebio SYBR Green Master Mix, 1 µL of each 10 µM primer (synthesized by Sangon Biotech) on a CFX96 instrument under 40 cycles: 95 °C for 1 min (pre-denaturation), then 95 °C/20 sec, 55 °C/20 sec, 72 °C/30 sec. Results were analyzed by melting curves for primer specificity and the 2^–∆∆Ct method for relative expression, with GraphPad Prism 10 used for plotting and statistical analysis (Table S1).

Statistical analysis

All analyses were performed using the R package (version 4.2.2) (https://www.r-project.org/). The Wilcoxon rank-sum test was used to compare differences between the two groups, whereas the t-test was applied for analyzing the RT-qPCR data.

Results

Integrated analysis uncovers 84 candidate genes

In training set 1 (GSE70768) and training set 2 (GSE113079), 939 DEGs1 and 4,884 DEGs2 were respectively detected. In training set 1, relative to the control group, 387 DEGs1 were activated and 552 DEGs1 were repressed in the PC group. In training set 2, compared with the control group, 2,773 DEGs2 were activated and 2,111 DEGs2 were repressed in the CHD group. Volcano plots displayed the distribution of DEGs1 and DEGs2 between samples, with red dots representing upregulated genes, such as TARP and AGR2 in training set 1, and OPN4 and BIRC7 in training set 2; green dots representing downregulated genes, such as ABP1 and MME in training set 1, and IL1A and PAK2 in training set 2; and gray dots indicating non-DEGs. Most DEGs1 and DEGs2 were concentrated on both sides of the volcano plots, indicating significant expression differences these genes (Figure 1A,1B). Heatmaps were employed to display the expression profiles of DEGs1 and DEGs2 across various samples. The density plot at the top illustrated the frequency distribution of the expression values of DEGs1 and DEGs2, while the clustering dendrogram on the left classified DEGs1 and DEGs2. Volcano plots were generated to visualize the distribution of genes according to their expression characteristics (Figure 1C,1D). Subsequently, the intersection of the 387 upregulated DEGs from PC and 2,773 upregulated DEGs from CHD was determined. Similarly, the intersection of 552 downregulated DEGs from PC and 2,111 downregulated DEGs from CHD was identified. The union of these two intersecting gene sets resulted in a total of 84 candidate genes (Figure 1E,1F; available online: https://cdn.amegroups.cn/static/public/tau-2025-519-1.xlsx).

Figure 1 Identification of DEGs in PC. (A) Volcano plot of DEGs between PC samples and normal controls. (B) Volcano plot of DEGs between CHD samples and normal controls. Red: upregulated DEGs; blue: downregulated DEGs; gray: genes without statistical significance. (C) Heatmap of DEGs between PC samples and normal controls. (D) Heatmap of DEGs between CHD samples and normal controls. Orange-red: case samples; blue: control samples; upper density heatmap color represents gene expression density across samples—darker red indicates higher density; lower heatmap rows represent genes—red: highly expressed genes, blue: lowly expressed genes. (E) Venn diagram of co-upregulated DEGs between PC and CHD. (F) Venn diagram of co-downregulated DEGs between PC and CHD. CHD, coronary heart disease; DEGs, differentially expressed genes; FC, fold change; PC, prostate cancer.

Deciphering candidate genes: GO, KEGG and PPI analyses

The candidate genes were subjected to GO enrichment analysis, yielding a total of 491 results, including 413 BPs, 25 CCs, and 53 MFs. The top 5 significantly enriched terms in each category were presented as follows: for BPs, they were regulation of blood circulation, muscle contraction, regulation of heart contraction, muscle system process, and heart contraction; for CCs, basolateral plasma membrane, basal plasma membrane, basal part of cell, sarcolemma, and myofibril; for MFs, phosphoric ester hydrolase activity, actin binding, phosphoric diester hydrolase activity, structural constituent of cytoskeleton, and protein phosphatase binding (Figure 2A). These results provided important insights into the roles of these genes in various biological processes, emphasizing that the candidate genes were significantly enriched in BPs, CCs, and MFs. At the same time, KEGG functional enrichment analysis was performed on the candidate genes, resulting in the identification of 41 associated pathways. The top 10 pathways with the most pronounced enrichment outcomes included cellular senescence, the cGMP-PKG signaling pathway, salivary secretion, the glucagon signaling pathway, the cAMP signaling pathway, adrenergic signaling in cardiomyocytes, gastric acid secretion, aldosterone synthesis and secretion, pc, and the regulation of TRP channels by inflammatory mediators (Figure 2B). These findings demonstrated that the candidate genes were strongly linked to cellular senescence, cardiovascular signal transduction, endocrine regulation, and inflammatory processes, laying the groundwork for uncovering their possible roles in disease development and as therapeutic targets.

Figure 2 Preliminary analysis of candidate genes. (A) GO enrichment results of DEGs. (B) KEGG pathway analysis of DEGs. (C,D) PPI networks of DEGs. DEGs, differentially expressed genes; GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; PPI, protein-protein interaction.

Upon analyzing the PPI network built from candidate genes, 40 candidate genes were identified as interconnected with a confidence score >0.4. The MCC algorithm was applied to pick out the top 20 candidate genes for network construction. Among these, interactions were detected between FOXO1, GPX3, and MSN proteins (Figure 2C,2D, available online: https://cdn.amegroups.cn/static/public/tau-2025-519-2.xlsx).

Machine learning-based screening of top 20 candidate genes from MCC analysis

In the training sets 1 (GSE70768) and 2 (GSE113079), the machine learning methods Boruta and RF were used to further screen the top 20 candidate genes from the MCC analysis. The Boruta algorithm, which evaluates the relevance of actual features by contrasting them with randomized shadow features, identified 20 genes (Boruta1) in training set 1, including ATP2B4, MB, FOXO1, PIK3R1, COMP, FMOD, KLK3, INSIG1, ATP1B1, SPDEF, SYNM, GPX3, MYL2, STOM, TGFB3, ADD3, HMGCS1, LPIN1, MSN, and CALM1, all of which showed significantly higher importance scores than the shadow features (Figure 3A). In training set 2, 18 genes (Boruta2) were identified, which were identical to Boruta1 except for the absence of COMP and FMOD (Figure 3B). The RF analysis, based on feature importance scores with a threshold >2, selected 15 genes (RF1) in training set 1: CALM1, SYNM, SPDEF, FOXO1, GPX3, PIK3R1, TGFB3, ATP2B4, ADD3, LPIN1, INSIG1, MB, FMOD, COMP, and MYL2 (Figure 3C,3D). In training set 2, 11 genes (RF2) were selected: MB, PIK3R1, LPIN1, MYL2, CALM1, ADD3, TGFB3, SPDEF, ATP2B4, GPX3, and HMGCS1 (Figure 3E,3F). The combined use of these two methods evaluated feature importance from different perspectives, providing multi-model support for the subsequent screening of feature genes.

Figure 3 Identification of biomarkers. (A) Boruta analysis of training set 1. (B) Boruta analysis of training set 2. (C,D) Diagnostic biomarkers for PC screened by RF algorithm. (E,F) Diagnostic biomarkers for CHD screened by RF algorithm. (G) Venn diagram of machine learning-intersected genes. CHD, coronary heart disease; PC, prostate cancer; RF, random forest.

Finally, the intersection of Boruta1, Boruta2, RF1, and RF2 was calculated, yielding 10 intersecting genes designated as feature genes, including: CALM1, SPDEF, GPX3, PIK3R1, TGFB3, ATP2B4, ADD3, LPIN1, MB, and MYL2 (Figure 3G).

Multi-dataset validation screening identifies key genes ATP2B4 and ADD3

The expression levels of 10 key genes were compared across various datasets: training set 1 (GSE70768), validation set 1 (GSE88808), training set 2 (GSE113079), and validation set 2 (GSE66360). For the ATP2B4 gene, the AUC values were 0.877 in training set 1, 0.893 in validation set 1, 0.880 in training set 2, and 0.763 in validation set 2. For the ADD3 gene, the AUC values were 0.822 in training set 1, 0.857 in validation set 1, 0.908 in training set 2, and 0.728 in validation set 2 (Figure 4A-4D). Additionally, the expressions of these two genes were significant (P<0.001) and exhibited consistent trends across all four datasets. Consequently, these two genes were recognized as key genes (Figure 4E-4H).

Figure 4 Validation of key genes across multiple datasets. (A) ROC analysis of biomarker diagnostic performance in training set 1. (B) ROC analysis in validation set 1. (C) ROC analysis in training set 2. (D) ROC analysis in validation set 2. (E) Expression of candidate biomarkers in training set 1. (F) Expression in validation set 1. (G) Expression in training set 2. (H) Expression in validation set 2. ***, P<0.001. AUC, area under the curve; CHD, coronary heart disease; PC, prostate cancer; ROC, receiver operating characteristic.

Nomogram models for PC and CHD risk assessment

A nomogram was employed to assess the risk of developing PC. The total score was 77.7, and the corresponding PC risk probability was 0.318. This meant that, according to the model, the individual had a relatively high likelihood of developing PC (Figure 5A). The AUC was 0.925, indicating that the model had a high level of discrimination (Figure 5B). A calibration curve was used to examine the cor between the model’s predicted probabilities and the actual probabilities. This demonstrated a strong agreement between the model’s predicted probabilities and the observed outcomes (Figure 5C). The decision curve exhibited favorable net benefits at different risk thresholds, validating the model’s robust clinical predictive effectiveness and practical applicability (Figure 5D).

Figure 5 Nomograms for assessing risk of coronary PC and CHD. (A) Nomogram model for PC biomarkers. (B) ROC curve of PC nomogram. (C) Calibration curve of PC nomogram. (D) DCA curve of PC nomogram. (E) Nomogram model for CHD biomarkers. (F) ROC curve of CHD nomogram. (G) Calibration curve of CHD nomogram. (H) DCA curve of CHD nomogram. ***, P<0.001. AUC, area under the curve; CHD, coronary heart disease; DCA, decision curve analysis; PC, prostate cancer; ROC, receiver operating characteristic.

Similarly, a nomogram was used to assess the risk of developing CHD. The total score was 60.8, and the corresponding CHD risk probability was 0.0523. This indicated that, according to the model, the individual had a relatively high likelihood of developing CHD (Figure 5E). The AUC was 0.912, indicating that the model also had a high level of discrimination (Figure 5F). A calibration curve was used to examine the cor between the model’s predicted probabilities and the actual probabilities. The Hosmer-Lemeshow test produced a P value of 0.468, and the calibration curve had a slope close to 1, indicating a strong agreement between the model’s predicted probabilities and the observed outcomes (Figure 5G). The good net benefits of the decision curve also confirmed its clinical practicality (Figure 5H). In summary, for both PC and CHD, the nomogram models used showed good discrimination, calibration, and clinical predictive efficacy, highlighting their potential value in practical applications.

ADD3 and ATP2B4: chromosomal and subcellular localization insights

Chromosomal localization analysis had determined the positions of 2 key genes, ADD3 and ATP2B4, on the chromosomes, with each located on chromosomes 10 and 1, respectively. The chromosomal positioning of these genes was crucial for understanding their genetic characteristics and roles in PC and CHD (Figure 6A).

Figure 6 Chromosomal and subcellular localization of ADD3 and ATP2B4. (A) Chromosomal locations of ADD3 and ATP2B4. (B) Subcellular localization of ADD3. (C) Subcellular localization of ATP2B4.

The subcellular localization results of ADD3 and ATP2B4 showed that ADD3 had the highest score in the nucleus, followed by the cytoplasm, extracellular region, and endoplasmic reticulum, with the least in the mitochondria; ATP2B4 also had the highest score in the nucleus, followed by the cytoplasm, with significantly lower scores in the endoplasmic reticulum, extracellular region, and mitochondria. Specifically, ADD3 was predominantly located in the nucleus (the highest proportion), while ATP2B4 was mainly distributed in both the nucleus and cytoplasm (both had significantly higher scores than other subcellular structures) (Figure 6B,6C). These findings offered a visual foundation for the subcellular localization of crucial genes. This was of great significance for further delving into the biological processes these genes participated in and their possible action mechanisms.

Multi-dataset analysis of disease-related gene pathway enrichment features

For the multi-dataset analysis of disease-related gene pathway enrichment features, in the training set 1 (GSE70768), GSEA analysis uncovered that the key genes ADD3 and ATP2B4 were respectively enriched in 56 and 100 pathways. The top 5 significantly enriched pathways for ADD3 included ribosome, focal adhesion, vascular smooth muscle contraction, pathways in cancer, and N-glycan biosynthesis. The top 5 significantly enriched pathways for ATP2B4 included ribosome, focal adhesion, extracellular matrix-receptor interaction, pathways in cancer, and cytokine-cytokine receptor interaction (Figure 7A,7B).

Figure 7 GSEA of ADD3 and ATP2B4. (A) The top 5 pathways enriched before GSEA for ADD3 in training set 1. (B) The top 5 pathways enriched before GSEA for ATP2B4 in training set 1. (C) GSVA analysis of training set 1. (D) The top 5 pathways enriched before GSEA for ADD3 in training set 2. (E) The top 5 pathways enriched before GSEA for ATP2B4 in training set 2. (F) GSVA analysis of training set 2. GSEA, gene set enrichment analysis; GSVA, gene set variation analysis.

GSVA results showed that there were 26 significantly enriched biological processes between PC group and control group. Among them, 8 pathways such as myc target gene set v2, MYC target gene set v1, and E2F target gene set were significantly upregulated, suggesting potential associations with enhanced regulation of cell proliferation, DNA repair, and metabolic activity. Including late estrogen response, heme metabolism, and TNF-α signaling via NF-κB, 18 pathways were significantly downregulated, involving functional abnormalities in hormone response, apoptosis, signal transduction, and tissue microenvironment regulation. These results provided insights into the biological characteristics of PC samples and potential key genes (Figure 7C).

In the training set 2 (GSE113079), GSEA showed that the key genes ADD3 and ATP2B4 were enriched in 69 and 77 pathways respectively. The top 5 significantly enriched pathways for ADD3 included RNA degradation, ubiquitin mediated proteolysis, neuroactive ligand receptor interaction, olfactory transduction, and spliceosome. The top 5 significantly enriched pathways for ATP2B4 included natural killer cell mediated cytotoxicity, neuroactive ligand receptor interaction, ubiquitin mediated proteolysis, olfactory transduction, and spliceosome (Figure 7D,7E).

The GSVA results showed that there were 41 significantly enriched biological processes between the CHD group and the control group. Among them, 33 pathways such as protein secretion and oxidative phosphorylation were significantly upregulated, indicating that the biological processes related to these pathways were active in CHD and might be closely associated with physiological activities such as cellular secretory function and oxidative phosphorylation. Including hedgehog signaling and myogenesis, 8 pathways were significantly downregulated, indicating that the biological processes associated with these pathways were suppressed in CHD. The functions involving signal transduction, muscle generation, etc., were relatively weak in the samples. Collectively, these findings unveiled the active and repressed biological pathways in CHD, offering insights into exploring the biological characteristics and underlying mechanisms of CHD (Figure 7F; available online: https://cdn.amegroups.cn/static/public/tau-2025-519-3.xlsx).

Molecular regulation and genetic association networks of ADD3 and ATP2B4

The results of the molecular regulatory network showed that 21 miRNAs were predicted for ADD3 and 11 miRNAs for ATP2B4. Then, 10 TFs were predicted. Among them, ATP2B4 corresponded to 6 TFs and ADD3 to 4 TFs, with 2 TFs shared by both. The constructed TFs-key gene-miRNA network showed that ADD3 and ATP2B4 were connected to each other through their associated TFs and miRNAs, forming a complex multi-level regulatory network structure. This provided important clues for in-depth analysis of the regulatory mechanisms of key gene expression and exploration of potential molecular targets related to diseases (Figure 8A; available online: https://cdn.amegroups.cn/static/public/tau-2025-519-4.xlsx).

Figure 8 Molecular regulatory networks of ADD3 and ATP2B4. (A) TFs-mRNA-miRNA regulatory network. (B) GeneMANIA network. GeneMINIA, gene multiple association network integration algorithm; TFs, transcription factors.

The GeneMANIA analysis results showed the association network of ADD3 and ATP2B4 with other genes, where key genes were located in the center and presented in the form of pie charts, with their colors representing the relationships with co-expressed genes. The color of the lines connecting the genes indicated the biological processes they were involved in together. The network involved biological functions such as smooth muscle contraction, regulation of muscle system processes, and regulation of actin filament capping. This diagram intuitively presented how key genes were interconnected with other genes through different types of relationships and played roles together in various biological processes (Figure 8B).

Immune infiltration analysis in PC and CHD

The immune infiltration analysis revealed significant differences in the composition of immune cell populations between the PC group and the control group. Specifically, the PC group had greater percentages of activated mast cells, monocytes, resting mast cells, and resting memory CD4 T cells. This demonstrated that the immune cell infiltration composition in the PC group diverged from that of the control group, suggesting possible changes in the PC group’s immune microenvironment (Figure 9A). Five types of immune cells with differential abundance were detected between the PC and control groups: memory B cells (P<0.05), plasma cells (P<0.01), M0 macrophages (P<0.0001), M1 macrophages (P<0.0001), and resting mast cells (P<0.05). These differentially infiltrated immune cells provided important clues for further investigations into the PC-related immune microenvironment and potential pathogenic mechanisms (Figure 9B). Cor analysis among the differentially abundant immune cells revealed a significant positive cor between m0 macrophages and m1 macrophages (cor =0.320, P<0.0001), indicating that these two cell types tended to increase or decrease concurrently in the samples. Conversely, a significant negative cor was observed between m0 macrophages and resting mast cells (cor =−0.313, P<0.0001), suggesting opposing trends in their changes (Figure 9C). The cor analysis between the key genes ADD3/ATP2B4 and the differentially abundant immune cells demonstrated that ATP2B4 was significantly negatively correlated with m1 macrophages (cor =−0.38, P<0.0001). This finding revealed an inverse relationship between ATP2B4 gene expression and the quantity or activity of m1 macrophages, offering critical insights into understanding the regulatory mechanisms of the immune microenvironment in related diseases and identifying potential therapeutic targets. However, no significant cor were detected between the ADD3 gene and the differentially abundant immune cells, which might aid in deciphering the functional divergence of key genes across different immune cell types (Figure 9D).

Figure 9 Immune infiltration analysis in PC and CHD groups. (A) Stacked plot of 22 immune cell types in PC. (B) Differential immune cell analysis between PC and controls. (C) Correlation analysis of differential immune cells in PC. (D) Heatmap of correlations between biomarkers and differential immune cells in PC. (E) Stacked plot of 22 immune cell types in CHD. (F) Differential immune cell analysis between CHD and controls. (G) Correlation analysis of differential immune cells in CHD. (H) Heatmap of correlations between biomarkers and differential immune cells in CHD. ns, P>0.05; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. CHD, coronary heart disease; PC, prostate cancer.

The results of the immune infiltration analysis indicated that there were significant differences in the composition of immune cell populations between the CHD group and the control group. Specifically, the CHD group had higher proportions of monocytes, naive B cells, and activated memory CD4 T cells, suggesting that the CHD group had characteristic differences in immune cell infiltration composition compared to the control group. Such differences might be closely related to the development and immunopathological processes of CHD (Figure 9E). Between the CHD and control group, 10 differentially abundant immune cell types were identified, namely CD8 T cells (P<0.001), naive CD4 T cells (P<0.0001), activated memory CD4 T cells (P<0.0001), regulatory (Tregs) T cells (P<0.0001), gamma delta T cells (P<0.05), resting natural killer (NK) cells (P<0.001), activated NK cells (P<0.01), monocytes (P<0.0001), resting mast cells (P<0.05), and activated mast cells (P<0.05). These differentially abundant immune cells provided important clues for further exploring the CHD-related immune microenvironment and potential pathogenic mechanisms (Figure 9F). The cor analysis among the differentially abundant immune cells showed that resting NK cells and activated memory CD4 T cells had a significant positive cor (cor =0.451, P<0.0001), while CD8 T cells and monocytes had a significant negative cor (cor =−0.744, P<0.0001). These significant cor suggested that these cells interacted with each other or jointly participated in biological processes under specific conditions (Figure 9G). Analysis between the key genes ADD3 and ATP2B4 and the differentially abundant immune cells demonstrated that ATP2B4 had a significant positive cor with CD8 T cells (cor =0.40, P<0.0001) and a significant negative cor with naive CD4 T cells (cor =−0.46, P<0.0001). Meanwhile, the ADD3 gene had a significant positive cor with resting NK cells (cor =0.41, P<0.0001) and a significant negative cor with Monocytes (cor =−0.34, P<0.0001) (Figure 9H). These results offered valuable insights into the regulatory roles of the key genes ADD3 and ATP2B4 within immune cells, contributing to a more comprehensive understanding of the intricate immune microenvironment involved in the related diseases (available online: https://cdn.amegroups.cn/static/public/tau-2025-519-5.xlsx).

Molecular docking and drug prediction for PC/CHD genes

In this study, 22 potential drugs that might effectively treat PC and CHD were identified. Among them, 20 drugs were predicted by ADD3, mainly associated with 3-hydroxy capric acid, DL-175, ZQ-16, etc., with the highest interaction score reaching 2.61. meanwhile, ATP2B4 predicted 2 drugs, mainly related to Zn²⁺ and pregnenolone sulphate, with the highest interaction score reaching 5.22 (Figure 10A; Table 1). These findings provided important clues regarding gene-drug associations for the drug development and precision treatment of PC and CHD, facilitating further exploration of personalized treatment strategies based on gene targets.

Figure 10 Molecular docking and drug prediction for PC/CHD genes. (A) Biomarker-drug interaction network. (B) Molecular docking results of ADD3 with DL-175. (C) Molecular docking results of ATP2B4 with pregnenolone sulfate. CHD, coronary heart disease; PC, prostate cancer.

In this study, molecular docking was performed on the protein targets of two key genes and the drugs with the highest interaction scores. All docking interactions were successful, and the docking scores were all ≤−5 kcal/moL (Table 2). Among them, the docking score between ADD3 and DL-175 was −6.7 kcal/mol, while that between ATP2B4 and pregnenolone sulphate was −9.0 kcal/mol. This provided important structural basis and theoretical support for revealing the action mechanisms of potential therapeutic drugs for PC and CHD and promoting precision drug development based on gene targets. In addition, the docking results of ADD3 and DL-175 showed the spatial structure of their binding and marked the relevant sites of f106 and r431; while the docking results of gngt2 and pregnenolone sulphate marked the relevant sites of g806 and g217 (Figure 10B,10C).

RT-qPCR reveals expression changes of ADD3 and ATP2B4 genes in PC and CHD

The melting curve analysis showed that the primers for ATP2B4, ADD3 and the internal reference GAPDH had good specificity, with smooth single peaks in the melting curves, indicating they were suitable for quantitative analysis. Gene expression detection revealed that the relative expression levels of both ATP2B4 and ADD3 genes in the PC + CHD group were lower than those in the Control group, and the differences were statistically significant (P<0.05). Specifically, the expression level of ATP2B4 in the Control group was 1±0.2361, while that in the PC + CHD group was 0.6649±0.1505 (P=0.03). The expression level of ADD3 in the Control group was 1±0.3271, and that in the PC + CHD group was 0.4724±0.1904 (P=0.01). These results indicated that the expression of ATP2B4 and ADD3 genes was significantly inhibited in the PC + CHD group (Table 3; Figure 11A,11B).

Figure 11 Expression changes of ADD3 and ATP2B4 in PC and CHD patients. (A) Expression changes of ADD3. (B) Expression changes of ATP2B4. *, P<0.05. CHD, coronary heart disease; PC, prostate cancer.

Discussion

CHD and cancer are among the major global health burdens and causes of death. Cardio-oncology, as an emerging interdisciplinary field, focuses on the adverse effects of cancer treatments on the cardiovascular system and the interactions between the two. CHD and cancer share common risk factors and pathogenesis (38) Studies show that CHD is associated with an increased risk of cancer, and some research also suggests that CHD occurring after a cancer diagnosis may lead to cancer progression (39). This bidirectional relationship is particularly complex in studies on the comorbidity of PC and CHD (40). One of the treatment methods for PC, androgen deprivation, may increase the risk of cardiovascular disease (15,41,42). The use of related medications is associated with an increased risk of cardiovascular events such as CHD, potentially due to mechanisms such as testosterone surges (43). At the same time, CHD is also a risk factor for PC, and the two may interact through common mechanisms such as metabolic disorders and inflammatory responses. The underlying mechanisms of their interaction are not yet fully understood. This article innovatively identifies comorbid genes for PC and CHD, and constructs a cross-disease risk prediction model.

Key genes and molecular mechanisms in PC and CHD

This research identified ADD3 and ATP2B4 as critical genes associated with PC and CHD, based on bioinformatics analysis and experimental confirmation. The protein encoded by ADD3 belongs to the adducin family, which is involved in cytoskeletal reorganization and cell migration (44,45). Studies have identified the K572Q mutation in ADD3 as a contributing factor to impaired myogenic response and disrupted autoregulation of renal and cerebral blood flow, which in turn heightens the risk of hypertension-related kidney disease, cerebrovascular disorders, and cognitive impairment (46). Additionally, a study (47) identified sequence variations in ADD3 in rats that may play a causal role in impaired myogenic response and autoregulation in renal and cerebral circulation.

ATP2B4 belongs to the P-type primary ion-transporting ATPase family and is responsible for removing divalent calcium ions from eukaryotic cells, playing a critical role in intracellular calcium homeostasis (48,49). Meanwhile, studies suggest that targeted modulation of ATP2B4 function may open new therapeutic avenues to promote or inhibit neovascularization in angiogenic diseases. Genome-wide association studies have identified ATP2B4 as a key gene in severe malaria resistance (50). Additionally, research has confirmed calcium overload in red blood cells from patients with multiple myeloma accompanied by osteolytic lesions, which is associated with the downregulation of ATP2B4 by multiple myeloma exosomal miR-4261 (51). These findings enhance our understanding of the molecular mechanisms underlying PC and CHD, providing new directions for future research and clinical therapies.

Cross-disease regulatory role of key genes

Through bioinformatics analysis, ADD3 and ATP2B4 were identified as common key genes shared by PC and CHD, offering a groundbreaking perspective for comorbidity research of the two diseases. The ADD3 gene is enriched in pathways related to vascular smooth muscle contraction and cancer, suggesting its potential role in promoting disease progression in PC by regulating angiogenesis and the tumor microenvironment. On one hand, ADD3 activates the Ras/MAPK signaling pathway, promoting the phenotypic switch of vascular smooth muscle cells from a contractile to a synthetic state. This transition is accompanied by increased secretion of matrix metalloproteinases, which degrade the extracellular matrix to facilitate tumor cell invasion (52,53). On the other hand, in CHD, the vascular smooth muscle contraction pathway enriched by ADD3 is positively associated with coronary artery spasm and plaque rupture risk. ADD3 enhances the interaction between calmodulin and myosin light chain kinase (54). Amplifying vascular reactivity to vasoconstrictive agents such as endothelin-1, leading to abnormal vascular tone. These functions suggest that ADD3 may act as a common driver of vascular pathology in both PC and CHD. In PC, ATP2B4 is enriched in the cytokine-cytokine receptor interaction pathway. It inhibits the activity of calcineurin, thereby reducing IL-2 secretion and suppressing effector cell proliferation. Simultaneously, by expelling intracellular Ca²⁺, ATP2B4 disrupts calcium signaling essential for dendritic cell maturation, leading to impaired tumor antigen presentation (55,56). This dual mechanism weakens immune responses and contributes to the formation of an immunosuppressive microenvironment, which is closely associated with PC’s resistance to immunotherapy. In CHD, ATP2B4 is enriched in the natural killer cell-mediated cytotoxicity pathway (56,57). Studies suggest that ATP2B4 may enhance the clearance of abnormal cells by maintaining intracellular calcium homeostasis, while also modulating adaptive immune responses by preventing the overactivation of naive T cells. Thus, ATP2B4 exhibits a dual function in modulating inflammatory responses in CHD.

Discussion on key genes and immune cells

In PC, infiltration of M0/M1 macrophages is increased, and ATP2B4 expression shows a significant negative correlation with the number of M1 macrophages. This suggests that ATP2B4 may alleviate inflammatory burden in the tumor microenvironment by inhibiting the polarization of macrophages toward the pro-inflammatory M1 phenotype. Macrophage polarization is a critical process in tumor progression; M1 macrophages promote angiogenesis and tumor cell invasion by secreting IL-6, TNF-α, and matrix metalloproteinases (58). In this study, the negative regulatory role of ATP2B4 may be mediated through inhibition of the NF-κB signaling pathway, which is a central regulator of macrophage polarization and whose excessive activation is closely linked to immune evasion in PC (59,60). Additionally, the positive correlation between mast cells and monocytes suggests a synergistic effect through the release of pro-inflammatory mediators such as histamine, further intensifying the inflammatory response in the tumor microenvironment. This finding is consistent with previous reports showing that mast cells promote monocyte recruitment via the TGF-β pathway (61,62).

In CHD, the negative correlation between CD8⁺ T cells and monocytes suggests that cytotoxic T cells may suppress the inflammatory cascade within atherosclerotic plaques by eliminating pro-inflammatory monocytes. This enhanced immune surveillance is associated with the ability of CD8⁺ T cells to secrete IFN-γ and induce apoptosis in monocytes (63). The positive correlation between ADD3 and resting NK cells implies a protective role through the enhancement of innate immune surveillance—NK cells can recognize damage-associated molecular patterns and eliminate apoptotic endothelial cells, thereby stabilizing plaques and preventing rupture (64). Notably, the negative correlation between ATP2B4 and naive CD4⁺ T cells may indicate its role in suppressing their differentiation into pro-inflammatory Th1/Th17 subsets, thus mitigating adaptive immunity-mediated vascular endothelial injury (65). This contrasts with the pathological mechanism in CHD, where overactivated T cells exacerbate arterial wall inflammation. These dynamic interactions among immune cells highlight how key genes contribute to the pathological processes of both PC and CHD through multidimensional regulation of the immune microenvironment, offering new directions for immune cell-targeted therapies.

In this experiment, the study observed that the expression levels of ADD3 and ATP2B4 were significantly downregulated in patients with comorbid PC and CHD compared to the control group (P<0.05). This result is highly consistent with the gene regulatory trends predicted by previous bioinformatics analyses. These findings not only validate the reliability of the bioinformatics approach but also suggest that ADD3 and ATP2B4 may serve as key regulatory nodes in the development and progression of PC and CHD comorbidity. Although the expression levels of ADD3 and ATP2B4 were significantly reduced in patients with PC and CHD comorbidity, this result holds important clinical significance. It further clarifies the critical role of these genes in disease onset and progression and provides a clear direction for in-depth exploration of their molecular mechanisms. At the same time, the discovery of this significant downregulation offers potential targets for developing gene-targeted therapeutic strategies, contributing to the future advancement of precision medicine. By specifically modulating the expression of these genes, it is expected that the prognosis of patients with CHD complicated by PC can be improved, thereby offering better treatment outcomes and quality of life. This study provides new insights into developing gene-targeted therapies and improving prognosis for patients with CHD and PC comorbidity.

Despite the promising results, there are certain limitations in this study, such as a limited sample size, lack of preclinical validation, and reliance on public database data, which to some extent affect the generalizability and reliability of the findings. Future research will focus on functional and clinical validation in larger cohorts and further elucidate the specific functional mechanisms of the identified key genes and their roles within disease networks, with the aim of providing a more solid theoretical basis for the precision treatment of PC and CHD.

Conclusions

This study, based on transcriptomics and experimental validation, explored the key genes and molecular mechanisms of comorbidity between PC and CHD for the first time. By analyzing transcriptomic data from the GEO database, and combing with machine learning algorithms and experimental validation, ADD3 and ATP2B4 were successfully identified as key genes for PC and CHD. These genes show significant expression differences in both PC and CHD. Furthermore, this study constructed a cross-disease nomogram risk prediction model based on these key genes, which demonstrates good predictive performance and suggests a new mechanism of immune-vascular interactions, providing new tools for risk assessment of PC and CHD. Through drug prediction and molecular docking analysis, several potential therapeutic drugs were predicted and their binding stability with the key genes was validated. These findings not only offer new perspectives for understanding the comorbid mechanisms of PC and CHD but also lay the foundation for developing new diagnostic biomarkers and therapeutic targets.

Acknowledgments

We gratefully acknowledge the support of Tianjin Natural Science Foundation (25JCLMJC00310) and Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-029A).

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-519/rc

Peer Review File: Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-519/prf

Funding: This work was supported by grants from the Tianjin Natural Science Foundation (25JCLMJC00310) and Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-029A).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-519/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. The study was approved by the Ethics Committee of the Second Hospital of Tianjin Medical University {approval number: Science Review [2025] No. (026)}. The patients provided their written informed consent to participate in this study.

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. Bergengren O, Pekala KR, Matsoukas K, et al. 2022 Update on Prostate Cancer Epidemiology and Risk Factors-A Systematic Review. Eur Urol 2023;84:191-206. [Crossref] [PubMed]
2. Baca SC, Singler C, Zacharia S, et al. Genetic determinants of chromatin reveal prostate cancer risk mediated by context-dependent gene regulation. Nat Genet 2022;54:1364-75. [Crossref] [PubMed]
3. Haider M, Leow JJ, Nordström T, et al. Emerging tools for the early detection of prostate cancer. BJUI Compass 2025;6:e70081. [Crossref] [PubMed]
4. Wei JT, Barocas D, Carlsson S, et al. Early Detection of Prostate Cancer: AUA/SUO Guideline Part I: Prostate Cancer Screening. J Urol 2023;210:46-53. [Crossref] [PubMed]
5. Taylor JM, Chen VE, Miller RC, et al. The Impact of Prostate Cancer Treatment on Quality of Life: A Narrative Review with a Focus on Randomized Data. Res Rep Urol 2020;12:533-46. [Crossref] [PubMed]
6. Henríquez I, Roach M 3rd, Morgan TM, et al. Current and Emerging Therapies for Metastatic Castration-Resistant Prostate Cancer (mCRPC). Biomedicines 2021;9:1247. [Crossref] [PubMed]
7. Corradini D, Brizi L, Gaudiano C, et al. Challenges in the Use of Artificial Intelligence for Prostate Cancer Diagnosis from Multiparametric Imaging Data. Cancers (Basel) 2021;13:3944. [Crossref] [PubMed]
8. Zhang X, Wang Y, Liu L, et al. Efficacy of Wen-Dan Decoction in the treatment of patients with coronary heart disease: A protocol for systematic review and meta-analysis. Medicine (Baltimore) 2022;101:e28041. [Crossref] [PubMed]
9. Al-Azzawy SM, McEvoy JW, Johansson I, et al. Global detection and management of dysglycaemic patients with coronary artery disease results from the INTERASPIRE survey from 14 countries across six WHO regions. Cardiovasc Diabetol 2025;24:327. [Crossref] [PubMed]
10. Naushad Z, Malik J, Mishra AK, et al. Artificial Intelligence in Cardiovascular Health: Insights into Post-COVID Public Health Challenges. High Blood Press Cardiovasc Prev 2025;32:475-94. [Crossref] [PubMed]
11. Li Y, Cao GY, Jing WZ, et al. Global trends and regional differences in incidence and mortality of cardiovascular disease, 1990-2019: findings from 2019 global burden of disease study. Eur J Prev Cardiol 2023;30:276-86. [Crossref] [PubMed]
12. López-Fernández T, Farmakis D, Ameri P, et al. European Society of Cardiology Core Curriculum for cardio-oncology. Eur J Heart Fail 2024;26:754-71. [Crossref] [PubMed]
13. Moryousef J, Duivenvoorden W, Leong D, et al. Comprehensive review of cardiovascular disease in prostate cancer: epidemiology, risk factors, therapeutics and prevention strategies. Prostate Cancer Prostatic Dis 2025;28:632-8. [Crossref] [PubMed]
14. Miro C, Di Giovanni A, Murolo M, et al. Thyroid hormone and androgen signals mutually interplay and enhance inflammation and tumorigenic activation of tumor microenvironment in prostate cancer. Cancer Lett 2022;532:215581. [Crossref] [PubMed]
15. Fradin J, Kim FJ, Lu-Yao GL, et al. Review of Cardiovascular Risk of Androgen Deprivation Therapy and the Influence of Race in Men with Prostate Cancer. Cancers (Basel) 2023;15:2316. [Crossref] [PubMed]
16. DE Nunzio C. Androgen deprivation therapy and cardiovascular risk in prostate cancer. Minerva Urol Nephrol 2022;74:508-17. [PubMed]
17. Corona G, Filippi S, Bianchi N, et al. Cardiovascular Risks of Androgen Deprivation Therapy for Prostate Cancer. World J Mens Health 2021;39:429-43. [Crossref] [PubMed]
18. Crocetto F, Barone B, Caputo VF, et al. BRCA Germline Mutations in Prostate Cancer: The Future Is Tailored. Diagnostics (Basel) 2021;11:908. [Crossref] [PubMed]
19. Hu J, Zhou J, Liang Y, et al. Recent Advances and Perspectives of Metabolomics-Based Investigations in Coronary Heart Disease. Curr Atheroscler Rep 2025;27:63. [Crossref] [PubMed]
20. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [Crossref] [PubMed]
21. Gustavsson EK, Zhang D, Reynolds RH, et al. ggtranscript: an R package for the visualization and interpretation of transcript isoforms using ggplot2. Bioinformatics 2022;38:3844-6. [Crossref] [PubMed]
22. Forslund SK, Chakaroun R, Zimmermann-Kogadeeva M, et al. Combinatorial, additive and dose-dependent drug-microbiome associations. Nature 2021;600:500-5. [Crossref] [PubMed]
23. Chen H, Boutros PC. VennDiagram: a package for the generation of highly-customizable Venn and Euler diagrams in R. BMC Bioinformatics 2011;12:35. [Crossref] [PubMed]
24. Wu T, Hu E, Xu S, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2021;2:100141. [Crossref] [PubMed]
25. Baralić K, Petkovski T, Piletić N, et al. Exploring Toxicity of Per- and Polyfluoroalkyl Substances (PFAS) Mixture Through ADMET and Toxicogenomic In Silico Analysis: Molecular Insights. Int J Mol Sci 2024;25:12333. [Crossref] [PubMed]
26. Öznacar T, Güler T. Prediction of Early Diagnosis in Ovarian Cancer Patients Using Machine Learning Approaches with Boruta and Advanced Feature Selection. Life (Basel) 2025;15:594. [Crossref] [PubMed]
27. Yang X, Li Y, Liu L, et al. Prediction of respiratory diseases based on random forest model. Front Public Health 2025;13:1537238. [Crossref] [PubMed]
28. Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011;12:77. [Crossref] [PubMed]
29. Núñez E, Steyerberg EW, Núñez J. Regression modeling strategies. Rev Esp Cardiol 2011;64:501-7. [PubMed]
30. Sui Z, Wu X, Du L, et al. Characterization of the Immune Cell Infiltration Landscape in Esophageal Squamous Cell Carcinoma. Front Oncol 2022;12:879326. [Crossref] [PubMed]
31. Chen Y, He J, Pan X. Prediction of risk factors for preoperative deep vein thrombosis in patients with pelvic fracture. Front Surg 2025;12:1585460. [Crossref] [PubMed]
32. Krzywinski M, Schein J, Birol I, et al. Circos: an information aesthetic for comparative genomics. Genome Res 2009;19:1639-45. [Crossref] [PubMed]
33. Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 2013;14:7. [Crossref] [PubMed]
34. Ru Y, Kechris KJ, Tabakoff B, et al. The multiMiR R package and database: integration of microRNA-target interactions along with their disease and drug associations. Nucleic Acids Res 2014;42:e133. [Crossref] [PubMed]
35. Brunson JC. ggalluvial: Layered Grammar for Alluvial Plots. J Open Source Softw 2020;5:2017. [Crossref] [PubMed]
36. Onyeso OK, Odole AC, Scott DR, et al. Association between sociodemographic factors and mobility limitation among older adults: a systematic review and meta-analysis protocol. Syst Rev 2023;12:19. [Crossref] [PubMed]
37. Long SW, Li SH, Li J, et al. Identification of osteoporosis ferroptosis-related markers and potential therapeutic compounds based on bioinformatics methods and molecular docking technology. BMC Med Genomics 2024;17:99. [Crossref] [PubMed]
38. Wilcox NS, Amit U, Reibel JB, et al. Cardiovascular disease and cancer: shared risk factors and mechanisms. Nat Rev Cardiol 2024;21:617-31. [Crossref] [PubMed]
39. Koelwyn GJ, Newman AAC, Afonso MS, et al. Myocardial infarction accelerates breast cancer via innate immune reprogramming. Nat Med 2020;26:1452-8. [Crossref] [PubMed]
40. Kakkat S, Pramanik P, Singh S, et al. Cardiovascular Complications in Patients with Prostate Cancer: Potential Molecular Connections. Int J Mol Sci 2023;24:6984. [Crossref] [PubMed]
41. Halfmann MC, Emrich T. Towards lower radiation and contrast media dose CT angiography of the aorta by artificial intelligence-supported iterative reconstructions. Eur J Radiol 2022;151:110327. [Crossref] [PubMed]
42. El-Taji O, Taktak S, Jones C, et al. Cardiovascular Events and Androgen Receptor Signaling Inhibitors in Advanced Prostate Cancer: A Systematic Review and Meta-Analysis. JAMA Oncol 2024;10:874-84. [Crossref] [PubMed]
43. Dragomir A, Touma N, Hu J, et al. Androgen Deprivation Therapy and Risk of Cardiovascular Disease in Patients With Prostate Cancer Based on Existence of Cardiovascular Risk. J Natl Compr Canc Netw 2023;21:163-71. [Crossref] [PubMed]
44. Suo Z, Ma X, Ding Y, et al. Posttranscriptional inhibition of γ-adducin promotes the proliferation and migration of osteosarcoma cells. Tumori 2022;108:600-8. [Crossref] [PubMed]
45. Laochareonsuk W, Chiengkriwate P, Sangkhathat S. Single nucleotide polymorphisms within Adducin 3 and Adducin 3 antisense RNA1 genes are associated with biliary atresia in Thai infants. Pediatr Surg Int 2018;34:515-20. [Crossref] [PubMed]
46. Gonzalez-Fernandez E, Fan L, Wang S, et al. The adducin saga: pleiotropic genomic targets for precision medicine in human hypertension-vascular, renal, and cognitive diseases. Physiol Genomics 2022;54:58-70. [Crossref] [PubMed]
47. Fan F, Pabbidi MR, Ge Y, et al. Knockdown of Add3 impairs the myogenic response of renal afferent arterioles and middle cerebral arteries. Am J Physiol Renal Physiol 2017;312:F971-81. [Crossref] [PubMed]
48. Delgado-Coello B, Luna-Reyes I, Méndez-Acevedo KM, et al. Analysis of cholesterol-recognition motifs of the plasma membrane Ca(2+)-ATPase. J Bioenerg Biomembr 2024;56:205-19. [Crossref] [PubMed]
49. Yoshie S, Kuriyama M, Maekawa M, et al. ATP2B4 is an essential gene for epidermal growth factor-induced macropinocytosis in A431 cells. Genes Cells 2024;29:512-20. [Crossref] [PubMed]
50. Thiam A, Nisar S, Adjemout M, et al. ATP2B4 regulatory genetic variants are associated with mild malaria. Malar J 2023;22:68. [Crossref] [PubMed]
51. Bian S, Zhang X, Lin L, et al. Exosomal MiR-4261 mediates calcium overload in RBCs by downregulating the expression of ATP2B4 in multiple myeloma. Front Oncol 2022;12:978755. [Crossref] [PubMed]
52. Wu HB, Wang ZW, Shi F, et al. Avβ3 Single-Stranded DNA Aptamer Attenuates Vascular Smooth Muscle Cell Proliferation and Migration via Ras-PI3K/MAPK Pathway. Cardiovasc Ther 2020;2020:6869856. [Crossref] [PubMed]
53. Singh KK, Matkar PN, Pan Y, et al. Endothelial long non-coding RNAs regulated by oxidized LDL. Mol Cell Biochem 2017;431:139-49. [Crossref] [PubMed]
54. Kiang KM, Leung GK. A Review on Adducin from Functional to Pathological Mechanisms: Future Direction in Cancer. Biomed Res Int 2018;2018:3465929. [Crossref] [PubMed]
55. Bedu-Addo G, Meese S, Mockenhaupt FP. An ATP2B4 polymorphism protects against malaria in pregnancy. J Infect Dis 2013;207:1600-3. [Crossref] [PubMed]
56. Hattangady NG, Foster J, Lerario AM, et al. Molecular and Electrophysiological Analyses of ATP2B4 Gene Variants in Bilateral Adrenal Hyperaldosteronism. Horm Cancer 2020;11:52-62. [Crossref] [PubMed]
57. Prager I, Watzl C. Mechanisms of natural killer cell-mediated cellular cytotoxicity. J Leukoc Biol 2019;105:1319-29. [Crossref] [PubMed]
58. Dong M, Zhou C, Ji L, et al. AG1296 enhances plaque stability via inhibiting inflammatory responses and decreasing MMP-2 and MMP-9 expression in ApoE-/- mice. Biochem Biophys Res Commun 2017;489:426-31. [Crossref] [PubMed]
59. Wang L, He C. Nrf2-mediated anti-inflammatory polarization of macrophages as therapeutic targets for osteoarthritis. Front Immunol 2022;13:967193. [Crossref] [PubMed]
60. Das S, Mukherjee T, Mohanty S, et al. Impact of NF-κB Signaling and Sirtuin-1 Protein for Targeted Inflammatory Intervention. Curr Pharm Biotechnol 2025;26:1207-20. [Crossref] [PubMed]
61. Nakano N, Saida K, Hara M, et al. Mucosal Mast Cell-Specific Gene Expression Is Promoted by Interdependent Action of Notch and TGF-β Signaling. J Immunol 2021;207:3098-106. [Crossref] [PubMed]
62. Campillo-Navarro M, Leyva-Paredes K, Donis-Maturano L, et al. Listeria monocytogenes induces mast cell extracellular traps. Immunobiology 2017;222:432-9. [Crossref] [PubMed]
63. Salerno F, Guislain A, Cansever D, et al. TLR-Mediated Innate Production of IFN-γ by CD8+ T Cells Is Independent of Glycolysis. J Immunol 2016;196:3695-705. [Crossref] [PubMed]
64. Zhu Z, Zeng Z, Song B, et al. Identification of diagnostic biomarkers and immune cell profiles associated with COPD integrated bioinformatics and machine learning. J Cell Mol Med 2024;28:e70107. [Crossref] [PubMed]
65. Acharya S, Timilshina M, Jiang L, et al. Amelioration of Experimental autoimmune encephalomyelitis and DSS induced colitis by NTG-A-009 through the inhibition of Th1 and Th17 cells differentiation. Sci Rep 2018;8:7799. [Crossref] [PubMed]