Integrated bioinformatics and clinical validation of oxidative stress-associated genes in atrial fibrillation and chronic kidney disease

Abstract

Atrial fibrillation (AF) and chronic kidney disease (CKD) frequently coexist and worsen cardiovascular and renal outcomes. Oxidative stress (OS) contributes to both conditions, yet the OS-associated genes shared between the two conditions remain unclear. The present study aimed to identify OS-associated genes common to AF and CKD and explore potential therapeutic targets for AF-CKD comorbidity. Public microarray data for AF and CKD were obtained from the Gene Expression Omnibus. Differentially expressed genes (DEGs) were intersected with OS-associated genes to identify candidates, which then underwent functional enrichment and pathway analyses. A protein-protein interaction network was constructed to identify hub genes and three machine learning algorithms (Least Absolute Shrinkage And Selection Operator, support vector machine and random forest) obtained the final targets. Cell-type Identification by Estimating Relative Subsets of RNA Transcripts was used to estimate immune cell composition and assess the associations between target gene expression and immune cell subsets. Gene set variation analysis was used to delineate pathways associated with target genes. Transcription factor and microRNA regulatory networks were inferred using NetworkAnalyst and the Connectivity Map (CMap) database was queried to identify candidate small-molecule modulators. Target genes were validated in the peripheral blood of controls and patients with AF, CKD or concomitant AF and CKD. A total of 75 overlapping DEGs were identified between AF and CKD, 28 of which were associated with OS. Through a screening process integrating the three machine learning algorithms, three hub target genes were identified: BTG anti-proliferation factor 2 (BTG2), ZFP36 zinc finger CCCH-type (ZFP36) and natriuretic peptide B (NPPB). CMap analysis further suggested a number of candidate therapeutic compounds, such as histone deacetylase inhibitors, cyclin-dependent kinase inhibitors and proteasome inhibitors. In clinical samples, BTG2 and ZFP36 mRNA expression levels, as well as the concentrations of N-terminal pro-B-type natriuretic peptide, encoded by natriuretic peptide B (NPPB), were found to be significantly different between the control and disease groups. The present study identified BTG2, ZFP36 and NPPB as key OS-associated genes in AF and CKD. Furthermore, the three target genes were validated in human peripheral blood, supporting their potential as candidate biomarkers and therapeutic targets for cardiorenal comorbidities.

Introduction

Atrial fibrillation (AF) and chronic kidney disease (CKD) are notable global public health concerns and their co-occurrence has garnered increasing attention. Epidemiological data reveal that AF is present in 15-20% of patients with CKD, with the prevalence further increasing to 15-40% among those undergoing renal replacement therapy, such as dialysis (1,2). Despite the clinical importance and frequent comorbidity of AF and CKD, the molecular mechanisms associating these two conditions remain largely unclear, thus advancements in targeted therapeutic interventions are limited.

Previous studies have demonstrated marked associations between oxidative stress (OS) and both AF and CKD. An elevated oxidative burden may contribute to atrial electrical and structural remodeling through redox-sensitive alterations in atrial electrophysiology and fibrotic remodeling, thereby increasing susceptibility to AF and contributing to its maintenance (3). In CKD, excessive reactive oxygen species can damage lipids, proteins and DNA and promote inflammatory and fibrotic changes, thereby aggravating renal injury and accelerating disease progression (4). Antioxidant vitamins and other pharmacological agents have been investigated as potential approaches to reduce OS-related injury in AF and CKD (3,4). However, evidence specifically evaluating OS-targeted therapies in patients with concomitant AF and CKD remains limited (5). OS may represent a pathological association between these two conditions. In CKD, systemic alterations, such as the buildup of uremic solutes, ionic imbalance, imbalance of neuroendocrine signaling and overactivation of the renin-angiotensin-aldosterone system, leads to excessive OS and inflammatory activation. These processes not only exacerbate renal injury but also may foster a proarrhythmic substrate in the atrium (6). Nevertheless, AF and CKD have often been studied separately in different tissues and patient populations, leaving their shared OS-associated genes and molecular interactions insufficiently defined.

The present study aimed to elucidate OS-related molecular links between AF and CKD. Common differentially expressed genes (DEGs) identified from AF- and CKD-related GEO datasets were intersected with OS-associated genes retrieved from GeneCards, followed by functional-enrichment, PPI, small-molecule and machine-learning analyses. Candidate genes were further investigated through immune-infiltration, pathway and regulatory-network analyses and externally validated by RT-qPCR in peripheral-blood samples from patients at the Affiliated Zhongshan Hospital of Dalian University (Dalian, China).

Materials and methods

Data source

Expression profile datasets analyzed in the present study were acquired from the GEO (https://www.ncbi.nlm.nih.gov/geo/). As the training datasets, the AF dataset GSE115574(7) and the CKD dataset GSE37171(8) were used. For independent validation, the AF datasets GSE41177(9) and GSE14975(10) and the CKD datasets GSE66494(11) and GSE62792(12) were included. In the GEO AF-associated datasets (GSE115574, GSE41177 and GSE14975), all patients were diagnosed with persistent AF. With regard to the GEO CKD-associated datasets, GSE37171 consisted of patients with stage 5 CKD, GSE62792 included patients with stages 2-5 CKD and GSE66494 enrolled patients with CKD but did not provide CKD stage stratification in the original report (pathological severity grading was used instead). Specific information regarding datasets is presented in Table I.

Table I.

Gene Expression Omnibus microarray chip information.

Dataset characteristics	GSE115574	GSE41177	GSE14975	GSE37171	GSE66494	GSE62792
Disease	AF	AF	AF	CKD	CKD	CKD
Platform	GPL570	GPL570	GPL570	GPL570	GPL6480	GLP10558
Species	Homo	Homo	Homo	Homo	Homo	Homo
Tissue	Atrial tissue	Atrial tissue	Atrial tissue	Blood	Renal specimens	Blood
Samples in AF group	28	32	5	75	58	12
Samples in control group	31	6	5	40	8	6
AF type	Persistent AF	Persistent AF	Persistent AF	-	-	-
CKD stage	-	-	-	5	Not reported	2-5
PMID	34737791	23183193	20117462	40130991	2631775	29808448

AF, atrial fibrillation; CKD, chronic kidney disease; PMID, PubMed Identifier.

GeneCards is an integrated resource that compiles gene-centric information from numerous repositories. Genes associated with OS were recognized by their association with OS-associated pathways and diseases. OS-associated genes were retrieved from the GeneCards platform (version 5.24; https://www.genecards.org/) by querying the keyword ‘oxidative stress’, yielding a total of 7,648 genes. GeneCards with a relevance score >1 were retained for subsequent analyses.

Differential expression analysis and identification of DEGs

To identify genes with significantly differential expression, the ‘limma’ package (version 3.58.1) (13) in R software (Posit Software, PBC) was used to identify genes. DEGs were considered significant when exhibiting an adjusted P<0.05, calculated using the Benjamini-Hochberg method and log₂ fold change >0.263. After filtering, volcano plots and heatmaps were used for data visualization. These graphical representations were generated using ‘ggplot2’ (14) and ‘pheatmap’ (15) in R software (Posit Software, PBC).

Screening of candidate key genes

To comprehensively pinpoint key genes shared by AF and CKD a hierarchical intersection process was constructed based on dual disease datasets. First, DEG analysis was conducted on independent AF and CKD datasets according to a preset threshold of P value <0.05 and |log₂FC| ≥0.263(16). The respective upregulated and downregulated genes were obtained, and four sets of AF-upregulated/AF-downregulated and CKD-upregulated/CKD-downregulated genes were identified. Subsequently, to ensure consistent directions of expression change between the two diseases, only codirectional overlapping DEGs, comprising AF/CKD-upregulated and AF/CKD-downregulated genes, were retained. These genes were collectively defined as the codirectional overlapping gene set of AF- and CKD-associated DEGs and were used as the candidate genes. To focus on OS- associated mechanisms, this gene set was further intersected with OS-associated genes to obtain the set of codirectional OS-associated genes, hereafter referred to as the core gene set, which was subsequently used for functional enrichment analyses, network construction and further validation.

Construction of a protein-protein interaction (PPI) network

Core genes were first submitted to the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) online resource (https://cn.string-db.org/) to acquire a predicted PPI map. The exported interaction data were subsequently reconstructed and edited using Cytoscape software (version 3.9.1) (17).

Functional enrichment analysis

Key candidate genes that overlapped among AF, CKD and OS-associated DEGs were further examined through functional enrichment analyses. To characterize the biological processes and signaling cascades associated with these genes, overrepresentation analyses for Gene Ontology (GO) terms (18) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (19) were carried out. These procedures were implemented in R software (Posit Software, PBC) using the ‘clusterProfiler’ package (version 4.10.0) (20), and enrichment terms with a P<0.05 and a Benjamini-Hochberg-adjusted false discovery rate (FDR) <0.25 were considered to indicate a statistically significant difference. In addition, single-Gene Set Enrichment Analysis was performed with ‘clusterProfiler’, drawing on gene sets from the Molecular Signatures Database (‘c2.cp.kegg_legacy.v2025.1.Hs.symbols.gmt’, https://www.gsea-msigdb.org/gsea/msigdb). Statistically significant enrichments were identified using a dual threshold of P<0.05 and a Benjamini-Hochberg-adjusted P-value FDR<0.25. For each disease cohort, the top five significantly enriched entries in each category were retained and displayed. Enrichment outcomes were then graphically summarized with the ‘enrichplot’ R package (version 1.22.0) (21).

Connectivity map (CMap) analysis

To predict potential small-molecule agents targeting shared mechanisms in CKD and AF, the CMap resource (https://clue.io) was queried. Upregulated and downregulated genes associated with CKD, AF and OS were uploaded to the cMAP platform. Compounds, with expression profiles inversely correlated with these gene signatures, were considered potential therapeutic candidates. An enrichment score was calculated for each compound to quantify the degree of negative connectivity, with stronger negative scores indicating a greater potential to counteract disease-associated transcriptional patterns. On the basis of these scores, the 10 highest-ranking compounds were selected as small-molecule compounds.

Identification of target genes

To further select target genes, three distinct machine learning techniques were implemented: Support vector machine (SVM), Least Absolute Shrinkage And Selection Operator (LASSO) and random forest (RF). In the present study, the SVM algorithm was used to improve the classification accuracy of multidimensional data by employing the R package ‘e1071’ (Posit Software, PBC) (22). The SVM model used a linear kernel (kernel=‘linear’) for C classification with ‘cost=10’, ‘cachesize=500’ and ‘scale=FALSE’. Furthermore, fivefold cross-validation (k=5) with ‘halve.above=100’ was applied and the optimal feature subset was determined according to the minimum cross-validation error. In addition, LASSO regression utilized the ‘glmnet’ R package (Posit Software, PBC) to implement L1 regularization, thereby improving both the prediction performance and the effectiveness of feature selection. The random seed was fixed in 2025. Further tenfold cross-validation was conducted using the ‘cv.glmnet (nfolds=10)’ function to determine the optimal penalty parameter λ and the value corresponding to the minimum cross-validated binomial deviance (λ.min) was selected. Genes with non-zero coefficients at λ.min were retained as LASSO-selected genes. In addition, RF analysis was performed using the ‘randomForest’ package (23) with ‘ntree=1,000’. Gene importance was assessed through the ‘Mean Decrease Gini’ score and genes with importance scores >1 were retained. For all three analyses, the random seed was set to 2025. UpSet plots were generated to display the intersection of hub genes produced by the three machine learning methods.

Assessment of target gene robustness

Robustness of the selected target genes was assessed in both the training and validation cohorts by examining differences in their expression between the disease and control groups, with P<0.05 being considered to indicate a statistically significant difference. To assess diagnostic accuracy, ROC curves were created in R software (Posit Software, PBC) with the ‘pROC’ package (24) and discriminative ability was summarized by the corresponding area under the ROC curve (AUC).

Construction of diagnostic nomograms

To estimate the diagnostic utility of the target genes, nomograms based on their expression profile were constructed for each disease (AF and CKD). Nomogram construction and calibration were conducted using the ‘rms’ R package (Posit Software, PBC) (25). Calibration curves were plotted to examine the discriminative ability and validation of the nomograms.

Immune cell correlation analysis

Immune-cell proportions were estimated using the Cell-type Identification by Estimating Relative Subsets of RNA Transcripts (CIBERSORT) procedure with gene expression data (26). The associations between hub gene expression and these immune cell fractions were analyzed with single-gene correlation assessment and visualized in a heatmap. Given that this analysis was based on transcriptomic deconvolution, the results were interpreted as descriptive associations rather than evidence of direct gene-mediated immune regulation.

Regulatory network construction

To explore the potential regulatory mechanisms, transcription factor (TF)-mRNA and TF-microRNA (miRNA)-mRNA regulatory networks were constructed. TF-gene interactions were predicted using the JASPAR and TRANSFAC databases (27,28).

Patient sample collection

Blood samples were collected between September 2025 and April 2026 from control subjects (n=50), patients with AF (n=50), patients with CKD (n=48) and patients with both AF and CKD (n=47) at the Affiliated Zhongshan Hospital of Dalian University (Dalian, China). Efforts were made to match the groups for age, sex and key comorbidities to minimize potential confounding effects. The numbers of men/women were 30/20 in the control group, 31/19 in the AF group, 30/18 in the CKD group and 31/16 in the AF + CKD group, respectively. The median age was 59.5 (range: 39.0-76.0 years) in the control group, 62.0 (range: 40.0-82.0 years) in the AF group, 61.0 (range: 46.0-84.0 years) in the CKD group and 62.0 (range: 45.0-85.0 years) in the AF + CKD group. All patients with AF in the AF and AF + CKD groups had persistent AF. The CKD and AF + CKD groups comprised patients with CKD stages 2-5. AF was diagnosed on the basis of a standard 12-lead electrocardiogram and/or Holter electrocardiographic recordings, together with a review of medical records. Persistent AF was defined as AF sustained for >7 days or requiring cardioversion after ≥7 days. The CKD and AF + CKD groups comprised patients with CKD stages 2-5. CKD was diagnosed according to the Kidney Disease: Improving Global Outcomes criteria as abnormalities of kidney structure or function persisting for ≥3 months, defined by an estimated glomerular filtration rate (eGFR) <60 ml/min/1.73 m² and/or markers of kidney damage. CKD was staged according to eGFR categories. Participants with acute kidney injury, paroxysmal or transient AF, active infection, malignant disease, autoimmune disease, severe hepatic dysfunction, or other prespecified conditions that could affect gene-expression profiles were excluded. Detailed information regarding the cohort is provided in Table SI. Ethical approval was acquired from the Institutional Research Ethics Committee of the Affiliated Zhongshan Hospital of Dalian University (approval no. PJ-KY2025-194-1) and all participants provided written informed consent in accordance with the Declaration of Helsinki.

Reverse transcription-quantitative PCR (RT-qPCR) analysis of target genes

Total RNA was extracted from whole-blood samples following erythrocyte lysis using the RNA extraction reagent (Wuhan Servicebio Technology Co., Ltd.) according to the manufacturer's instructions. RNA quality was assessed using a NanoDrop 2000 (Thermo Fisher Scientific, Inc.). A total of 1 µg of RNA was reverse transcribed with 5x SweScript All-in-One SuperMix for qPCR (Wuhan Servicebio Technology Co., Ltd.) with genomic DNA removal (Wuhan Servicebio Technology Co., Ltd.) under the following conditions: 25˚C for 5 min, 42˚C for 30 min and 85˚C for 5 sec. RT-qPCR was conducted on a LightCycler 96 (Roche Diagnostics) using SYBR Green Master Mix (Wuhan Servicebio Technology Co., Ltd.). The thermal cycling program consisted of: 95˚C for 30 sec, followed by 40 cycles of 95˚C for 15 sec and 60˚C for 30 sec. Melting curve analysis verified the selectivity of the primers. Relative expression was quantified by the 2^-ΔΔCq approach using GAPDH as the reference gene (29). The primer sequences are presented in Table SII. Each sample was analyzed in both biological and technical triplicate. Data analysis and visualization were conducted using GraphPad Prism (version 10; Dotmatics).

Statistical analysis

Statistical analyses were performed using R software (version 4.3.3; R Foundation for Statistical Computing) in RStudio (version 2025.05.1+513; Posit Software, PBC). Continuous variables are presented as mean ± standard deviation (SD) or median (range), as appropriate, and categorical variables are presented as n (%). Differences among the four groups were assessed using one-way analysis of variance followed by Tukey's multiple-comparisons test. P<0.05 was considered to indicate a statistically significant difference. Each patient sample was analyzed in technical triplicate and the individual patients were considered biological replicates.

Results

DEG analysis

Fig. 1 outlines the overall analytical workflow of the present study. DEGs between the disease and control groups were identified using the ‘limma’ R package. In the AF dataset, 523 DEGs met the P<0.05 and log₂ FC>0.263 thresholds, including 295 upregulated transcripts and 228 downregulated transcripts. In the CKD dataset, 8,877 DEGs were identified under the same criteria, comprising 2,970 upregulated and 5,907 downregulated genes. DEG patterns are summarized and displayed as volcano plots and clustered heatmaps (Fig. 2A-D).

Figure 1.

Overview of the present study design. AF, atrial fibrillation; CKD, chronic kidney disease; OS, oxidative stress; DEG, differentially expressed genes; LASSO, least absolute shrinkage and selection operator; RF, random forest; SVM, support vector machine; BTG2, BTG anti-proliferation factor 2; NPPB, natriuretic peptide B; ZFP36, ZFP36 zinc finger CCCH-type; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; PPI-protein-protein interaction; GSVA, gene set variation analysis; RT-qPCR, reverse transcription quantitative-PCR.

Figure 2.

Codirectional DEGs shared by AF and CKD and identification of OS-associated candidates. Volcano plots depicting DEGs between (A) patients with AF and matched controls in the dataset GSE115574 and between (D) patients with CKD and controls in the dataset GSE37171. Red and blue points correspond to significantly increased and decreased transcripts, respectively, whereas gray points denote genes whose transcriptional changes did not meet the significance criteria (log₂ FC>0.263; P<0.05). Heatmaps illustrating the expression patterns of DEGs in (B) AF vs. control atrial tissue and in (E) CKD vs. control samples. Higher and lower expression levels are indicated by red and blue shading, respectively. (C) Venn diagram illustrating the intersection of upregulated and downregulated DEGs in the AF and CKD datasets. A total of 75 codirectional overlapping genes were identified and defined as AF-CKD codirectional DEGs. (F) Venn diagram summarizing the intersection between AF-CKD codirectional DEGs and OS-associated genes. A total of 28 common genes were retained and regarded as the core gene set for subsequent analyses. AF, atrial fibrillation; CKD, chronic kidney disease; DEGs, differentially expressed genes; FC, fold change; Up, upregulated; Down, downregulated; Not, not significantly differentially expressed; OS, oxidative stress.

Identification of key genes

To identify common molecular signals that are directionally consistent between AF and CKD, the upregulated and downregulated DEGs for the two diseases and identified their codirectional intersections were extracted. As presented in Fig. 2E, 75 codirectional overlapping genes (36 upregulated and 39 downregulated genes) were identified in this manner. Furthermore, these 75 genes were intersected with 7,648 OS-associated genes retrieved from the GeneCards database; ultimately, 28 codirectional overlapping genes associated with OS were identified (Fig. 2F). These 28 genes (17 upregulated and 11 downregulated) were ultimately defined as the target genes of the present study and were used for subsequent analysis.

Functional enrichment analysis

A panel of 28 hub genes was subjected to GO and KEGG enrichment to explore the key biological functions shared by AF and CKD. According to the GO results (Table II and Fig. 3A), these genes were enriched predominantly in biological processes (BP) associated with ‘collagen fibril organization’, ‘nuclear-transcribed mRNA catabolic process, deadenylation-dependent decay’, ‘nuclear-transcribed mRNA poly(A) tail shortening’, ‘regulation of nuclear-transcribed mRNA poly(A) tail shortening’ and ‘positive regulation of nuclear-transcribed mRNA poly(A) tail shortening’. With regard to cellular components, the genes were notably associated with terms such as ‘collagen-containing extracellular matrix’, ‘lysosomal lumen’, ‘Golgi lumen’, ‘Z disc’ and ‘I band’. With regard to molecular function, the genes were notably associated with terms such as ‘extracellular matrix structural constituent’, ‘cytokine binding’, ‘transcription corepressor activity’, ‘extracellular matrix structural constituent conferring compression resistance’ and ‘growth factor binding’. KEGG pathway analysis further indicated that these genes were enriched in extracellular matrix ‘(ECM)-receptor interaction’, ‘Focal adhesion’ and ‘RNA degradation’ (Table II and Fig. 3A).

Table II.

Results of GO and KEGG enrichment analysis for 28 key differentially expressed genes.

Ontology	ID	Description	Gene ratio	Bg ratio	P-value	P.adjust
BP	GO:0060213	Positive regulation of nuclear-transcribed mRNA poly(A) tail shortening	3/28	13/18,800	8.38x10-7	0.0006
BP	GO:0060211	Regulation of nuclear-transcribed mRNA poly(A) tail shortening	3/28	15/18,800	1.33x10-6	0.0006
BP	GO:0030199	Collagen fibril organization	4/28	62/18,800	2.07x10-6	0.0006
BP	GO:0000289	Nuclear-transcribed mRNA poly(A) tail shortening	3/28	28/18,800	9.45x10-6	0.0020
BP	GO:0000288	Nuclear-transcribed mRNA catabolic process, deadenylation-dependent decay	3/28	63/18,800	0.0001	0.0188
CC	GO:0062023	Collagen-containing extracellular matrix	5/28	429/19,594	0.0003	0.0147
CC	GO:0043202	Lysosomal lumen	2/28	97/19,594	0.0084	0.0979
CC	GO:0005796	Golgi lumen	2/28	104/19,594	0.0096	0.0979
CC	GO:0030018	Z disc	2/28	127/19,594	0.0141	0.0979
CC	GO:0031674	I band	2/28	139/19,594	0.0167	0.0979
MF	GO:0005201	ECM structural constituent	5/27	172/18,410	4.59x10-6	0.0005
MF	GO:0019955	Cytokine binding	4/27	141/18,410	5.05x10-6	0.0027
MF	GO:0003714	Transcription corepressor activity	4/27	191/18,410	0.0002	0.0059
MF	GO:0030021	ECM structural constituent conferring compression resistance	2/27	22/18,410	0.0005	0.0128
MF	GO:0019838	Growth factor binding	3/27	139/18,410	0.0011	0.0235
KEGG	hsa04512	ECM-receptor interaction	3/14	88/8,164	0.0004	0.0117
KEGG	hsa04510	Focal adhesion	3/14	201/8,164	0.0044	0.0635
KEGG	hsa03018	RNA degradation	2/14	79/8,164	0.0078	0.0754

GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; ECM, extracellular matrix; BP, biological processes; CC, cellular component; MF, molecular function; P.adjust, adjusted P-value; Bg, background.

Figure 3.

Enrichment profiling and interaction landscape of the core genes. (A) Overrepresentation analysis of the core gene set in GO and KEGG. GO annotations are grouped into three domains: BP, CC and MF. Bar length reflects the Gene ratio and bar color denotes the corresponding GO/KEGG category. (B) Interaction map of the core genes at the protein level generated from the Search Tool for the Retrieval of Interacting Genes/Proteins resource and rendered in Cytoscape. Each node represents an individual gene product; nodes with deeper colors indicate higher degrees of connection within the network. GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; BP, biological process; CC, cellular component; MF, molecular function; ECM, extracellular matrix.

PPI

All 28 key genes were submitted to STRING to construct a PPI network. After the isolated nodes were discarded, the final network contained 14 nodes and 16 edges. The interactions among these genes were displayed using Cytoscape (Fig. 3B). The 14 nodes comprised COL1A1, FMOD, NES, PXDN, RELN, SH3KBP1, ZFP36, TOB1, FHL2, BTG2, BGN, COMP, CSRP3 and NPPB.

Identification of target genes

To identify target genes, 14 genes obtained from the PPI interaction network were subsequently subjected to LASSO, random forest and SVM analyses in each of the two disease datasets. In the AF dataset, LASSO regression identified 10 candidate genes (FMOD, RELN, SH3KBP1, ZFP36, TOB1, FHL2, BTG2, COMP, CSRP3 and NPPB) based on the cross-validation curve and coefficient profiles (Fig. 4A and B). Random Forest analysis ranked the top 10 genes according to the Mean Decrease Gini index (CSRP3, COMP, NPPB, SH3KBP1, RELN, BTG2, NES, FHL2, ZFP36 and TOB1; Fig. 4C). SVM-recursive feature elimination identified 12 candidate genes (COMP, TOB1, ZFP36, CSRP3, COL1A1, SH3KBP1, PXDN, FHL2, BTG2, FMOD, NPPB and NES; Fig. 4G and H). In the CKD dataset, LASSO regression identified six candidate genes (FMOD, RELN, ZFP36, BTG2, BGN and NPPB; Fig. 4D and E). Random Forest analysis ranked the top 10 genes as BGN, COMP, TOB1, NPPB, RELN, PXDN, BTG2, ZFP36, FHL2 and COL1A1 (Fig. 4F), whereas SVM-recursive feature elimination identified 13 candidate genes (BGN, NPPB, RELN, FMOD, PXDN, SH3KBP1, ZFP36, FHL2, BTG2, COL1A1, NES, TOB1 and COMP; Fig. 4I and J). As presented in Fig. 4K, after the intersections were verified, three target genes, namely, ZFP36, BTG2 and NPPB, were identified.

Figure 4.

Screening AF- and CKD-associated target genes with integrated machine learning approaches. LASSO regression analysis of candidate genes in the AF dataset. (A) The tenfold cross-validation for parameter (λ) selection showing the minimum partial likelihood deviance. (B) LASSO coefficient profiles as a function of log (λ). (C) RF analysis identifying the top-ranking AF-associated genes based on the ‘Mean Decrease Gini’ index. LASSO regression for feature selection in the CKD dataset. (D) The tenfold cross-validation curve for optimal λ selection and (E) corresponding coefficient profiles. (F) RF analysis identifying the most important CKD-associated genes ranked by the ‘Mean Decrease Gini’ index. SVM analysis of the AF dataset. (G) Number of selected features vs. classification accuracy. (H) Cross-validation error curve showing the optimal feature subset. Support vector machine-recursive feature elimination analysis of the CKD dataset. (I) Number of selected features vs. classification accuracy. (J) Cross-validation error curve indicating the optimal number of features. (K) UpSet plot showing the intersection of gene sets identified by the LASSO, RF and SVM models in the AF and CKD datasets. The genes shared across all three algorithms were considered target genes. LASSO, least absolute shrinkage and selection operator; RF, random forest; SVM, support vector machine; AF, atrial fibrillation; CKD, chronic kidney disease; CV, cross validation.

Validation of target genes

Expression of the three target genes was first examined. As presented in Fig. 5A and B, NPPB was found to be significantly upregulated in the disease groups, whereas BTG2 and ZFP36 were found to be significantly downregulated (compared with control groups). In the AF dataset (GSE115574), ROC analysis yielded AUCs of 0.758 for NPPB, 0.671 for BTG2 and 0.657 for ZFP36 (Fig. 5C). Similarly, in the CKD dataset (GSE37171), the AUCs were 0.924 for NPPB, 0.877 for BTG2 and 0.886 for ZFP36 (Fig. 5D), indicating relatively favorable discriminative performance in this dataset. To further assess the generalizability of these genes, the present study additionally examined two AF datasets (GSE14975 and GSE41177; Fig. 5C) and two CKD datasets (GSE62792 and GSE66494; Fig. 5D). In these additional cohorts, all three genes exhibited AUC values >0.60. However, given the very limited sample sizes in some external datasets, the present ROC results should be considered preliminary and interpreted with caution. Therefore, the external analyses of the present study provide only exploratory gene-level support rather than rigorous external validation.

Figure 5.

Validation and diagnostic performance evaluation of core genes and construction of nomogram models for AF and CKD. Violin plots of BTG2, NPPB and ZFP36 expression in the (A) AF (GSE115574) and (B) CKD (GSE37171) datasets. Red indicates patients with diseases, and blue indicates controls. Statistical significance is indicated as follows: ^*P<0.05 and ^***P<0.001. ROC analyses illustrating the diagnostic performance of the three core genes across numerous (C) AF (GSE115574, GSE14975, and GSE41177) and (D) CKD (GSE37171, GSE62792, and GSE66494) datasets. The AUC values for each gene are labeled in the plots. Development and evaluation of prediction nomograms derived from BTG2, NPPB and ZFP36 in (E) the AF (EGSE115574) and (F) CKD (SE37171) datasets. The left panels display the nomogram models, the middle panels show the calibration curves for assessing the predictive consistency and the right panels present the ROC curves for evaluating the discriminatory performance of the models. ROC, receiver operating characteristic; AUC, area under the curve; BTG2, BTG anti-proliferation factor 2; NPPB, natriuretic peptide B; ZFP36, ZFP36 zinc finger CCCH-type; AF, atrial fibrillation; CKD, chronic kidney disease; FPR, false positivity rate; TPR, true positivity rate.

Nomogram development based on target genes

A nomogram integrating the scores of the three target genes was constructed (Fig. 5E and F). The calibration curves exhibited acceptable agreement between the predicted and observed outcomes in the present dataset (Fig. 5E and F). The calculated AUCs for the nomogram were 0.821 for AF and 0.966 for CKD (Fig. 5E and F), indicating favorable discriminative ability in the current dataset. However, as no independent external cohort was available for model-level validation, the performances of the nomograms should be interpreted with caution.

Immune microenvironment analysis

CIBERSORT algorithms were utilized to profile immune cell infiltration in the disease and control groups of both datasets; these profiles are presented as stacked bar plots (Fig. 6A and B). The results for the AF group revealed significant increases in the numbers of regulatory T cells and monocytes (Fig. 6C). In the CKD dataset, significant differences were observed in the estimated fractions of a number of immune cell subsets, including B cell populations, plasma cells, CD8⁺/γδ T cells, resting natural killer cells, monocytes and M0 macrophages (Fig. 6D).

Figure 6.

Immune infiltration analysis and regulatory network prediction of target genes in AF and CKD. Stacked bar charts depicting the estimated fractions of 22 immune cell populations inferred by the Cell-type Identification by Estimating Relative Subsets of RNA Transcripts algorithm in the (A) AF (GSE115574) and (B) CKD (GSE37171) datasets. Different colors represent distinct immune cell subtypes. Boxplots comparing the levels of infiltrating immune cells between patients and controls in the (C) AF and (D) CKD cohorts. Red and blue indicate the disease and control samples, respectively. Statistical significance is indicated as follows: ^***P<0.001, ^**P<0.01 and ^*P<0.05. Heatmaps displaying the correlations between target genes (BTG2, NPPB and ZFP36) and immune cell infiltration levels in the (E) AF and (F) CKD datasets. Red denotes a positive correlation and blue denotes a negative correlation. (G) Interaction network between target genes TFs predicted using the NetworkAnalyst platform. (H) An integrated regulatory map associating target genes, TFs, and microRNAs was generated in NetworkAnalyst. Nodes represent genes, TFs, and miRNAs, while edges reflect their predicted regulatory associations. AF, atrial fibrillation; CKD, chronic kidney disease; TF, transcription factor; NK, natural killer; BTG anti-proliferation factor 2; NPPB, natriuretic peptide B; ZFP36, ZFP36 zinc finger CCCH-type.

To further assess the associations between hub-gene expression and the CIBERSORT-inferred immune cell fractions, Spearman's rank correlation analysis was performed, and the results were visualized using correlation heatmaps. In the AF dataset, NPPB expression was found to be positively correlated with activated mast-cell fractions and negatively correlated with regulatory T cells and resting mast cells. BTG2 expression was negatively associated with resting mast cells fractions, whereas ZFP36 expression was positively associated with follicular helper T-cell and activated mast-cell fractions and negatively associated with resting dendritic-cell and resting mast-cell fractions (Fig. 6E). In the CKD dataset, NPPB expression was positively correlated with plasma cell and monocyte fractions, whereas BTG2 and ZFP36 expression were both positively correlated with neutrophil fractions (Fig. 6F). Collectively, these results suggest that the hub gene expression is associated with distinct immune cell composition patterns in the AF and CKD datasets. As the present results were derived from bulk transcriptomic deconvolution and correlation analyses, they should be interpreted as descriptive associations rather than evidence of direct effects of these genes on immune cell recruitment, activation, polarization or disease progression.

TF-gene associations and TF-miRNA cooperative regulatory networks

TF-gene associations were obtained using NetworkAnalyst. TF-gene interactions were analyzed for each targeted gene (ZFP36, BTG2 and NPPB). The interactions between the TFs and key genes are presented in Fig. 6G and H. The network contained 16 nodes and 15 edges. Among them, ZFP36 was regulated by five TFs [SP1, upstream transcription factor 2, transcription factor AP-2, paired box 2 (PAX2) and histone H4 transcription factor (HINFP)]. BTG2 was regulated by six TFs [HINFP, FOXC1, E2F transcription factor 1 (E2F1), POU class 2 homeobox 2, AT-rich interaction domain 3A (ARID3A) and RELA proto-oncogene, NF-κB subunit (RELA)] and NPPB was regulated by four TFs [FOXC1, sterol regulatory element binding transcription factor 2, TP53 and nuclear factor κ B subunit 1 (NFKB1); Fig. 6G]. In addition, the TF-miRNA coregulatory network was analyzed to identify potential interaction effects of miRNAs and TFs on target genes. This network, which may underlie the regulation of DEGs, consisted of 142 nodes and 145 edges (Fig. 6H and Table SIII). Together, these predicted regulatory interactions may provide insights into the potential transcriptional and post-transcriptional mechanisms underlying the regulation of OS-associated genes shared by AF and CKD.

Candidate small-molecule compounds for AF and CKD

To screen potential small-molecule therapeutics for AF and CKD, the overlapping upregulated and downregulated genes between AF_DEGs and CKD_DEGs were submitted to the CMap database. Based on the CMap scores, ten compounds, namely, vorinostat, MG-132, BNTX, AZD-5438, R-547, alvocidib, 5-iodotubercidin, bortezomib, genz-644282 and AT-7519, were prioritized as predicted small-molecule compounds potentially associated with the shared molecular mechanisms of AF and CKD (Fig. 7A). The predicted targets and mechanisms of action, together with the chemical structures of these molecules, are summarized in Fig. 7B and C. These compounds may represent candidate pharmacological leads capable of modulating the shared transcriptional signatures associated with AF and CKD. However, these computational predictions do not establish their efficacy or safety in patients with AF and CKD and require further experimental validation.

Figure 7.

Screening candidate therapeutic agents using CMap. (A) Heatmap illustrating the ten most strongly enriched small-molecule compounds identified by CMap on the basis of the expression pattern of the target genes. The x-axis shows the numerous cell lines included in the CMap resource and the y-axis lists the corresponding compounds. The colors indicate connectivity scores, with red indicating a positive association and blue indicating a negative association to the gene signature associated with the disease. (B) Summary table of the 10 candidate drugs, including the Pert_id, compound name and pharmacological description of each. The identified compounds included HDAC inhibitors, CDK inhibitors, proteasome inhibitors and other signaling modulators, suggesting their potential therapeutic effects in atrial fibrillation and chronic kidney disease. (C) The two-dimensional molecular structures of predicted compounds (vorinostat, MG-132, BNTX, AZD-5438, R-547, alvocidib, 5-iodotubercidin, bortezomib, Genz-644282 and AT-7519), showing their chemical configurations. cMAP, Connectivity Map; HDAC, histone deacetylase; Pert_id, perturbation ID.

Pathway enrichment profiles of BTG2, ZFP36 and NPPB identified by gene set variation analysis (GSVA)

GSVA results indicated an association between target genes and distinct pathway modules across the datasets (Fig. 8). In AF, high ZFP36 expression was enriched in ‘KEGG_SMALL_CELL_LUNG_CANCER’, ‘KEGG_EITHER_LIPID_METABOLISM’ and ‘KEGG_EPITHELIAL_CELL_SIGNALING_IN_HELICOBACTER_PYLORI_INFECTION’ and was inversely associated with ‘KEGG_TAURINE_AND_HYPOTAURINE_METABOLISM’ and ‘KEGG_MATURITY_ONSET_DIABETES_OF_THE_YOU NG’ (Fig. 8A). BTG2 in AF preferentially engaged ‘KEGG_CELL_ADHESION_MOLECULES_CAMS, ‘KEGG_HEMATOPOIETIC_CELL_LINEAGE’ and ‘KEGG_AUTOIMMUNE_THYROID_DISEASE’ and was depleted in ‘KEGG_STARCH_AND_SUCROSE_METABOLISM’ (Fig. 8B). NPPB in AF aligned with ‘KEGG_GLYCOSAMINOGLYCAN_BIOSYNTHESIS_CHONDROITIN_SULFATE’, ‘KEGG_COLORECTAL_CANCER’ and ‘KEGG_TGF_BETA_SIGNALING_PATHWAY’, while repressing ‘KEGG_DRUG_METABOLISM_CYTOCHROME_P450, ‘KEGG_METABOLISM_OF_XENOBIOTICS_BY_CYTOCHROME_P450’ and ‘KEGG_RETINOL_METABOLISM’ (Fig. 8C). In CKD, ZFP36 was associated with ‘KEGG_NOTCH_SIGNALING_PATHWAY’ and host-pathogen pathways including ‘KEGG_LEISHMANIA_INFECTION’ and ‘KEGG_PATHOGENIC_ESCHERICHIA_COLI_INFECTION’ and showed a negative association with ‘KEGG_DRUG_METABOLISM_CYTOCHROME_P450’, ‘KEGG_METABOLISM_OF_XENOBIOTICS_BY_CYTOCHROME_P450’ and ‘KEGG_TYROSINE_METABOLISM’ (Fig. 8D). BTG2 in CKD favored morphogenesis/adhesion and infection pathways (including ‘KEGG_DORSO_VENTRAL_AXIS_FORMATION’, ‘KEGG_ADHERENS_JUNCTION’ and ‘KEGG_PATHOGENIC_ESCHERICHIA_COLI_INFECTION’) and dow nregulated ‘KEGG_TYROSINE_METABOLISM’ and ‘KEGG_DRUG_METABOLISM_CYTOCHROME_P450’ (Fig. 8E). NPPB in CKD was found to be associated with metabolic regulatory pathways (including ‘KEGG_MATURITY_ONSET_DIABETES_OF_THE_YOUNG’, ‘KE GG_GLYCOSPHINGOLIPID_BIOSYNTHESIS_LACTO_AND_NEOLACTO_SERIES’ and ‘KEGG_PPAR_SIGNALING_PATHWAY’) and further suppressed ‘KEGG_AMINO_SUGAR_AND_NUCLEOTIDE_SUGAR_METABOLISM’, ‘KEGG_MTOR_SIGNALING_PATHWAY’ and ‘KEGG_NON_SMALL_CELL_LUNG_CANCER’ (Fig. 8F).

Figure 8.

GSVA-based functional profiling of target genes in AF and CKD datasets. GSVA enrichment results for (A) ZFP36, (B) BTG2 and (C) NPPB in the AF dataset (GSE115574). The bars represent the top enriched KEGG pathways that were significantly correlated with each gene. Red bars correspond to pathways that were positively associated with gene expression, whereas blue bars denote pathways showing negative associations. GSVA enrichment results for (D) ZFP36, (E) BTG2 and (F) NPPB in the CKD dataset (GSE37171). The length of each bar reflects the enrichment score for the corresponding KEGG pathway. Red and blue, respectively, indicate pathways that were positively and negatively associated with the corresponding genes. GSVA, gene set variation analysis; AF, atrial fibrillation; CKD, chronic kidney disease; BTG2, BTG anti-proliferation factor 2; NPPB, natriuretic peptide B; ZFP36, ZFP36 zinc finger CCCH-type; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Clinical validation of target genes in patient specimens

Expression levels of the genes ZFP36, BTG2 and NPPB were evaluated in peripheral blood samples using RT-qPCR. As presented in Fig. 9A, ZFP36 expression was found to be significantly greater in the AF, CKD and AF + CKD groups compared with the control group (5.08±1.52; P<0.05; 5.99±1.904, P<0.01; and 28.66±15.55, P<0.001, respectively, vs. 1.01±0.268). The AF + CKD group exhibited the highest relative ZFP36 mRNA level among the four groups. As presented in Fig. 9B, BTG2 expression was significantly greater in the CKD group (2.10±0.95; P<0.001) and the AF + CKD group (3.27±2.56; P<0.0001) compared with the control group (1.05±0.35). As NPPB mRNA was barely detectable by qPCR in whole blood, its secreted product, NT-proBNP, was instead assessed. NT-proBNP levels were found to be significantly greater in the CKD group (19,036.00±7,823.00 ng/ml; P<0.0001) and the AF + CKD group (20,261.00±10,033.00 ng/ml; P<0.0001) compared with the control group (130.40±76.23 ng/ml; Fig. 9C). These findings provide clinical evidence consistent with the bioinformatics results, showing differential expression of ZFP36 and BTG2 in patients with concomitant AF and CKD. Because NPPB mRNA was barely detectable in whole blood, the increased NT-proBNP levels provide only indirect evidence of NPPB-related cardiac stress rather than direct confirmation of NPPB transcription.

Figure 9.

Validation of BTG2, ZFP36, and NT-proBNP expression levels in peripheral blood. Quantitative PCR analysis of (A) BTG2 and (B) ZFP36 mRNA expression levels in human whole blood from the control, AF, CKD and AF + CKD groups. The data are presented as relative mRNA levels normalized to those in the internal control. (C) Serum concentrations of NT-proBNP, encoded by NPPB, measured in the same four groups. Statistical significance is indicated as follows: ^*P<0.05, ^**P<0.01, ^***P<0.001 and ^****P<0.0001. BTG2, BTG anti-proliferation factor 2; ZFP36, ZFP36 zinc finger CCCH-type; NPPB, natriuretic peptide B; AF, atrial fibrillation; CKD, chronic kidney disease.

Discussion

Both AF and CKD incidence rates are increasing, particularly in aging populations. In addition, these conditions frequently coexist due to overlapping risk factors and pathophysiological mechanisms (30,31). Accumulating evidence has established OS as a key mechanistic association between AF and CKD (32,33). Accordingly, modulation of OS-associated pathways could offer a new therapeutic avenue for patients with both conditions.

GO and KEGG pathway enrichment of the genes shared by AF, CKD and OS revealed common molecular mechanisms predominantly involving ECM remodeling, RNA metabolism and adhesion-associated signaling pathways. GO analysis revealed enrichment in terms such as ‘collagen fibril organization’ and ‘collagen-containing extracellular matrix’, reflecting enhanced ECM metabolic activity and a tendency toward fibrosis. Previous studies have indicated that CKD can trigger the TGF-β/Smad pathway, upregulating the expression of type I collagen and α-smooth muscle actin, thereby promoting atrial interstitial fibrosis and structural remodeling and increasing AF susceptibility (34,35). The present results support this hypothesis. Notably, excessive deposition of the fibrotic matrix not only alters the mechanical properties of tissues but also fosters a microenvironment in the myocardium and kidney that favors OS and local inflammation (36,37). In addition, RNA metabolism-associated BP terms, such as ‘nuclear transcribed mRNA catabolic process, deadenylation-dependent decay’ and ‘poly(A) tail shortening’ indicated that RNA stability regulation and posttranscriptional control may represent important adaptive mechanisms in response to OS. Previous evidence has demonstrated that small RNAs can promote mRNA demethylation and degradation, thereby regulating ion channel gene expression and contributing to atrial electrical remodeling (38). Importantly, sRNAs not only involved in altering the electrical activity of the atrium but also regulate gene networks associated with fibrosis, inflammatory signaling and tubular epithelial cell injury, thus promoting CKD progression and indicating an association between cardiac and renal remodeling at the molecular level (39). KEGG enrichment pointed to pronounced involvement of the ‘ECM-receptor interaction’ and ‘Focal adhesion’ pathways, highlighting the importance of cell-matrix contacts and adhesion-mediated signaling in the overlapping pathogenesis of AF and CKD. Consistent with previous reports, dysregulation of ECM-receptor signaling has been associated with atrial structural remodeling and the return of AF after treatment (40). In addition, focal adhesion kinase has been shown to drive AF remodeling through activation of the Akt/ribosomal protein S6 kinase pathway (41). Similarly, altered cell-extracellular matrix interactions, particularly integrin-mediated adhesion signaling, have been implicated in glomerular injury and fibrotic remodeling in CKD, including focal segmental glomerulosclerosis (42,43). In summary, the present findings suggest that the interaction among ECM remodeling, adhesion signaling and RNA regulatory pathways forms the molecular basis of OS-mediated cardiac-renal remodeling. These mechanisms provide new insights into the pathophysiological associations between AF and CKD and may help identify potential therapeutic targets for future studies.

BTG2, belongs to the anti-proliferative gene family and is a classical downstream target of p53, with key roles in cell cycle regulation, OS response, apoptosis and metabolic homeostasis (44,45). Previous studies have demonstrated that the role of BTG2 in cardiac and renal diseases is dependent on cell type and disease context dependency. In the cardiovascular system, BTG2 upregulation can reduce MMP2/MMP9-dependent ECM degradation and restrain abnormal cardiomyocyte proliferation, thereby exerting anti-inflammatory and antifibrotic effects that attenuate cardiac hypertrophy and interstitial remodeling (46). In a number of chronic kidney injury models, BTG2 is often upregulated and can promote the expression of collagen I/III and α-smooth muscle actin, thereby accelerating glomerulosclerosis and interstitial fibrosis (47,48). These results suggest that BTG2 may be involved in pathological processes such as inflammation, OS and ECM remodeling, but its specific role is not entirely consistent across different organs and disease states. Transcriptome analysis from public databases in the present study revealed that BTG2 expression tended to decrease in the disease group. However, peripheral blood RT-qPCR validation revealed that BTG2 expression levels were significantly higher in the AF + CKD group compared with the control group, while no significant difference was observed between the AF group and the control group. Therefore, despite BTG2 being considered a candidate gene for AF- and CKD-related transcriptional changes, evidence supporting its role as a stable diagnostic marker or key hub gene remains limited. Further validation in independent cohorts and clarification of its potential function through functional studies are still needed.

ZFP36 (tristetraprolin) is a pivotal RNA-binding protein that identifies AU-rich elements in target mRNAs and promotes their deadenylation and decay, thereby exerting central control over inflammation, OS and metabolic regulation (49). Despite direct research on the role of ZFP36 in the onset, maintenance and atrial remodeling processes of AF being limited, previous studies have demonstrated that the expression of ZFP36 is elevated in atherosclerosis-associated endothelial cells and peripheral blood mononuclear cells of patients undergoing maintenance hemodialysis (50,51). These findings were accompanied by differential regulation of inflammation-associated signaling pathways, including Toll-like receptor, MAPK and chemokine signaling (50), whereas ZFP36 was shown to reduce inflammatory responses in aortic endothelial cells (51).

These results suggest that abnormal expression of ZFP36 may be associated with alterations in the peripheral immune and inflammatory microenvironment related to CKD. In addition, the ZFP36 family member ZFP36L1 (Tis11b) can regulate the stress response of tubular cells under hyperosmotic conditions by reducing the stability of mineralocorticoid receptor mRNA (52). In the present study, peripheral blood ZFP36 mRNA expression was found to be significantly increased in the AF group, CKD group and AF + CKD group compared with the control group, with the highest expression observed in the AF + CKD group. These results suggest that abnormal ZFP36 expression may be associated with systemic inflammation, stress response or changes in the composition of circulating immune cells in the comorbid state of AF and CKD. Considering that ZFP36 is involved in the posttranscriptional regulation of inflammation-associated mRNAs, its elevated expression may reflect a regulatory response induced by inflammatory stimuli; however, this hypothesis does not establish whether ZFP36 serves a protective or pathogenic role. In addition, whole-blood ZFP36 levels may be influenced by factors such as the stage of renal function, volume status, comorbidities, pharmacological treatments and the proportion of blood cell subpopulations (50,53). Therefore, the results of the present study merely suggests that ZFP36 is associated with the inflammation-stress characteristics in the context of comorbid AF and CKD; its cellular origin, downstream targets and specific functions still require further validation.

NPPB encodes pre-proBNP, which is processed to produce BNP and NT-proBNP, primarily reflecting the state of myocardial stress and is associated with cardiac load, fibrosis, OS and metabolic regulation. Previous studies have indicated that NPPB expression is elevated in heart failure, hypertrophic cardiomyopathy and dilated cardiomyopathy and is often accompanied by increased ventricular wall tension and OS (54-56). In addition, the NPPB/BNP pathway may promote angiogenesis, inhibit fibrosis and facilitate tissue repair (57). Its expression can be regulated by OS, inflammation and epigenetic mechanisms, involving signaling pathways such as JAK/STAT3, NF-κB and p53 (58,59). In addition, NPPB has been identified as a shared hub gene in interactome analyses linking obesity, type 1 diabetes and cardiovascular disease (60), and circulating NPPB levels have been associated with incident cardiovascular disease (61). Bioinformatics analysis in the present study revealed that NPPB was upregulated in the AF and CKD-associated datasets; however, no reliable amplification signal was obtained by peripheral blood qPCR validation, which may be due to NPPB being expressed primarily in ventricular cardiomyocytes and the low abundance of its transcript in circulating blood cells (58,62). Therefore, the present study further measured plasma NT-proBNP levels to provide supplementary, indirect evidence for alterations in the NPPB/BNP-associated pathway. The present results revealed that plasma NT-proBNP levels were significant elevated in the AF, CKD and AF + CKD groups, which is consistent with previous findings that NT-proBNP is associated with the onset and recurrence of AF and adverse cardiovascular outcomes (63-66). Despite direct evidence associating NPPB with CKD still being limited, existing studies have shown that elevated levels of BNP and NT-proBNP are associated with decreased renal function and adverse cardiorenal outcomes (67-69). However, the elevated NT-proBNP levels in the CKD and AF + CKD groups should still be interpreted with caution. In addition to increased cardiac pressure or volume load, reduced renal clearance and systemic accumulation due to impaired kidney function can also markedly affect its concentration. Therefore, elevated NT-proBNP does not directly indicate enhanced NPPB transcription or increased myocardial secretion; especially in patients with AF + CKD, cross-sectional measurements make it difficult to distinguish the relative contributions of cardiac stress, changes in volume status and renal impairment. Overall, in the present study, NT-proBNP testing provides only indirect and exploratory clinical support for abnormalities in the NPPB/BNP-associated pathway and further validation with atrial or myocardial tissue testing as well as larger independent cohorts is still needed.

To identify potential small-molecule compounds associated with the shared molecular signatures of AF and CKD, CMap analysis was conducted on the basis of core OS-associated genes and ten compounds with possible modulatory effects (vorinostat, MG-132, BNTX, AZD-5438, R-547, alvocidib, 5-iodotubercidin, bortezomib, Genz-644282 and AT-7519) which could be categorized into three major pharmacological classes, were predicted. Among these compounds, vorinostat is a histone deacetylase (HDAC) inhibitor that has displayed potential protective effects in both AF and CKD. HDAC inhibition can alleviate atrial fibrosis, regulate calcium homeostasis and reduce inflammation and OS, thereby preventing AF-associated structural and electrical remodeling (70-73). In CKD, HDAC inhibitors modulate epigenetic networks and suppress TGF-β and inflammation, attenuating renal fibrosis and functional decline (74,75). Thus, targeting HDAC-mediated epigenetic regulation may offer a therapeutic strategy with dual cardiorenal benefits for AF and CKD comorbidity. In addition, AZD-5438, R-547, alvocidib (flavopiridol) and AT-7519 are representative CDK inhibitors. CDK inhibitors, originally developed as anticancer agents, have recently been explored for their anti-inflammatory potential. Despite direct evidence of their effects on AF and CKD remains limited, studies have suggested that CDK inhibitors exhibit promising activity in suppressing neutrophil-mediated inflammatory responses (76,77). As neutrophils, the most abundant leukocytes, serve a central role in inflammation, which is a key mechanistic association between AF and CKD (78,79), CDK inhibitors may represent novel candidates for treating AF-CKD comorbidity. However, their precise mechanisms and safety profiles require further investigation. In addition, proteasome inhibitors, such as MG-132 and bortezomib, primarily used in multiple myeloma therapy, have been shown to have immunomodulatory and anti-inflammatory effects on renal diseases, reducing proteinuria and increasing the glomerular filtration rate in patients with IgA nephropathy (80). Collectively, the compounds predicted by CMap may act through diverse mechanisms, such as epigenetic modulation, cell cycle regulation and proteostasis control, to alleviate OS and inflammation, providing new insights into multitarget therapeutic strategies for AF and CKD comorbidity.

TFs control gene expression through interactions with defined DNA motifs and through the orchestration of numerous physiological processes. Through bioinformatics analysis, the present study revealed a number of key TFs associated with AF and CKD, including SP1, E2F1, RELA (NF-κB), TP53, PAX2 and ARID3A, revealing common regulatory mechanisms involving OS, inflammatory responses and fibrosis. As members of the NF-κB family, RELA and NFKB1 serve central roles in the development of AF and CKD. Studies have demonstrated that NF-κB participates in renal fibrosis, cellular senescence and inflammatory responses through certain mechanisms, including histone lactylation, PANoptosis regulation and the sirtuin 1 pathway (81-83). The TF SP1 is also central to the development of fibrosis as well as OS. Accumulating evidence has suggested that SP1 facilitates disease progression in CKD together with AF by regulating renal interstitial fibrosis, antioxidant defense and cellular senescence (84-86). TP53 was identified as a key hub gene in AF (87), with abnormalities closely associated with disease prognosis (88). PAX2 and ARID3A serve important roles in kidney disease. Studies have shown that PAX2 is reactivated after acute kidney injury, whereas ARID3A, a novel congenital anomalies of the kidney and urinary tract pathogenic gene, is closely associated with immune infiltration in patients with AF (89-91). The present study elucidated the regulatory roles of key OS-association transcription factors in AF and CKD, providing a theoretical basis for precision therapy.

The present study exhibits a number of limitations. The overall sample size for public database analysis and clinical validation was limited and the stability and generalizability of the results need to be further verified in a larger population. The present study integrated transcriptomic data from different tissues, which is more conducive to revealing comorbidity-associated OS characteristics at the systemic level; however, biological heterogeneity among different tissue sources may still affect the interpretation of the results. Despite the candidate genes having received preliminary support in independent datasets and clinical samples, combined predictive models, especially the nomogram, still lack validation in independent external cohorts, and their predictive performance requires further assessment. In addition, immune infiltration analysis, mechanism inference and small molecule compound prediction are primarily based on bioinformatics methods and further functional experiments are still needed for validation. Notably, no stratified analyses were performed by CKD stage or AF subtype. In the present clinical cohort, AF cases were restricted to persistent AF, and the current sample size was not adequately powered for subgroup comparisons by CKD stage; therefore, the effects within different CKD stages or potential AF subtype-specific differences could not be evaluated.

In conclusion, the present study highlights the close molecular association between AF and CKD from the perspective of OS-associated gene expression. A total of 28 pivotal genes were identified, among which ZFP36, BTG2 and NPPB may be key mediators indicating an association between OS and disease progression. These genes represent potential biomarkers and therapeutic targets, offering new insights into the shared pathogenic mechanisms of AF and CKD. Overall, the present findings advance the understanding of how OS contributes to the interplay between cardiac and renal dysfunction, providing a basis for future development of targeted interventions.

Funding Statement

Funding: Funding was received from the Natural Science Foundation of Liaoning Province (grant no. 2024012006-JH4/4800) and the Interdisciplinary Discipline Construction Project of Dalian University (grant no. DLUXK-2025-QN-004).

Availability of data and materials

The data generated in the present study may be requested from the corresponding author.

Authors' contributions

JL and ZX contributed equally to this work. JL and ZX conceived and designed the study and performed the data analysis. HY contributed to the interpretation of the data and critically revised the manuscript for important intellectual content. FS, RW, YB and HH participated in data acquisition and interpretation. XY and YW contributed to the conception and design of the study, supervised the research, interpreted the data and critically revised the manuscript. JL and ZX confirm the authenticity of all the raw data. All authors read and approved the final manuscript.

Ethics approval and consent to participate

The collection of venous blood samples from healthy controls and patients with AF, CKD or concomitant AF and CKD was approved by the Ethics Committee of Affiliated Zhongshan Hospital of Dalian University (Dalian, China; approval no. PJ-KY2025-194-1) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to sample collection.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.