Causal Relationship Between Psoriatic Arthritis and Coronary Heart Disease: A Mendelian Randomization and Bioinformatics Analysis

Abstract

Objective

Psoriatic arthritis (PsA) often coexists with coronary heart disease (CHD). We integrated Mendelian randomization (MR), bioinformatics and reverse network pharmacology to dissect their causality, biomarkers, candidate therapeutics and transcription factors (TFs).

Methods

With PsA and CHD GWAS statistics, five MR methods including inverse variance weighted (IVW), weighted mode, weighted median, simple mode, MR Egger were used to assess causality, followed by sensitivity analyses. We analyzed the differentially expressed genes (DEGs) from the PsA and CHD GEO datasets and identified co-DEGs through intersection. GO, KEGG and GeneMANIA analyses elucidated their functions and pathways. We screened therapeutic candidates through DSigDB and reverse network pharmacology, and mined co-DEGs-related TFs from TRRUST.

Results

The MR analysis revealed genetically predicted PsA confers increased CHD risk (IVW, OR = 1.098, 95% CI 1.016, 1.187, p = 0.018), whereas reverse MR showed no causal effect of CHD on PsA. Our study revealed 71 DEGs from GSE61281 (PsA) and 1116 DEGs from GSE66360 (CHD). The seven intersecting co-DEGs participated in immune recruitment, migration, and inflammatory activation, linked to NF-κB, TLR, and CLR pathways. The drug prediction analysis indicated that prednisone, tretinoin, fumaric acid, adenosine, fenofibrate, and herbal medicines including Salvia miltiorrhiza Bunge, Conioselinum anthriscoides “Chuanxiong”, Achyranthes bidentata Blume, along with herbal ingredients like kaempferol and naringenin, might treat the comorbidities. Notably, these candidates were computational predictions requiring subsequent experimental validation. TF prediction for BCL2A1, CSTA and LY96 highlighted related TFs.

Conclusion

Our study validates the PsA-CHD correlation and identifies shared novel molecular characteristics. It offers genetic evidence supporting a causal effect of PsA on CHD (not definitive clinical causation). The predicted drugs and herbs can become candidate targets for subsequent research. Due to the limitations of the European population and the limited sample size of public transcriptome data, our findings require further external validation.

Introduction

Psoriatic arthritis (PsA) is a chronic inflammatory musculoskeletal disorder associated with psoriasis (PsO), which affects 7% to 40% of patients with PsO.¹ Although 15% of patients suffer from PsA concurrently with PsO, cutaneous psoriasis typically emerges 10 years before the onset of PsA.² PsA imposes a substantial disease burden worldwide. Studies have shown that the global prevalence of PsA is approximately 0.1% to 1.0%,³ whereas the prevalence ranges from 2% to 4% among adults in the United States.⁴ In the age group of 40 to 50 years, the prevalence is nearly comparable between males and females.⁵ PsA is classified as a spondyloarthropathy, with diverse clinical manifestations including axial arthritis, peripheral arthritis, dactylitis, enthesitis, and occasionally nail dystrophy.⁶ PsA is associated with comorbidities such as cardiovascular diseases, uveitis, inflammatory bowel disease, and osteoporosis.^7–9 The optimal management of PsA involves controlling musculoskeletal and skin symptoms, while addressing associated comorbidities to improve the long-term prognosis of patients.

Coronary heart disease (CHD) is a common progressive cardiovascular disorder. Its etiology is related to atherosclerosis or plaque buildup, which leads to gradual narrowing or occlusion of the coronary arteries, thereby causing myocardial ischemia, hypoxia, and cardiomyocyte necrosis.¹⁰ Its pathogenesis involves abnormal lipid metabolism, inflammation, and endothelial dysfunction, triggered by risk factors like smoking, diabetes, dyslipidemia, and hypertension. Its incidence is increasing year by year, making it the most common cause of death globally.¹¹ The main symptoms of CHD include shortness of breath, chest tightness and chest pain. It is closely related to an increased risk of fatal myocardial infarction, arrhythmia, heart failure, stroke or revascularization.¹² Successful treatment of CHD relies on early revascularization combined with comprehensive control of comorbidities to reduce risks and optimize patient prognosis.

Recent studies have shown that patients with PsA are more likely to suffer from premature cardiovascular diseases, and cardiovascular diseases are the most common cause of death among PsA patients.¹³ According to Majed Khraishi et al, PsA is associated with significant cardiovascular comorbidities even in its early stages, with 8.7% of PsA patients being more susceptible to CHD.¹⁴ Therefore, early detection and management of these comorbidities can reduce morbidity, disability and quality of life of patients. Similarly, according to Mishari T Alrubaiaan et al, it is determined that PsA is associated with cardiovascular morbidity. The authors emphasizes that future research should consider these cardiovascular diseases when evaluating PsA.¹⁵ In addition, a study found that in the Czech Republic, patients with PsA have higher cardiovascular risk factors, with CHD occurring in 4.9% of these individuals.¹⁶ Increasing epidemiological evidence indicates that patients with PsA have a significantly increased risk of CHD and major adverse cardiovascular events, independent of conventional cardiovascular risk factors such as hypertension, dyslipidemia, and smoking.^17–19 Meanwhile, some scholars believe that the characteristic of PsA is persistent systemic inflammation and disrupted immune homeostasis. During the progression of PsA, proinflammatory cytokines including IL-17 and TNF-α released into the systemic circulation cause endothelial damage, lipid deposition, and the formation of atherosclerotic plaques, thereby significantly increasing the risk of CHD.^20–25 Beyond these effector cytokines, hyperactivated TLR/NF-κB pathways serve as the shared pathological axis linking PsA to CHD, collectively triggering immune dysregulation and metabolic disorders to aggravate atherosclerotic lesions and vascular dysfunction.²⁶ These factors may be potential causes of PsA comorbidity CHD. While many studies have identified a correlation between PsA and CHD, these findings are mainly based on observational studies and the results are not very robust. To date, there have been no studies using genetics and bioinformatics to explore the shared molecular signatures and potential pathogenetic links between the two. Furthermore, both PsA and CHD impose significant health burdens globally, yet their exact comorbidity pathogenesis remains largely unknown. Therefore, identifying the correlation between PsA and CHD, as well as the risk factors for comorbidity, can help elucidate their pathogenesis and determine effective prevention and treatment strategies.

Mendelian randomization (MR) represents a statistical approach to assess causal links between exposures (particular risk factors) and outcomes (specific phenotypes) by leveraging instrumental variables (IVs), which can reduce common confounding factors in observational studies.²⁷ Unlike forward pharmacology that screens targets from known drugs, reverse network pharmacology predicts therapeutic candidates based on disease-related genes and pathways. This method can efficiently screen multi-target drugs and herbal components targeting the common pathogenic pathways of PsA and CHD, providing comprehensive treatment options for both conditions. While MR can robustly evaluate genetic causality and eliminate observational bias, it cannot reveal the potential therapeutic targets and underlying molecular mechanisms of this comorbidity. To address this limitation, we integrated MR analysis with bioinformatics and reverse network pharmacology. Specifically, bidirectional MR was applied to verify the causality between PsA and CHD; bioinformatics analysis was performed to screen shared pathogenic genes and perturbed signaling pathways; reverse network pharmacology further excavated candidate drugs and herbal ingredients for comorbidity intervention. This analytical framework integrates causal validation, mechanistic exploration and translational drug screening, providing more systematic and deeper biological insights than causal inference alone. It is worth noting that there have been no studies using bidirectional MR to assess the causal association between PsA and CHD to date. Previous cohort studies only reported epidemiological correlations susceptible to confounding factors, and existing bioinformatics studies failed to validate causal relationships. Different from these works, our study first performed bidirectional MR to confirm the causal effect of PsA on CHD, further excavated shared molecular markers and predicted potential therapeutic substances, providing causal evidence and mechanistic clues for PsA-related cardiovascular complications. Therefore, our research findings can deepen understanding of the comorbidity mechanism between PsA and CHD. Discovering shared biomarkers and therapeutic candidates can help perform precise risk stratification for patients with both diseases, optimize comorbidity management strategies, provide theoretical support for translational research targeting this comorbid state, and provide fresh and robust scientific evidence for the comprehensive treatment of both conditions. Meanwhile, the predicted targets and drugs have potential clinical value and can become future research targets.

Materials and Methods

MR Analysis

Data Collection

We adopt the two-sample MR approach to investigate the relationship between PsA and CHD. The summary data of genetic information related to PsA was derived from the publicly available genome-wide association study (GWAS) analyses, with the selected dataset having the identifier ieu-b-5116, comprising 26,351 samples (5,065 cases and 21,286 controls), all of which belong to the European ancestry. Meanwhile, the IEU OpenGWAS database provided CHD data with the identifier ebi-a-GCST000998, a study of 86,995 European descent, including 22,233 cases and 64,762 controls. Table 1 has summarized the two GWAS datasets used in this study. GWAS summary statistics were retrieved 12 December 2025 from legacy OpenGWAS (API v4, old: https://gwas.mrcieu.ac.uk/, new: https://opengwas.io/datasets/), all datasets harmonized to GRCh37.

Table 1.

Details on GWAS Datasets That Were Used for the MR Analysis

Trait	GWAS ID	Case/Control	Sample Size	Population	Year
PsA	ieu-b-5116	5,065/21,286	26,351	European	2022
CHD	ebi-a-GCST000998	22,233/64,762	86,995	European	2011

Selection of Instrumental Variables

As is well known, in the entire MR research process, three assumptions should be satisfied, namely the relevance hypothesis, the independence hypothesis and the exclusivity hypothesis. To ensure that the causality conclusion between PsA and CHD was accurate and effective, the quality control procedures below were adopted by us to filter suitable IVs: (1) The IVs were closely related to the corresponding exposures.²⁸ Single nucleotide polymorphisms (SNPs) linked to PsA were chosen as potential IVs at the significance threshold p < 5×10⁻⁸. (2) To ensure that exposure-related IVs were independent of each other, we used the European sample data as a reference panel (1,000 Genomes Project) to exclude the linkage disequilibrium (LD) between the SNPs to avoid the offset caused by them, the parameters were set to r² < 0.001 and LD distance > 10,000 kb.²⁹ (3) F-statistic was computed to evaluate the IVs and exposure relationship strength, and a higher F-statistic prompted a greater effect of IVs. F-statistic > 10 took into account to be strong IVs and retained, while F-statistic ≤ 10 considered for weak IVs and discarded.³⁰ The instrument selection for reverse analysis was the same as that for forward analysis. Statistical power was calculated via the web-based mRnd calculator (https://shiny.cnsgenomics.com/mRnd/). After LD clumping, 17 independent instrumental SNPs were obtained, collectively explaining 5.58% of the phenotypic variance of PsA (R2 IV = 0.055784). Combined with GWAS sample sizes of PsA and CHD, as well as IVW OR estimate, the statistical power reached 99.98% at α = 0.05. Each instrumental SNP yielded an F-statistic > 10, ruling out weak instrument bias and verifying sufficient statistical power for causal inference.

MR Analysis

The MR analysis in this study was rigorously performed using various methods to improve the validity and robustness of the findings. We adopted five statistical MR models, including inverse variance weighted (IVW), weighted mode, weighted median (WME), simple mode, and MR Egger,³¹ with IVW set as the primary analytical method for the following reasons. Firstly, when all instrumental SNPs met the three core MR assumptions and there was no horizontal pleiotropy, IVW could achieve optimal statistical power and generate unbiased causal estimates. Secondly, the MR-Egger intercept test excluded the significant horizontal pleiotropy in our instrumental SNPs, which met the prerequisite conditions for reliable IVW inference. Moreover, we adopted the random-effects IVW model to obtain conservative and robust effect estimates in the presence of heterogeneity, while the fixed-effects IVW model was applied when heterogeneity was absent.^32,33 As no significant heterogeneity was detected in our study (p > 0.05), fixed-effects IVW was employed as the primary analytical approach. Finally, through sensitivity analyses using the MR-Egger and weighted median methods, we further verified the robustness of the results based on IVW. Therefore, the IVW approach was preferred, and the other four models served as complementary causal evaluation methods applicable under distinct assumptions regarding horizontal pleiotropy. If the results of these supplementary methods are consistent with the estimation results of IVW, the effect estimation robustness can be enhanced. To assess whether there was any bias in the obtained causality, we applied several methods to perform sensitivity analysis. Firstly, the intercept of MR-Egger regression was used to confirm the existence of pleiotropy.³⁴ Secondly, Cochran’s Q test was used to determine the presence of heterogeneity.³⁵ Third, the “leave-one-out” analysis was used to judge whether a single SNP would affect the main causality.³⁶ Throughout the entire MR analysis processes, we made use of the “TwoSampleMR (v0.6.8)” and “gwasglue (v0.0.0.9000)” packages in R v4.4.1.

Bioinformatic Analysis

Processing of Datasets

From the Gene Expression Omnibus (GEO) database, the PsA-related dataset GSE61281 and CHD-related dataset GSE66360 were extracted. The selected datasets both contain samples from Homo sapiens and were selected after rigorous screening. Specifically, the GSE61281 dataset contained 52 whole blood samples, including 20 patients with PsA, 20 patients with cutaneous psoriasis without arthritis (PsC), and 12 healthy individuals. Only data from patients with PsA and healthy individuals were used for subsequent differential expression analysis, while samples from patients with PsC were excluded from our analyses. The GSE66360 dataset included 99 whole blood samples: 49 patients with CHD and 50 healthy controls. GEO datasets were retrieved from NCBI GEO on 25 December 2025 (https://www.ncbi.nlm.nih.gov/geo/).

Screening for Relevant Targets for PsA and CHD

The “limma (v3.60.4)” package was used for differential expression analysis to identify Differential Expression Genes (DEGs) between patients with PsA, PsC, and healthy controls, as well as between patients with CHD and healthy controls. Each dataset was processed separately. Duplicated probes corresponding to the same gene were aggregated by the avereps function of limma. Genes with average expression values equal to zero across all samples were discarded to eliminate biologically silent transcripts. We evaluated the need for log transformation of raw expression data based on expression quantile distributions. Non-log-transformed matrices were subjected to log₂(X + 1) transformation, and intra-dataset normalization was conducted using the normalizeBetweenArrays function from the limma package. Samples within each dataset were divided into disease and control groups. Differential expression analysis was conducted using the limma framework. The false discovery rate (FDR, Benjamini-Hochberg correction) was applied during differential expression analysis to control false positives arising from multiple testing. We set the filtering threshold to |log₂ fold change (FC)| > 0.585 and FDR < 0.05, and visualized PsA-DEGs and CHD-DEGs using volcano plots and heatmaps. The “ggvenn (v0.1.10)” package was used to construct Venn diagrams to extract overlapping co-DEGs between PsA and CHD patients, which provided a molecular basis for previous MR analysis. Box plots depicting the expression levels of candidate co-DEGs were plotted via limma, and the “pROC (v1.19.0.1)” package was utilized to calculate the area under the ROC curve (AUC) for verifying the diagnostic performance of co-DEGs. The “circlize (v0.4.16)” package was applied to create circle plots exhibiting the chromosomal loci of DEGs. All analyses were performed using R v4.4.1.

GO Terms and KEGG Pathway Enrichment Analyses

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed on the co-DEGs obtained. The analyses were conducted using the “GO plot (v1.0.2)” package, “enrichplot (v1.24.4)” package, and “clusterProfiler (v4.12.6)” package in the R v4.4.1 software statistical environment in order to determine the functions and pathways associated with these DEGs. The Benjamini-Hochberg FDR correction was calculated for all enriched terms to control false positives caused by multiple testing. Enrichment was filtered based on raw p < 0.05, while FDR-adjusted P values were retained. GO term analysis is a widely used gene annotation tool, covering three aspects: molecular function (MF), cellular component (CC), and biological process (BP).³⁷ KEGG enrichment analysis is a specialized resource for storing genetic pathway information across different species.³⁸ To visualize the results of GO and KEGG enrichment analyses, circle plots and circos plots were respectively constructed. Subsequently, a co-expression network of the co-DEGs was constructed using GeneMANIA (https://genemania.org/), which was accessed and downloaded on 29 December 2025. The network construction adopted default parameters of the GeneMANIA web server. Given the limited number of intersecting co-DEGs obtained in this study, all overlapping genes were regarded as hub genes involved in the common pathogenic process of PsA and CHD comorbidity. It provides comprehensive information on genetic interactions, physical interactions, pathway co-localization, co-expression networks, as well as predicted and shared protein domain data.³⁹

Drug and TFs Prediction

Using the Drug Signatures Database (DSigDB) (https://dsigdb.tanlab.org/DSigDBv1.0/) and with the help of the “Biocmanager (v1.30.25)” package, potential candidate drugs for treating PsA and CHD comorbidities were screened. All records from DSigDB were mapped to the hub gene set, and compounds targeting one or more hub genes were preserved. Duplicated compounds were removed to generate the final candidate drug list. The interaction dataset was imported into Cytoscape v3.7.2 for visualization of the hub genes-predicted drugs network. Meanwhile, reverse network pharmacology was adopted to screen herbal therapeutic candidates against the hub genes of PsA and CHD comorbidity. Herbal ingredients predicted to interact with hub genes were retrieved from traditional Chinese medicine pharmacology databases. The interaction information among hub genes, herbal active components and corresponding Chinese herbal medicines was imported into Cytoscape v3.7.2 to construct a gene-ingredient-herb network. Molecular docking was conducted using the Cavity-detection guided Blind Docking version 2 (CB-Dock2) (https://cadd.labshare.cn/cb-dock2/index.php) web server to validate candidate. Molecular interaction visualization was performed with PyMOL 2.0 (Schrödinger, USA). Then, we used the Transcriptional Regulatory Relationships Unraveled by Sentence-based Text mining (TRRUST) (https://www.grnpedia.org/trrust/) database to further explore the regulatory mechanisms underlying the co-DEGs.⁴⁰ TRRUST enables the prediction of transcriptional regulatory networks through the identification of target genes, their cognate transcription factors (TFs), and the regulatory interactions between these two classes of molecules. We identified the TFs related to the hub genes and used Cytoscape v3.7.2 software to construct a gene-TF network map. The above web tools, including DSigDB, CB-Dock2 and TRRUST, were run with default parameters and accessed on 31 December 2025. Cytoscape was used merely for network visualization without additional plugins, relying on its default layout mode.

Results

MR Analytic Result

Selection of Instrumental Variables

Following the IVs selection criteria, we confirmed SNPs related to PsA as IVs. Based on the series of quality control processes described above, 17 unique SNPs of PsA were finally screened for this MR analysis. Moreover, the F-statistics of all filtered IVs ranged from 27.06 to 199.54, all satisfying the conventional threshold of F > 10, which reduces the risk of weak instrument bias. Accordingly, the selected IVs are robust, dependable and informative for estimating causal effects. The main information of the IVs mentioned above was presented in Supplementary Table S1.

Causal Effects of PsA on CHD

We utilized the five statistical models mentioned above for MR analysis, while using odds ratio (OR) to indicate the degree and trend of association between PsA and CHD. Genetically predicted PsA was related to an increased risk of CHD, according to the IVW technique [OR = 1.098, 95% confidence interval (CI): (1.016, 1.187), p = 0.018]. All five methods showed positive ORs and similarly directional correlations were detected utilizing another method WME [OR = 1.113, 95% CI: (1.009, 1.228), p = 0.032], despite the results of the remaining methods, including the MR Egger [OR = 1.101, 95% CI: (0.862, 1.407), p = 0.497], simple mode [OR = 1.124, 95% CI: (0.992, 1.275), p = 0.140] and weighted mode [OR = 1.115, 95% CI: (0.993, 1.252), p = 0.139], were not statistically significant. We conducted a series of comprehensive sensitivity analyses to ensure our obtained results robustness and reliability. The Cochran’s Q test indicated that there was no heterogeneity in the selected IVs (p > 0.05), and the funnel plot displayed general symmetry, which also indicated that there was almost no heterogeneity. Notably, the MR-Egger regression intercept term results did not deviate from 0, suggesting that the IVs filtered by us did not exhibit horizontal pleiotropy (p > 0.05). Meanwhile, after the Leave-one-out analysis, we found that no single SNP had a strong impact on the total estimations, indicating that all SNPs made significant contributions to the causality establishment. Detailed information was presented in Figure 1 and Table 2, and Supplementary Table S2.

Figure 1.

Plots of Mendelian randomization (MR) estimates illustrating the causal effects of PsA on CHD. (A) The causal effect of PsA on CHD (nSNP, number of SNPs employed in this study; CI, confidence interval; OR, odds ratio, bold text indicates statistically significant findings); (B) MR effect size of SNPs associated with PsA and their associated risk for CHD (black dots and lines show the point estimate and 95% CI of the causal effect for each SNP. Bottom red lines indicate pooled causal estimates (95% CI) from MR-Egger and IVW analyses, respectively); (C) MR test scatterplot of five methods (the y-axis and x-axis, respectively, represent the impact of IVs on the outcome and exposure. The slope illustrates the effect of exposure on the outcome); (D) MR leave-one-out analysis of PsA on CHD (the black horizontal line represents the estimates after removing each IV, while the red horizontal line represents the overall estimate of IVW); (E) Funnel plot of individual SNP analyses (symmetry indicates the absence of heterogeneity in IVs).

Table 2.

Sensitivity Analysis of the Relationship Between PsA and CHD

Exposure	Outcome	Heterogeneity Test				Pleiotropy Test
		IVW		MR-Egger		MR-Egger Intercept
		Q	Q_pval	Q	Q_pval	Intercept	pvalue
PsA	CHD	1.302	0.861	1.301	0.729	−0.0005	0.985

Causal Effects of CHD on PsA

Reverse analysis was conducted to further evaluate the causality between CHD and PsA. The IVW technique demonstrated that there was no evidence for the causal effect of CHD on PsA [OR = 0.980, 95% CI: (0.877, 1.096), p = 0.728]. Moreover, no statistical significance was observed in the results from the remaining MR methods. Sensitivity analyses indicated that after using the MR-Egger regression intercept term to analyze the selected IVs, no horizontal pleiotropy was found (p > 0.05). Meanwhile, Cochran’s Q test suggested no significant heterogeneity in these IVs (p > 0.05), and the funnel plot also showed overall symmetry, indicating almost no heterogeneity. Meanwhile, the result from Leave-one-out analysis displayed that no single SNP has a large impact on the total estimations, which were shown in Supplementary Figure S1, Supplementary Tables S3 and S4.

Bioinformatic Analysis

Screening for Relevant Targets for PsA and CHD

Analyzing the PsA-related dataset GSE61281, a total of 71 DEGs were obtained, of which 33 genes were upregulated and 38 genes were downregulated. Analyzing the CHD-related dataset GSE66360, a total of 1116 DEGs were identified, of which 682 genes were upregulated and 434 genes were downregulated. By using Venn diagrams for intersection analysis, seven co-DEGs shared by PsA and CHD were successfully identified, namely CLEC4D, CSTA, BCL2A1, SAMSN1, LY96, MS4A4A, LOC100131541. Of these, 6 genes were upregulated and 1 gene (LOC100131541) was downregulated. Moreover, box plots were used to analyze the seven co-DEGs to display their expression levels. The AUC values of the seven co-DEGs all exceeded 0.7 with favorable 95% CIs, further supporting the reliability of these genes. Through circle plots, we presented the chromosomal loci of these genes. The above relevant information was shown in Figures 2 and 3, Supplementary Tables S5 and S6. These findings support the close association between PsA and CHD at the genetic level, and suggest that PsA may cause CHD.

Figure 2.

DEGs screening. (A) Volcano plot of DEGs for PsA; (B) Heatmap of DEGs for PsA; (C) Volcano plot of DEGs for CHD; (D) Heatmap of DEGs for CHD; (E) Venn graphic of upregulated DEGs that PsA and CHD have in common; (F) Venn graphic of downregulated DEGs that PsA and CHD have in common.

Figure 3.

Co-DEGs characteristics. (A) Box plot of co-DEGs for PsA (** p < 0.01, *** p < 0.001); (B) Box plot of co-DEGs for CHD (*** p < 0.001); (C) AUC of co-DEGs for PsA; (D) AUC of co-DEGs for CHD; (E) Chromosomal loci of co-DEGs for PsA; (F) Chromosomal loci of co-DEGs for CHD.

Enrichment Analysis Results

We conducted enrichment analysis on seven co-DEGs to gain a deeper understanding of their functional significance. GO analysis identified 7 CC, 12 MF, and 33 BP terms. Notably, the results revealed that these hub genes are primarily involved in pathogen recognition of bacteria and fungi, immune cell activation and regulation, and innate immune response (BP); localized in immune cell granules, plasma membrane rafts, and skin barrier-related structures (CC); and exert pattern recognition, molecular binding, and signaling regulatory activities (MF). Enrichment p-values were negative log₁₀-transformed and visualized in a circle plot. In this plot, the outermost circle shows GO IDs (BP, CC, MF), the second circle represents the number of genes on each GO term, the third circle shows the number of enriched co-DEGs, and the innermost circle denotes the proportion of co-DEGs relative to the total genes. Darker red coloring of the second circle indicates more significant enrichment. Additionally, KEGG pathway analysis identified 8 enriched pathways, including Apoptosis, C-type lectin receptor signaling pathway, Toll-like receptor (TLR) signaling pathway, and NF-κB signaling pathway. Corresponding enrichment p-values were similarly negative log₁₀-transformed and visualized in a circos plot. The left and right semicircles represent co-DEGs and pathways, respectively, with different colors indicating distinct pathways. Connecting lines indicate gene-pathway enrichment associations. Asterisks mark enrichment significance: ***p < 0.001, **p < 0.01, *p < 0.05. To further investigate the functional interactions of these co-DEGs, we utilized GeneMANIA to construct a gene interaction network. The network displayed co-localization (2.6%) and co-expression (97.4%), as shown in Figure 4 and Supplementary Figure S2.

Figure 4.

Enrichment analyses. (A) GO enrichment analysis circle plot of co-DEGs; (B) KEGG pathway analysis circos plot of co-DEGs (*p < 0.05, *** p < 0.001); (C) Co-expression network of co-DEGs constructed using GeneMANIA.

Drug and TFs Prediction Results

When predicting potential medications for treating PsA and CHD comorbidities, a total of 80 drugs were identified by searching co-DEGs using the DSigDB database, mainly including prednisone, tretinoin, fumaric acid, adenosine, fenofibrate and so on. We employed a bubble map to visually present the top 30 drugs with the most significant enrichment. In this map, the color and size of bubbles reflect the significance of gene enrichment. Meanwhile, we constructed a drug regulatory network utilizing Cytoscape software, which the red ellipses represent hub genes and the green squares represent therapeutic drugs. In addition, we also employed reverse network pharmacology to predict the Chinese herbal medicines and ingredients for treating their comorbidity. A total of 135 Chinese herbal medicines were screened out, mainly including Conioselinum anthriscoides “Chuanxiong”, Achyranthes bidentata Blume, Salvia miltiorrhiza Bunge and so on. 4 ingredients were identified, namely naringenin, Deoxycholic Acid, kaempferol, and eriodictyol. Similarly, a gene-ingredients-herb network was constructed using Cytoscape software, which the red ellipses represent hub genes, the yellow diamonds represent ingredients, and the blue rectangles represent Chinese herbal medicines. Subsequently, CB-Dock2 was applied for molecular docking to verify the reliability of the above results at the molecular level. Generally, a binding affinity below −5.0 kcal/mol indicates favorable binding capacity of the ligand to the target, and a more negative Vina score corresponds to a more stable Protein-ligand complex. We presented the docking results of kaempferol and BCL2A1 that we were most concerned about, with a Vina score of −7.1 kcal/mol. The docking scores are shown in Supplementary Table S7. To investigate the regulatory relationship between hub genes and TFs, we utilized the TRRUST database, which is a comprehensive resource for the interactions between TFs and target genes. By using this database, we identified potential TFs that exhibit regulatory potential on BCL2A1, CSTA, LY96, and constructed a gene-TF network map using Cytoscape software. In this map, the brown diamonds represent hub genes, and orange octagons represent TFs. The relevant information was shown in Figure 5, Supplementary Tables S8 and S9.

Figure 5.

Drug and TFs prediction. (A) Bubble map of potential medications for treating comorbidities; (B) Network diagram of hub genes and predicted drugs; (C) Network diagram of gene-ingredient-herb; (D) Molecular docking of kaempferol with BCL2A1; (E) Regulatory network of hub genes and TFs.

Discussion

It is well established that PsA onset is tightly linked to genetic susceptibility, environmental stimuli, and dysregulated immune responses.⁴¹ Genetic susceptibility involves HLA-B27 and multiple non-HLA loci related to both NF-κB and IL-17/IL-23 signaling.⁴² Environmental stimuli including gut dysbiosis, obesity, and infection trigger aberrant immune activation.⁴³ These two factors jointly induce immune dysregulation, leading to the heterogeneous clinical phenotypes of PsA affecting the joints, entheses, skin, and gut.

CHD is predominantly driven by coronary atherosclerosis, which is associated with genetic susceptibility, immune-mediated inflammatory responses, age, obesity, hypertension, diabetes, and dyslipidemia.⁴⁴ Dysregulated activation of T/B lymphocytes and autoantibodies (eg, anti-cardiolipin antibodies) aggravate oxidized low-density lipoprotein-mediated endothelial dysfunction. Aberrant immune cells (innate and adaptive), the NLRP3 inflammasome, and Treg/Th17 imbalance drive plaque progression.^45,46 Such cascading immune dysfunctions and inflammatory responses further exacerbate CHD initiation and progression.

Accumulating evidence suggests a close association between PsA and CHD mediated by immune-inflammatory pathways, laying a solid theoretical foundation for the present study. However, confounders and ethical concerns prevent traditional observational epidemiologic approaches from drawing valid causal inferences. MR analysis can compensate for these shortcomings, and bioinformatic validation can further strengthen the reliability of the research results.⁴⁷ This study used MR analysis to investigate the causality between PsA and CHD, with its results subsequently validated by bioinformatic analyses. To our knowledge, this is the first study to elucidate the causality between PsA and CHD using public genetic datasets and to explore their comorbidity via bioinformatic approaches.

In this study, two-sample MR analysis revealed a causal association between genetically predicted PsA and increased CHD risk using the IVW method (OR = 1.098, 95% CI 1.016, 1.187, p = 0.018). This is basically consistent with the results of the previous several clinical studies. Notably, this modest OR suggests a weak causal effect of PsA on CHD. Chronic inflammation, immune dysregulation and genetic susceptibility are considered to be the common pathogenic mechanisms of these two diseases. Although the causal effect estimate was modest, persistent chronic inflammation related to PsA may exert cumulative adverse impacts on the vasculature, facilitating the progression of atherosclerosis and increasing CHD susceptibility. Our MR findings further deepen the understanding of PsA-CHD comorbidity by disclosing the genetic basis of this association, providing new insights into how PsA genetic susceptibility contributes to CHD development. In the MR analysis, we did not find that the susceptibility to PsA would be altered by genetic predisposition to CHD. This result is consistent with clinical observational studies. Current evidence does not support a causal role of CHD in the onset of PsA.

With advances in bioinformatics, such analytical tools have played a significant role in elucidating disease mechanisms and identifying potential therapeutic targets. In the present study, multiple bioinformatic approaches, including Venn diagrams, enrichment analysis, and reverse network pharmacology, were applied to provide insights into the potential connection between PsA and CHD. Analysis of microarray datasets from patients with PsA and CHD identified seven co-DEGs. These hub genes may act as key nodes linking the two diseases and modulate chronic inflammation and immune dysfunction. CLEC4D is a C-type lectin receptor gene, which has the function of mediating innate immune responses and inflammatory cell recruitment. Its overexpression can promote proinflammatory cytokine release, accelerating joint inflammation and atherosclerosis.^48–50 CSTA encodes cystatin A, which inhibits cysteine proteases. It enhances inflammatory cell infiltration, leading to skin-joint lesions and vascular endothelial damage.^51–53 BCL2A1 is an anti-apoptotic gene, which protects inflammatory cells from apoptosis. It can prolong inflammation and promote foam cell formation in atherosclerotic plaques.^54–56 SAMSN1 regulates the activation and proliferation of immune cells. Its abnormal expression can impair immune homeostasis and participate in the pathological processes of synovitis and atherosclerosis.⁵⁷ LY96 is a co-receptor of TLR4, which can amplify inflammatory signaling.⁵⁸ Upregulation of its expression can exacerbate immune-mediated tissue damage and endothelial dysfunction.^59,60 MS4A4A can participate in immune cell chemotaxis and activation. Its dysregulation can promote inflammatory infiltration in joints and coronary vessels.^61,62 LOC100131541 is a long non-coding RNA-related gene that regulates gene expression involved in inflammation and lipid metabolism,⁶³ indirectly facilitating the progression of both PsA and CHD. These co-DEGs reflect the shared genetic background of PsA and CHD, and their dysregulated expression may be a key driver of this comorbidity, consistent with the observation that PsA-related systemic inflammation accelerates atherosclerosis.

GO enrichment of PsA and CHD revealed significant correlations with a series of biological processes, including immune cell recruitment, migration, and initiation of inflammatory responses. These findings are consistent with other studies, such as the GO analyses results suggesting that lipopolysaccharide (LPS)-mediated signaling regulation (BP entries), TLR4 signaling (BP entries), pattern recognition receptor activity (MF entries), and specific granule membrane (CC entries) are all related to the comorbidity of PsA and CHD. TLR4-LPS signaling promotes proinflammatory cascades in PsA synovitis and CHD atherosclerosis.^64–66 Pattern recognition receptors amplify systemic inflammation, linking PsA immune dysregulation to endothelial dysfunction and plaque progression.^67,68 Neutrophil degranulation and granule membrane components sustain inflammatory responses in both diseases.^69,70 These results support a common immune-inflammatory pathogenesis for both conditions. Furthermore, KEGG enrichment revealed several shared pathways underlying the comorbidity of PsA and CHD. Besides being related to the TLR signaling pathway, it was also most closely associated with the NF-κB signaling pathway, and the C-type lectin receptor signaling pathway. These findings are also consistent with other studies. The NF-κB pathway serves as the central hub linking chronic inflammation in these two diseases. Persistent NF-κB activation drives proinflammatory cytokine secretion in PsA synovitis and promotes endothelial dysfunction, foam cell formation, and atherosclerotic plaque instability in CHD.^71–73 The TLR and C-type lectin receptor signaling initiate innate immune responses, exacerbating systemic inflammation.^74–76 Overactivation of these pathways contributes to immune dysregulation in PsA and accelerates the development of atherogenesis by amplifying inflammatory cascades in the vascular walls. These results similarly suggest that aberrant immune-inflammatory signaling is an important mechanistic link between PsA and CHD. Synthesizing these bioinformatics analyses, we believe that there is a complex interaction between PsA and CHD, including multiple mechanisms and biological pathways, which further validates the results of the MR analyses in this study.

Currently, distinct therapeutic options exist for PsA and CHD, respectively. In this research, we identified candidate drugs with potential dual-action against PsA-CHD comorbidity. The candidate compounds identified in this study, including prednisone, tretinoin, fumaric acid, adenosine, and fenofibrate, exhibit promising therapeutic potential by targeting shared core pathways underlying both disorders. Prednisone is a classic glucocorticoid that can effectively alleviate systemic inflammatory response in patients with PsA by inhibiting proinflammatory cytokines such as TNF-α and IL-6. Such reduction in chronic inflammatory burden further lowers cardiovascular risk and stabilizes atherosclerotic plaques.^77,78 Tretinoin can modulate keratinocyte differentiation and alleviate PsA inflammation. It also regulates lipid metabolism and reduces vascular endothelial oxidative stress, conferring cardiovascular protection.^79–81 Fumaric acid plays anti-inflammatory effects by inhibiting NF-κB signaling and activating the Nrf2 pathway, relieving cutaneous and articular inflammation in PsA. It also inhibits vascular inflammation and improves endothelial function, reducing atherosclerosis risk.^82–84 Adenosine exerts immunosuppressive effects via the A2A receptor to alleviate immune activation in PsA. It also dilates coronary vessels and improves myocardial perfusion, yielding cardioprotective and anti-inflammatory effects.^84–86 Fenofibrate exerts anti-inflammatory and lipid-regulating effects. It inhibits the IL-17A signaling pathway, restores Treg/Th17 balance, and alleviates inflammation in PsA.⁸⁷ It also ameliorates dyslipidemia, lowers triglyceride levels, and retards atherosclerosis progression. Because these drugs exert anti-inflammatory, immunomodulatory, and plaque-preventing effects, they may enable effective treatment for patients with PsA complicated by CHD.

Chinese herbal medicine offers a relatively safer alternative to chemical drugs and exerts therapeutic effects via multiple targets and pathways. Our study identified multiple herbs, including Conioselinum anthriscoides “Chuanxiong”, Achyranthes bidentata Blume and Salvia miltiorrhiza Bunge, capable of treating PsA-CHD comorbidity, consistent with the traditional Chinese-medicine principle of “treating different diseases with the same therapy”. Of these, the four herbal-derived components hold potential therapeutic value for both conditions. In aggregate, these four phytochemicals, namely naringenin, deoxycholic acid, kaempferol and eriodictyol, exert dual protective effects against joint inflammation in PsA and atherosclerosis underlying CHD by regulating autophagy-related metabolic pathways,^88,89 bile-acid receptor signaling,^90,91 vascular proatherogenic signaling,^92,93 and NF-κB/AKT-mediated inflammatory cascades.^94,95 Further research in this area may provide a promising strategy for preventing and managing cardiovascular complications in patients with PsA.

This study has both its advantages and limitations. The present study is the first attempt to synergistically integrate GWAS and GEO data to elucidate the causal relationship between PsA and CHD through MR analysis and bioinformatics analysis. To ensure the robustness and timeliness of the research results, we used the recent large-scale GWAS data in our study. Additionally, the MR analysis in this research followed a strict quality control process based on three fundamental assumptions and sensitivity analysis. Moreover, by utilizing MR design, this study could effectively reduce the distractions of reverse causality and confounding factors on the results. The limitations of this study are as follows. Firstly, the research mainly focused on the genetic association between PsA and CHD, but some non-genetic factors, such as lifestyle and environmental factors, might also have certain impacts. Secondly, all genetic datasets analyzed in the present study were derived exclusively from individuals of European ancestry. Whether our causal conclusions and molecular mechanisms are applicable to other ethnic populations remains unclear, which may limit the generalizability of our findings. Further validation based on multi-ethnic cohorts is needed in future research. Thirdly, the lack of basic demographic information (eg, age and sex) of the original research subjects prevented further subgroup analyses, and whether this had an impact on the results requires more investigation. Additionally, despite multiple sensitivity analyses to assess horizontal pleiotropy, residual pleiotropic effects cannot be fully excluded, and shared inflammatory pathways as well as other unmeasured biological confounders may contribute to the observed molecular associations. Fourthly, all transcriptomic analyses relied on public datasets, and independent transcriptomic data or external clinical cohorts were unavailable for external validation. Notably, transcriptomic data were derived from peripheral whole-blood samples, which may not completely mirror tissue-specific pathological changes within joint synovium in PsA or coronary artery lesions in CHD. Fifthly, only seven co-DEGs were identified, which may weaken the robustness of downstream enrichment and network analyses. Importantly, it is critical to distinguish that bidirectional MR provides evidence for a causal association between PsA and CHD at the disease level, while transcriptomic analyses only uncover correlational molecular signatures. Despite the biological plausibility of the detected pathways, causality linking these molecular signatures to disease progression cannot be inferred merely from transcriptomic associations. Although the identified pathways are biologically plausible, causality linking these molecular signatures to disease progression cannot be inferred merely from transcriptomic associations. Sixthly, the inherent limitations of the statistical methods and bioinformatics tools used may affect the results. Seventhly, the biomarkers and TFs screened via bioinformatics prediction lack cellular, animal or clinical experimental verification. Collectively, the conclusions drawn from this study should be interpreted with caution before being verified by functional experiments and clinical trials. Finally, reverse network pharmacology serves only as a hypothesis-generating approach and cannot be considered direct evidence of therapeutic efficacy. Given the heterogeneity of herbal preparations, variations in raw material sources, dosage forms and administration regimens, these therapeutic candidates cannot be directly applied in clinical practice without sufficient verification. Subsequent research should systematically explore their active components, mechanisms of action and safety profiles. We advocate that all translational studies involving herbal therapeutics should be conducted following standardized protocols in a responsible fashion. Therefore, additional clinical studies are still imperative in order to validate our findings; subsequent studies should endeavor to conduct further subgroup analyses, and if conditions permit, high-quality randomized controlled trials can be considered to obtain more dependable conclusions.

In conclusion, by conducting MR and bioinformatics analyses, this study explores potential molecular signatures associated with the causal relationships between PsA and CHD, and provided comprehensive screening data for their association. Our research findings suggested that there was a causal relationship between PsA and CHD. Furthermore, key hub genes, biological processes, pathways, therapeutic drugs, and related TFs involved in the comorbidity of PsA and CHD were identified. These results deepen our understanding of the pathogenesis of these diseases and provide potential directions for subsequent research on immunotherapy and targeted interventions. Notably, all molecular signatures and therapeutic candidates identified via bioinformatic methods, including medicines and herbal ingredients, arise solely from computational predictions. These results remain hypothesis-generating and cannot be translated into clinical practice without experimental and prospective clinical validation. Moreover, restricted to European ancestry GWAS data, our results cannot be generalized to other ethnic populations, and MR still has inherent observational limitations. Future research should focus on conducting comprehensive mechanistic research and validating these findings in larger and more diverse populations.

Funding Statement

The Natural Science Foundation of Hubei Province (NO. 2023AFD136) and the Postdoctoral Scientific Research Foundation, General Hospital of Central Theater Command (NO. 20230919KY42) funded this study. This project is supported by the Shizhen Talent Program of Hubei Province for Scientific Research (Grant No: Hubei Health Document [2024] No. 256).

Data Sharing Statement

All data are available in the main text and Supplementary Materials. Additional data related to this paper may be requested from the corresponding author Xietian Yin upon reasonable request.

Ethical Approval

As the data involved in this MR study all came from publicly available GWAS summary statistics, it qualifies for exemption from formal approval according to Chinese national legislation. Specifically, Article 32 (Items 1 and 2) of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects (National Health Commission of China, effective February 18, 2023) stipulates that ethical review may be exempted for research involving human information/data or biological samples under the following conditions, provided that the research: (i) It utilizes legally accessible public data without individual identifiers; (ii) The analytical methodology poses no risk to personal rights or public interests.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors state that in this study there is no interest conflict.