Differential gene expression analysis reveals common biomarkers for systemic lupus erythematosus and atrial fibrillation

doi:10.21203/rs.3.rs-3589630/v1

Download PDF

Research Article

Differential gene expression analysis reveals common biomarkers for systemic lupus erythematosus and atrial fibrillation

https://doi.org/10.21203/rs.3.rs-3589630/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Atrial fibrillation (AF) is a significant outcome of systemic lupus erythematosus (SLE), increasing the chances of experiencing blood clotting events and unforeseen mortality. As the underlying mechanism of SLE companied with AF is still unknown, this study sought to uncover potential biomarkers that could be of significant value for individuals dealing with SLE and AF, employing thorough bioinformatics research as the primary approach.

Methods

The NCBI Gene Expression Omnibus database (GEO) was employed to retrieve a collection of five microarray datasets (GSE50772, GSE41177, GSE79768, GSE81622, and GSE2240). By employing the online analytical tool GEO2R, we conducted an analysis of GSE50772 to pinpoint genes that exhibited differential expression. Significant module genes were discovered by WGCNA (weighted gene co-expression network analysis). To identify DEGs in AF, the 'Limma' package was utilized. Function of the common DEGs was found by functional enrichment analysis. The candidate biomarkers were discovered by applying a machine learning technique. The investigation involved the utilization of Single Sample Gene Set Enrichment Analysis (ssGSEA) scores to perform functional enrichment analysis on the identified candidate biomarkers. To predict the risk of AF in individuals with SLE, a nomogram and a ROC curve were created. The analysis focused on examining the presence of immune cells infiltrating the training datasets of SLE and AF, while also conducting a consensus cluster analysis specifically for SLE.

Results

29 common DEGs were identified between SLE and AF. The identification and utilization of five potential biomarkers-ANKRD36B, SLC4A4, ANKRD12, MTUS1 and DSC1-led to the creation of a nomogram with area under the receiver operating characteristic curve 0.900-0.981 across all datasets. The dysregulated immune cell infiltration was associated with the biomarkers. Based on the consensus clustering analysis, it was concluded that three subtypes were the most suitable in terms of quantity. The biomarkers exhibited different expression patterns among the subtypes. Regarding immunological infiltration, each subtype possessed unique traits.

Conclusion

By employing various bioinformatics research approaches and machine learning techniques, our study identified five candidate biomarkers (ANKRD36B, SLC4A4, ANKRD12, MTUS1, DSC1). Additionally, a nomogram capable of predicting the likelihood of both SLE and AF was developed. The results of our study provide a foundation for future investigations on potential important genes for AF in individuals with SLE. Moreover, it was discovered that AF and SLE exhibited abnormal compositions of immune cells.

System lupus erythematosus

Atrial fibrillation

Bioinformatics analysis

Biomarkers

Immune infiltration

With its association to the production of autoantibodies, systemic lupus erythematosus (SLE) stands as a chronic autoimmune condition, inducing enduring inflammation and affecting individuals across the globe with a prevalence rate that varies from 0.02–0.15% [1, 2]. In patients with SLE, their immune system produces autoantibodies and triggers widespread inflammation throughout the body, which often leads to a significant risk of cardiovascular complications [3, 4]. In accordance with recent findings, the pattern of distribution of SLE fatality is bimodal, with cardiovascular disease (CVD) contributing an extra peak. A significant proportion of lupus patients, exceeding 15%, exhibit disturbances in their heart's rhythm, commonly alongside other heart-related complications or while their lupus is actively affecting them [5].

Out of all the other abnormalities in heart rhythm, atrial fibrillation (AF) has been the most extensively researched heart arrhythmia in the last century, providing valuable knowledge [6]. The incidence rises with aging and is the most prevalent arrhythmia in humans [7]. AF patients suffer from a far greater likelihood of embolic consequences. Stroke is a major danger that arises from AF. Furthermore, the consequences of AF-related stroke are more severe. The prevalence of AF is elevated among patients diagnosed with SLE in comparison to the general population, mainly due to the increased presence of risk factors such as hypertension, obesity, CVD, heart failure, valvular heart disease, and chronic kidney disease [8–10]. Systemic inflammation has been linked to the emergence of AF as a contributing element in its development [11–13]. SLE-related AF patients have the potential to severely shorten life expectancy and result in irreparable organ damage.

Extensive studies have predominantly focused on exploring the effects of conditions such as ischemic heart disease and strokes in populations of patients with SLE when it comes to researching CVD. That being said, the available data is quite limited in terms of providing a comprehensive understanding of the prevalence of AF in individuals with SLE when compared to the broader population [14, 15]. Consequently, delving into the pathogenesis of SLE and AF, discovering predictive biomarkers for early disease detection, and implementing timely intervention therapies on these genes are crucial. This comprehensive approach seeks to minimize the risk of thrombosis and ultimately improve patients' long-term outcomes. In the past decade, remarkable strides have been made in scientific research, thanks to the remarkable advancements in bioinformatics and machine-learning techniques. These cutting-edge tools have allowed scientists to delve into the very foundations of natural processes and identify potential biomarkers [16–19]. By harnessing the power of bioinformatics and machine learning, this study successfully identified key biomarkers for AF in individuals diagnosed with SLE.

2.1 Data acquisition

Figure 1 displays the flowchart of this study. Five microarray datasets (GSE50772, GSE41177, GSE79768, GSE81622, and GSE2240) were sought after in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database. The search was conducted using the relevant search terms 'systemic lupus erythematosus' or 'atrial fibrillation' [20]. Using the 'GEO query' R package, data from each microarray were accessed from GEO. Within the dataset GSE50772 [21] , our study involved examining the gene expression patterns of peripheral blood mononuclear cells (PBMCs) obtained from 61 patients diagnosed with SLE, in addition to 20 healthy individuals serving as controls. The dataset GSE41177 [22] included tissue samples from 3 individuals with normal heart rhythm and 16 patients experiencing AF, specifically focusing on the left atrial-pulmonary vein junction and left atrial appendage. The dataset GSE79768 [23] contained samples of atrial tissue from 6 patients with normal heart rhythm and 7 patients with AF. The dataset GSE81622 [24] included data on the gene expression of PBMCs from 30 individuals diagnosed with SLE. This dataset consisted of 25 individuals without any health issues serving as controls, along with 15 SLE patients who had lupus nephritis (LN) and 15 SLE patients without LN. The gene expression of right atrial appendage and left ventricular myocardium tissue samples from 10 AF patients and 20 healthy donor controls were extracted from the dataset GSE2240 [25]. Among the five datasets, the external validation datasets were the GSE81622 and GSE2240. The information in Table 1 outlines the microarray platform employed, the composition of the sample groups, and the respective sizes of the datasets.

Table 1 \| Fundamental details regarding the GEO datasets utilized in the research
GSE series	Sample size		Tissue types	Platform	Group
GSE series	Control	SLE	Tissue types	Platform	Group
GSE50772	20	61	PBMC	GPL570	Training dataset
GSE81622	25	30	PBMC	GPL10558	Validation dataset
	Control	AF
GSE41177	3	16	LAA+LAPV	GPL570	Training dataset
GSE79768	12	14	LAA+RAA	GPL570	Training dataset
GSE2240	20	10	RAA+left ventricular myocardium	GPL96	Validation dataset

SLE, system lupus erythematosus; AF, atrial fibrillation; PBMC, Peripheral blood mononuclear cell; LAA, left atrial appendage; LAPV, left atrial-pulmonary vein junction; RAA, right atrial appendage.

2.2 Data pre-processing

By undergoing data processing, the raw data matrix undergoes a significant refinement, resulting in a notable enhancement in the quality of information [26]. Missing data elimination, filtration, transformations and sample-based normalization are typical preprocessing techniques [27, 28]. Optimal utilization of suitable tools, methodologies, and datasets can markedly elevate the effectiveness, accuracy, and resilience of analytical processes [29, 30]. The collected datasets were intended for analysis using R software (Version 4.2.3; https://www.r-project.org/). The expression value of the first gene was chosen as the gene expression value when a probe was found to be related to different genes. Through the leverage of platform annotation profiles, the probes are converted into gene symbols, thereby eliminating any probes that do not correspond with gene symbols. If multiple probes match the same gene, the 'unique' function is used to remove duplicates and only one expression value is retained. All datasets have been Log2 transformed. Normalization of the arrays was carried out using the 'Limma' R package's normalizeBetweenArrays function. To consolidate the datasets GSE41177 and GSE79768, the 'merge' function was employed, alongside the R package 'sva' which helped alleviate the heterogeneity caused by distinct batches.

2.3 Detection of differentially expressed genes (DEGs)

In the integrated dataset (GSE41177 and GSE79768), the determination of DEGs between AF and the control was employed using the 'Limma' package [31]. The criteria for determining statistical significance were a |log fold change (FC)| of 0.6 and an adjusted P value of 0.05. The 'graphics' package in R was utilized to construct the volcano plot, a visual representation showcasing the DEGs identified from the AF training dataset.

2.4 Detection of important module genes using weighted gene co-expression network analysis (WGCNA)

We successfully generated a co-expression network utilizing the 'WGCNA' R package, primarily centered around the DEGs in the SLE training dataset [32]. To examine the genes closely linked to SLE when compared to healthy controls, we employed the 'goodSamplesGenes' function provided by WGCNA. This function helped us filter the DEG expression matrix in dataset GSE50772, enabling us to pinpoint the module genes that displayed significant associations with SLE. This process resulted in the creation of a scale-free co-expression network. Subsequently, any unusual samples were excluded, and modules were determined by employing hierarchical clustering and dynamic tree-cutting methods. In order to establish the adjacency, we applied the select-Soft-Threshold method and relied on the soft thresholding exponent derived from the co-expression similarity, moving forward with our analysis. Following that, the adjacency underwent a conversion process to establish a matrix of topological overlap (TOM), which was then utilized to calculate the gene ratio along with its corresponding dissimilarity (1-TOM). To organize genes based on their shared characteristics, we implemented a method called average linkage hierarchical clustering. This technique relied on TOM-based dissimilarity measures and aimed to construct a gene dendrogram. In order to maintain the validity of the gene groups, it was made certain that no group contained fewer than 30 genes. The eigengene network was simultaneously showcased.

2.5 Identification of common DEGs

By intersecting the significant module genes associated with SLE and the DEGs from the AF training dataset, we successfully identified the shared DEGs in both SLE and AF. The visualization of these common DEGs was performed using the R package 'venn'.

2.6 Enrichment analysis of the common DEGs

Renowned as a dependable reference, the KEGG, or the Kyoto Encyclopedia of Genes and Genomes, has gained widespread recognition for its extensive coverage and comprehensive analysis of gene functionalities [33]. By scrutinizing Gene Ontology (GO), researchers have categorized it into three key elements: biological process (BP), cellular component (CC), and molecular function (MF) [34]. Functional enrichment analysis was carried out by performing the 'ClusterProfiler' R package [35]. Using the R package 'ggplot2', the top 10 GO terms in each category were displayed, adhering to the screening criteria of a false discovery rate of 0.05 and an adjusted P value of 0.05. To anticipate gene connections and analyze gene function, GeneMANIA (http://www.genemania.org) [36] established a co-expression network for these common DEGs between SLE and AF.

2.7 Machine learning algorithms

To identify potential biomarkers, a machine learning algorithm known as Support Vector Machine-Recursive Feature Elimination (SVM-RFE) was utilized. Reports indicate that the SVM-RFE algorithm is suitable for datasets with limited samples as it removes unnecessary features and retains only those that impact the result [37].

2.8 Validation and Single-sample gene-set enrichment analysis (ssGSEA) of potential biomarkers

The evaluation of DEGs in each dataset was conducted using the Wilcoxon rank-sum test, which was chosen based on the SVM-RFE analysis carried out in the earlier stage. To illustrate the findings, the R packages 'ggplot2' and 'ggpubr' were utilized, allowing for clear and informative visualizations. Through the utilization of ssGSEA, we managed to establish a strong correlation between potential biomarkers and hallmark gene sets. At the beginning, we acquired the Hallmark gene sets from MSigDB, a comprehensive database (https //www.gsea-msigdb.org/gsea/msigdb/human/genesets.jsp?collection=H), providing comprehensive insights and illustrations of 50 distinct and well-defined biological states or processes. Later on, the potential biomarkers were examined in relation to the hallmark gene sets using ssGSEA, a technique performed with the aid of the R package 'GSVA' [38].

2.9 Assessment of diagnostic effectiveness of biomarkers

To estimate the diagnostic effectiveness of each potential biomarker, the ROC curves were generated through the R package 'pROC' [39]. The diagnostic accuracy of distinguishing disease samples from control samples was assessed in each dataset by calculating the area under the ROC curve (AUC) as well as the corresponding 95% confidence interval (CI). Following that, the 'rms' R package was utilized to produce the nomogram. Each gene's expression level is evaluated and assigned a score by the nomogram, and then we compute the total number of the scores to obtain a final result that could be applied for predicting the probability of developing SLE and AF. In the meantime, we created the nomogram's ROC curve. The ideal value of AUC for estimating the likelihood of SLE and AF was more than 0.7.

2.10 Evaluation of immune infiltration

The algorithm 'Cibersort' was created by the Alizadeh Lab and the Newman Lab as an analytical tool. Using gene expression data as a foundation, this approach leverages the power to deduce gene expression patterns and provide an approximation of the relative abundance of distinct cell types within a complex cell mixture. This method allows for the determination of the distribution of 22 distinct immune cell types between the disease and control groups [40]. The distribution of each immune cell populations observed in each sample was showcased by barplot. By employing the Wilcoxon rank-sum test, we were able to establish the discrepancies in immune cell abundance between the groups afflicted with the illness and the control group. To present these findings in a clear and concise manner, a boxplot was employed as a visual representation. Performing the R package called 'corrplot', a heatmap was constructed to depict the correlation between different types of immune cells and diseases [41]. Employing heatmaps, a Spearman correlation analysis was conducted to assess the association between the abundance of immune cells and the expression levels of every biomarker, presenting the results in a visual format.

2.11 Consensus clustering analysis for SLE

By employing the R package 'ConsensusClusterPlus', we effectively utilized a range of techniques including constructing the consensus matrix plot, consensus cumulative distribution function (CDF) plot, and tracking plot. These methods allowed us to evaluate the relative changes in the area under the CDF curve, enabling us to determine the optimal number of clusters. The t-SNE (t-distributed stochastic neighbor embedding) technique was exerted to identify the variation in gene expression between different subtypes of SLE, as well as the unique genes associated with each subtype [42]. The resulting t-SNE plot was generated using the 'ggplot2' package for visualization [43].

Determination of DEGs in SLE/AF training datasets

A total of 2388 DEGs were detected between the SLE and control groups using GEO2R. Among these, 995 genes were found to be up-regulated, while 1393 genes were down-regulated. These findings were visually represented in Fig. 2A through a volcano plot. To eliminate inter-batch discrepancies before merging the gene expression datasets from GSE41177 and GSE79768, we use the 'sva' R package. Using the R package 'Limma', we identified a total of 1131 DEGs between healthy control individuals and patients with AF. Among these DEGs, 566 were found to be significantly down-regulated, while 565 were significantly up-regulated in AF (Fig. 2B).

DEGs identified in the SLE/AF training datasets. (A) The volcano plot of SLE training dataset. Downregulated genes are marked in light blue; upregulated genes are marked in red. (B) The volcano plot of AF training dataset. Downregulated genes are marked in green; upregulated genes are marked in red.

WGCNA of DEGs in GSE50772 datasets

The soft threshold power in the GSE50772 dataset was recorded at 14 (Fig. 3A). Afterwards, we proceeded to identify dynamic modules, making sure that each module contained a minimum of 30 genes (Figs. 3B). A total of seven important modules were clustered. Out of all the modules analyzed, the gray module was labeled as non-functional as it failed to effectively group genes together. Additionally, both the brown module and the turquoise module showed a significant negative correlation with SLE (r = -0.60, p = 8.4 × 10 − 9 and r = -0.50, p = 2.0 × 10 − 6 respectively, as shown in Fig. 3C). In SLE, there is a correlation between the brown/turquoise module and gene significance, as depicted in Fig. 3D.

WGCNA of DEGs in SLE training dataset. (A) The soft threshold selection. β = 0.9 was choose as the most appropriate threshold. (B) Gene cluster tree of different modules. (C) The association between brown/turquoise module membership and gene significance in SLE. (D) Correlation analysis of modules and groups.

Identification of common DEGs between SLE and AF

To delve deeper into our investigation, we isolated a combined total of 932 genes from both the brown and turquoise modules for further scrutiny. By intersecting the DEGs related to AF (n = 1131) and the genes in the brown/turquoise module associated with SLE (n = 932), we were able to identify a total of 71 DEGs that were shared between the two diseases. After excluding genes that displayed contrasting patterns of expression from the original group of 71 common DEGs, we ended up with a final tally of 29 DEGs (Fig. 4A, B, Additional file 1: Table S1).

The common DEGs of SLE and AF. (A) 71 DEGs were identified by intersecting the DEGs in AF and the brown/turquoise module genes in SLE. (B) 29 DEGs with same expression trends were retained.

Analysis of functional enrichment for common DEGs

Figure 5A displays the KEGG functional enrichment analysis findings for the 29 DEGs. The pathways associated with the 'FoxO signaling pathway', 'ras signaling pathway' and 'MAPK signaling pathway' stood out as the primary enrichment points for the 29 DEGs. The GO terms for BP that showed significant enrichment included 'Hepatocyte apoptotic process', 'positive regulation of carbohydrate metabolic process' and 'regulation of epithelial cell apoptotic process'. Furthermore, the terms 'bicellular tight junction', 'tight junction' and 'apical junction complex' exhibited enrichment for CC. Furthermore, the expression levels of the 29 DEGs showed a strong correlation with 'histone methyltransferase activity (H3-K9 specific)', 'growth factor binding', 'histone methyltransferase activity (H3-K36 specific)', and 'protein tyrosine kinase activity' (Fig. 5B-D). In addition, the co-expression network built by the GeneMANIA database showcased an intricate protein-protein interaction (PPI) network, comprising of 53.20% co-expression, 2.73% co-localization, 42.29% physical interactions and 1.77% genetic interactions (Fig. 5E).

Functional enrichment analysis and co-expression network of 29 common DEGs. (A) KEGG pathway analysis of common DEGs. (B-D) GO analysis (BP, CC, MF) of common DEGs. (E) DEGs and their co-expression genes were analyzed via GeneMANIA.

undefined

3.1 Determination of DEGs in SLE/AF training datasets

A total of 2388 DEGs were detected between the SLE and control groups using GEO2R. Among these, 995 genes were found to be up-regulated, while 1393 genes were down-regulated. These findings were visually represented in Figure 2A through a volcano plot. To eliminate inter-batch discrepancies before merging the gene expression datasets from GSE41177 and GSE79768, we use the 'sva' R package. Using the R package 'Limma', we identified a total of 1131 DEGs between healthy control individuals and patients with AF. Among these DEGs, 566 were found to be significantly down-regulated, while 565 were significantly up-regulated in AF (Figure 2B).

3.2 WGCNA of DEGs in GSE50772 datasets

The soft threshold power in the GSE50772 dataset was recorded at 14 (Figure 3A). Afterwards, we proceeded to identify dynamic modules, making sure that each module contained a minimum of 30 genes (Figures 3B). A total of seven important modules were clustered. Out of all the modules analyzed, the gray module was labeled as non-functional as it failed to effectively group genes together. Additionally, both the brown module and the turquoise module showed a significant negative correlation with SLE (r = -0.60, p = 8.4 × 10−9 and r = -0.50, p = 2.0 × 10−6 respectively, as shown in Figure 3C). In SLE, there is a correlation between the brown/turquoise module and gene significance, as depicted in Figure 3D.

3.3 Identification of common DEGs between SLE and AF

3.4 Analysis of functional enrichment for common DEGs

Figure 5A displays the KEGG functional enrichment analysis findings for the 29 DEGs. The pathways associated with the 'FoxO signaling pathway', 'ras signaling pathway' and 'MAPK signaling pathway' stood out as the primary enrichment points for the 29 DEGs. The GO terms for BP that showed significant enrichment included 'Hepatocyte apoptotic process', 'positive regulation of carbohydrate metabolic process' and 'regulation of epithelial cell apoptotic process'. Furthermore, the terms 'bicellular tight junction', 'tight junction' and 'apical junction complex' exhibited enrichment for CC. Furthermore, the expression levels of the 29 DEGs showed a strong correlation with 'histone methyltransferase activity (H3-K9 specific)', 'growth factor binding', 'histone methyltransferase activity (H3-K36 specific)', and 'protein tyrosine kinase activity' (Figure 5B-D). In addition, the co-expression network built by the GeneMANIA database showcased an intricate protein-protein interaction (PPI) network, comprising of 53.20% co-expression, 2.73% co-localization, 42.29% physical interactions and 1.77% genetic interactions (Figure 5E).

3.5 SVM-RFE algorithm of common DEGs

According to the SVM-RFE findings, the 27 top features of DEGs exhibited the greatest precision and the least amount of mistakes when identifying potential biomarkers for AF (Figure 6). The gene ranking of the 29 common DEGs is demonstrated in Additional file 2: Table S2. The top 27 genes were extracted for follow-up analysis.

3.6 Identification of candidate biomarkers

In each dataset, the expression of the 27 DEGs obtained above in each dataset were assessed employing the Wilcoxon rank-sum test. The R packages 'ggplot2' and 'ggpubr' were utilized to illustrate the outcomes in a visual format (Figure 7A-D). Candidate biomarkers were identified by isolating genes with significant expression (statistically significant defined as a p-value less than 0.05) from all datasets and then intersecting them. Finally, ANKRD36B, SLC4A4, ANKRD12, MTUS1 and DSC1 these five genes were identified as candidate biomarkers and underwent subsequent analysis (Figure 7E).

3.7 Single-sample gene-set enrichment analysis (ssGSEA)

According to the information provided in Figure 8A-D, it was observed that the five potential biomarkers exhibited significant down-regulation in both the training and validation datasets when compared to the control. The correlation analysis (Figure 8E) showed that several pathways in SLE, including 'CHOLESTEROL_HOMEOSTASIS', 'MTORC1_SIGNALING', 'INFLAMMATORY_RESPONSE', 'P53_PATHWAY', 'FATTY_ACID_METABOLISM', 'GLYCOLYSIS' and 'REACTIVE_OXYGEN_SPECIES_PATHWAY' had a significant negative association with the five candidate genes. These hallmark gene sets were widely acknowledged for their close association with lipid metabolism, the inflammatory response, as well as cell proliferation and differentiation. AF was significantly influenced by the inflammatory response, which was a pivotal element in its initiation and advancement. Additionally, these findings indicated that the occurrence of AF in individuals diagnosed with SLE was linked to the expression of the 5 potential biomarkers.

3.8 Assessment of the prognostic significance of the 5 biomarkers and nomograms

The training and validation datasets were subjected to ROC analysis to assess how effectively the 5 (ANKRD36B, SLC4A4, ANKRD12, MTUS1 and DSC1) biomarkers can predict the diagnostic accuracy for SLE and AF. In the SLE training dataset, the AUC values and the related 95% CI of the 5 biomarkers were depicted in Figure 9A: ANKRD36B (AUC 0.925, 95%CI 0.858–0.993), SLC4A4 (AUC 0.791, 95%CI 0.696–0.886), ANKRD12 (AUC 0.884, 95%CI 0.804–0.965), MTUS1 (AUC 0.787, 95%CI 0.680–0.893), DSC1 (AUC 0.815, 95%CI 0.772–0.908). The AUC values and the corresponding 95% CI of the biomarkers in AF training dataset were illustrated in Figure 9B: ANKRD36B (AUC 0.702, 95%CI 0.541–0.862), SLC4A4 (AUC 0.808, 95%CI 0.693–0.923), ANKRD12 (AUC 0.678, 95%CI 0.518–0.837), MTUS1 (AUC 0.728, 95%CI 0.582–0.875), DSC1 (AUC 0.685, 95%CI 0.541–0.830). In Figure 9C, we could see that the AUC value and 95%CI in SLE validation dataset of the biomarkers were: ANKRD36B (AUC 0.767, 95%CI 0.639–0.895), SLC4A4 (AUC 0.933, 95%CI 0.867–0.999), ANKRD12 (AUC 0.707, 95%CI 0.567–0.848), MTUS1 (AUC 0.739, 95%CI 0.604–0.873), DSC1 (AUC 0.853, 95%CI 0.746–0.961). The Figure 9D represented the AUC value and 95%CI of the biomarkers in AF validation dataset were: ANKRD36B (AUC 0.753, 95%CI 0.578–0.929), SLC4A4 (AUC 0.73, 95%CI 0.562–0.898), ANKRD12 (AUC 0.753, 95%CI 0.583–0.924), MTUS1 (AUC 0.837, 95%CI 0.689–0.984), DSC1 (AUC 0.88, 95%CI 0.77–0.99). A predictive nomogram was developed for each dataset, assigning a score to the relative expression of each gene. The total score was then calculated by adding up the individual scores of all the genes. (Figure 9E-H). The results of ROC analysis of the nomogram in the 4 datasets were shown in Figure 9I-L: SLE training dataset (AUC 0.995, 95%CI 0.985–1.000), AF training dataset (AUC 0.837, 95%CI 0.709–0.965), SLE validation dataset (AUC 0.955, 95%CI 0.907–1.000), AF validation dataset (AUC 0.943, 95%CI 0.862–1.000).

3.9 Immune infiltration analysis in SLE/AF training datasets

The boxplot (Figure 10A) and barplot (Additional file 3: Figure S1 A) indicated that the proportion of Monocytes, Neutrophils, B cells memory and Plasma cells was greater in SLE than in control groups, whereas the percentage of NK cells resting, T cells CD4 memory resting and Mast cells resting was lower. While in AF, there was a higher abundance of Monocytes and Neutrophils compared to the control groups, while T cells CD4 memory resting, T cells CD4 memory activated and Mast cells activated showed a decrease. This was demonstrated by both the barplot (Additional file 3: Figure S1 B) and boxplot (Figure 10B).

In the study on SLE, the correlation analysis highlighted a strong positive relationship between the activation of NK cells and Mast cells, while it was found that Macrophages M1 exhibited a significant positive correlation with resting Mast cells (Additional file 3: Figure S2 A). However, in AF many immune cells had significant correlations between them, according to correlation analysis. The Mast cells activated showed a notable and robust positive correlation with T cells CD4 naive. Moreover, Plasma cells exhibited a notable and powerful positive correlation with B cells memory. Furthermore, NK cells resting exhibited a significant and strong positive correlation with Neutrophils. In contrast, A clear and striking inverse connection was observed between monocytes and T cells gamma delta, while T cells regulatory displayed a noteworthy and powerful negative correlation with T cells gamma delta (Additional file 3: Figure S2 B).

It is worth noting that Figure 10C provides compelling evidence of a positive association between the five candidate biomarkers and the infiltration of T cells CD4 memory resting in SLE. In contrast, the study conducted in AF revealed a association between the five potential biomarkers and the infiltration of Monocytes, showing a negative correlation. However, a positive relationship was observed between these biomarkers and the infiltration of Macrophages M1 and T cells gamma delta. (Figure 10D).

3.10 Consensus clustering analysis

Our approach involved employing consensus clustering analysis to examine the DEG expression profile within the SLE dataset GSE50772. Figure 11A indicated that the strongest clustering effect between different groups was observed when the parameter k was assigned a value of 3. By employing the CDF, the k value was determined as the distribution reached its maximum, signifying the attainable stability in the given context. In Figure 11B, when k=3, the curve had a plateau. Moreover, the tracking plot (Figure 11C) provided insights into how projects were clustered together using consensus clustering at various k values. It could be seen from the figure that the sample clustering is clear when k=3. The maximum number of clusters was determined by calculating the average consistency within each group of clusters. According to Figure 11D, it became evident that when k=3, every cluster displayed a significantly high value. Therefore, we divided SLE into A, B and C three unsupervised clustering subgroups. In Figure 11E's t-SNE plot, it is worth mentioning that there was a significant differentiation observed among the three subtypes A, B, and C.

3.11 Analysis of subtypes of SLE

The expression of the five potential biomarkers in three subtypes was shown to vary significantly in Figure 12A. More precisely, when it came to gene expression, ANKRD12 and MTUS1 were found to be more prominent in subtype A as opposed to subtypes B and C, while subtype B exhibited a higher expression of SLC4A4 than subtypes A and C. Furthermore, in subtype C, there was a noticeable increase in the expression levels of ANKRD36B and DSC1 compared to subtypes A and B. The boxplot in Figure 12A and the heatmap in Figure 12B both displayed the same trend. In subtype A, we observed a significant increase in the presence of T cells CD4 memory resting and Neutrophils compared to subtypes B and C. On the other hand, subtype B showed a higher level of infiltration of NK cells resting when compared to subtypes A and C (Figure 12C). Furthermore, subtype C exhibited higher levels of infiltration by memory B cells and Plasma cells in comparison to subtypes A and B.

According to reports, atrial fibrosis, ectopy, chronic inflammation, and impaired conduction are all key factors in the pathophysiology of AF [44]. It is increasingly evidenced that chronic inflammation is a characteristic of AF. Numerous systemic conditions, including coronary artery disease, hypertension, SLE, obesity, surgery, and ablation, exhibit a connection to low-grade inflammation and elevated levels of proinflammatory cytokines [45]. Inflammation can lead to AF, which in turn can promote inflammatory response [45]. It is a common knowledge that SLE is an autoimmune disorder [1] that is accompanied by inflammation. These two prevalent disorders were the subjects of this study's application of machine learning and bioinformatics analysis techniques. This study focused solely on examining the presence of potential biomarkers in blood samples taken from SLE patients, using a reliable and valuable clinical technique, in order to ascertain the likelihood of AF occurrence. Once we had devised our nomogram, determining the cumulative score became as simple as adding up the scores assigned to each gene. In general, our nomogram has the potential to be utilized in clinical settings as it is important to promptly examine and treat SLE patients with elevated overall scores for an improved prognosis. A complete solution to analyze gene expression data is provided by the GEO2R online DEGs analysis tool and the 'Limma' R package [31]. The present investigation found 2388 DEGs between the SLE and control groups using GEO2R, whereas the 'Limma' R package identified 1131 DEGs in the AF training datasets. WGCNA serves as a scientific methodology in the realm of biology, enabling scientists to investigate the intricate network of gene connections across diverse sample sets. Through the construction of a network that shows the co-expression of genes and the identification of gene clusters that are associated, it is possible to ascertain the most pertinent connection between gene modules and phenotypes. Through this study, we discovered 947 noteworthy DEGs within SLE modules. We employed WGCNA to pinpoint gene clusters consisting of interconnected, associated and overlapping genes in both SLE and AF. A total of 71 shared risk genes associated with SLE and AF were identified by intersecting the DEGs associated with AF (n = 1131) and the module DEGs associated with SLE (n = 947). After deleting genes with opposite expression trends, 29 DEGs are left.

In order to shed light on the roles of these 29 shared DEGs in both SLE and AF, a comprehensive examination was undertaken through KEGG pathway and GO analysis. The DEGs were predominantly found in KEGG signaling pathways such as the 'FoxO signaling pathway', 'ras signaling pathway', and 'MAPK signaling pathway', indicating a potential association with immunoinflammatory responses. The FoxO signaling pathway is identified as one of the immunological metabolic signaling pathways in a study [46]. In quiescent immune cell populations, FoxO protein are more transcriptionally active, which encourages the development of the steady-state cytokine receptor IL7R and cellular transport molecules (such as CD62L, CCR7, and S1PR1) through KLF2. Recent research reveals that proper FoxO1 down-regulation is essential for T cells' internal environmental stability, which evidenced that FoxO protein is correlated with T cells response [47]. The latest study showed that the connection between the Ras signaling pathway [48] and acute inflammatory response has been established. Associated with Ras activation, RasGRP1, which falls under the RasGRP family, acts as a guanine nucleotide exchange factor [49, 50]. While RasGRP1 demonstrates significant expression primarily in T cells, it is also detectable in a range of other cell types such as B cells, Mast cells, NK cells, neuronal cells, and activated macrophages, albeit at varying levels [51–56]. The most recent discoveries in the field highlight the crucial contribution of RasGRP1 in effectively managing acute inflammation, as it stimulates the production of the proinflammatory cytokine IL-6. Its approach to tackling inflammation-related cancer involves competing with let-7a and utilizing both protein-dependent and protein-independent methods to significantly lower the chances of developing this type of cancer. This finding has the potential to provide insights into the infrequent association between acute inflammation and the development of tumors [48]. Moreover, the secretion of inflammatory cytokines by various types of immune cells is also influenced by the MAPK signaling pathway [57]. Collectively, numerous studies and pieces of evidence indicate that these 29 DEGs are associated with the immune pathogenesis of AF and SLE.

Furthermore, the GeneMANIA database was utilized to construct a co-expression network by employing 29 shared DEGs and predicting their roles. In the field of machine learning, the SVM-RFE technique is widely used to give priority to certain features and select the most crucial ones for classification purposes. It can remove unnecessary elements and retain only variables associated with the conclusion. SVM-RFE stands out as a top-notch method for selecting features, and it has proven effective in identifying key genes for various disorders [58, 59]. Among the genes analyzed in our research, a total of 27 DEGs demonstrated exceptional precision and minimal mistakes when identifying potential biomarkers. Setting the threshold for statistical significance at a P-value below 0.05, we proceeded to analyze the expression significance of these 27 genes within both the training and validation sets. Subsequently, we isolated the significantly expressed DEGs from each dataset and intersected them (Fig. 7A-E). Eventually, we obtained five candidate biomarkers (ANKRD36B, SLC4A4, ANKRD12, MTUS1, DSC1). In order to verify the significant expression of these five potential biomarkers in each dataset, we performed a recalculation of the biomarker expression in the 4 datasets (Fig. 8A-D). Afterwards, a nomogram was generated and the predictive ability for patients with SLE and AF in each dataset was evaluated using ROC analysis for the five potential biomarkers (Fig. 9A-L). During the ROC analysis, it was found that the five potential biomarkers and the nomograms exhibited exceptional specificity and sensitivity in identifying SLE and AF in every dataset.

The exact function of ANKRD36B and its role in AF or SLE are still not understood. Melanoma is linked to ANKRD36B according to the GeneCards database (https//www.genecards.org). SLC4A4 codes for NBCe1 protein, an electroneutral sodium bicarbonate transporter that plays a crucial role in maintaining pH balance and homeostasis in healthy tissues [60]. Epithelial ductal cells primarily express the bicarbonate transporter, which is associated with an acidic tumor microenvironment (TME) that promotes the advancement of cancer, resistance to therapy, and evasion of the immune system [61]. SLC4A4's function in SLE and AF, however, is still uncertain. Multiple malignancies have been linked to the tumor suppressor gene MTUS1 [62]. Cell polarity and microtubule balance during cardiac development are two processes that are affected by the microtubule-associated protein encoded by MTUS1 [63]. Speculation has arisen that MTUS1 may be involved in the initiation of AF due to the growing evidence connecting atrial fibrosis remodeling and the promotion of AF [59]. A study has found a connection between MTUS1 and autoimmune conditions, such as SLE, as well as cardiovascular disorders like spontaneous heart hypertrophy and lymphoproliferative disease, which exhibits similarities to SLE in mice [64]. A recent study has revealed that the protein coding by DSC1 (Desmocollin 1) is found to be extensively distributed throughout atherosclerotic regions, and it also has an impact on the formation of high-density lipoprotein [65]. The activity of atherosclerotic HDL is hindered when it comes in contact with DSC1-generated apolipoproteins (also known as apoA-I binding proteins) that effectively capture apoA-I. The apoA-I retention theory suggests that the expression of DSC1 by macrophages leads to the inactivation of apoA-I and facilitates the development of atherosclerosis [66]. However, there is a strong relationship between systemic lupus erythematosus and atherosclerosis [67]. Therefore, Therefore, DSC1 may be a target for atherosclerosis caused by SLE. However, several investigations found a positive connection between DSC1 and autoimmune illnesses [68, 69], suggesting that DSC1 and SLE are also linked in terms of immunity.

Immunologic disturbances caused by SLE, an autoimmune disease with a variety of clinical presentations, harm cardiac tissue, causing AF to appear and progress. AF has been consistently associated with the existence of immune cells and specific proteins that trigger an inflammatory response in the heart tissue and circulatory systems. However, it is yet to be determined if this inflammation in AF is directly linked to an autoimmune reaction [45]. To provide an instance, specific alterations in the genetic makeup of IL-1, IL-6, and IL-10 sequences, which play a crucial role in managing inflammatory cytokine levels, have been separately connected to atrial fibrillation in individuals [70–72]. Numerous case-control studies have indicated that individuals diagnosed with AF tend to display heightened levels of inflammatory indicators in comparison to individuals with a normal sinus rhythm, such as C-reactive protein (CRP), heat shock protein 27 (HSP27), interleukin-6 (IL-6), interleukin-8 (IL-8), and tumor necrosis factor (TNF). Additionally, neutrophil and lymphocyte ratios were also found to be enhanced in these patients [73–76]. Based on our examination of immune infiltration, AF exhibited a higher proportion of Monocytes and Neutrophils compared to the control group. On the flip side, AF exhibited a decreased percentage of resting memory CD4 T cells, activated memory CD4 T cells, and activated mast cells. The findings of our immune infiltration analysis (Fig. 10A, B) are in accord with those of the research mentioned above.

Then, we looked at the connection between Immunologic infiltration and the five candidate biomarkers (ANKRD36B, SLC4A4, ANKRD12, MTUS1, DSC1). FIGURE 12B shows that in SLE, there is a favorable correlation between the five identified genes and infiltrations of resting T cells CD4 memory. Within the AF framework, the five pivotal genes showcased a favorable connection with the infiltration of Macrophages M1 and T cells gamma delta, while simultaneously exhibiting an unfavorable association with monocyte infiltrations (Fig. 10D). Moreover, a bioinformatics study has discovered that SLC4A4 is among the central genes implicated in schizophrenia and is linked to the presence of immunological infiltration in individuals with schizophrenia [77]. As was already indicated, SLC4A4 is correlated with acidic TME [61]. As previously stated, MTUS1 deficiency is linked to autoimmune disorders [64]. The above-mentioned findings, along with various other studies, provide evidence to suggest that the five potential biomarkers have an immunological connection to both AF and SLE. As a result, they can not only anticipate the occurrence of SLE companied with AF but also effectively control the immune response during assistant therapy.

Finally, using the expression profiles, we distinguished three SLE subtypes. The five identified genes have different expression in three subtypes. In Fig. 12B, we could see subtype A exhibited higher expression levels of ANKRD12 and MTUS1 compared to subtypes B and C, whereas subtype B showed higher expression of SLC4A4 than subtypes A and C. Furthermore, subtypes A and B demonstrated lower levels of ANKRD36B and DSC1 expression when compared to subtype C. Through a comprehensive examination of the expression levels of the five key genes, we can effectively distinguish between various subtypes of SLE. Figure 12C showed a greater presence of CD4 memory resting T cells and Neutrophils in subtype A, while subtype B and C had lower infiltration. In comparison to subtypes A and C, subtypes B demonstrated a higher level of infiltration by resting NK cells, whereas subtypes A and B exhibited a lower infiltration of memory B cells and Plasma cells than subtype C. Figure 10A illustrates an increase in the abundance of Monocytes, Neutrophils, B cells memory, and Plasma cells in individuals with SLE. Conversely, a decrease in NK cells resting, T cells CD4 memory resting, and Mast cells resting was evident in SLE when compared to the control group. As the results presented above, the concentration of subtype C immune cells was found to be higher than that of the control group, reinforcing our findings. Immunotherapy is expected to be most beneficial for patients classified under subtype C, offering potential guidance for personalized medicine. Immunotherapy proves particularly advantageous for patients categorized under subtype C, providing a heightened potential for tailored and individualized treatment alternatives.

But there were certain restrictions on our investigation. Initially, through the amalgamation of diverse AF datasets, various sample types were incorporated, encompassing LAA (left atrial appendage), LAPV (left atrial-pulmonary vein junction), RAA (right atrial appendage), and left ventricular myocardium. This amalgamation introduced a potential bias that needed verification in our own clinical samples. Experimental studies are still necessary to validate the crucial biomarkers and underlying mechanisms, despite the utilization of validation datasets (GSE81622 and GSE2240) to assess the predictive significance.

Our study methodically unearthed five prospective central genes (ANKRD36B, SLC4A4, ANKRD12, MTUS1, DSC1) and devised a nomogram to gauge the probability of SLE accompanied by AF by way of diverse bioinformatic analyses approaches and advanced machine learning algorithms. The outcomes of our study serve as a stepping stone for future research endeavors, allowing for in-depth examinations of crucial genes possibly connected to AF in individuals diagnosed with SLE.

Acknowledgments

None.

Author Contributions

J. Ge and S. An devised the concept and supervised the study. R. Wang, J. Liu and T. Zhang designed the study. S. Yao, Q. Yang and T. Zhu performed the collection and analysis of data. R. Wang, J. Liu and T. Zhang participated in the process of image and interpretation of data. All the authors drafted the manuscript and approved its submission.

Funding

This study was supported by grants from the National Natural Science Foundation of China (No. 81970312).

Availability of data and materials

We exclusively sourced all the datasets utilized in this study from the GEO database (http:// www.ncbi.nlm.nih.gov/geo). The datasets supporting the conclusions of this article is included within the article and its additional file. R code and R data files for analysis are also available if required.

Ethical approval and consent to participate

The data analyzed in this study were from public databases, so ethical approval and consent participation were not required. All methods in this study were performed in accordance with the relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The author declares that no competing interest.

Kuo CF, Grainge MJ, Valdes AM, See LC, Luo SF, Yu KH, Zhang W, Doherty M: Familial Aggregation of Systemic Lupus Erythematosus and Coaggregation of Autoimmune Diseases in Affected Families. JAMA Intern Med 2015, 175(9):1518-1526.
Rees F, Doherty M, Grainge M, Davenport G, Lanyon P, Zhang W: The incidence and prevalence of systemic lupus erythematosus in the UK, 1999-2012. Ann Rheum Dis 2016, 75(1):136-141.
Hak AE, Karlson EW, Feskanich D, Stampfer MJ, Costenbader KH: Systemic lupus erythematosus and the risk of cardiovascular disease: results from the nurses' health study. Arthritis Rheum 2009, 61(10):1396-1402.
Bartels CM, Buhr KA, Goldberg JW, Bell CL, Visekruna M, Nekkanti S, Greenlee RT: Mortality and cardiovascular burden of systemic lupus erythematosus in a US population-based cohort. J Rheumatol 2014, 41(4):680-687.
Teixeira RA, Borba EF, Pedrosa A, Nishioka S, Viana VS, Ramires JA, Kalil-Filho R, Bonfá E, Martinelli Filho M: Evidence for cardiac safety and antiarrhythmic potential of chloroquine in systemic lupus erythematosus. Europace 2014, 16(6):887-892.
Lau DH, Linz D, Sanders P: New Findings in Atrial Fibrillation Mechanisms. Card Electrophysiol Clin 2019, 11(4):563-571.
Staerk L, Sherer JA, Ko D, Benjamin EJ, Helm RH: Atrial Fibrillation: Epidemiology, Pathophysiology, and Clinical Outcomes. Circ Res 2017, 120(9):1501-1517.
Benjamin EJ, Levy D, Vaziri SM, D'Agostino RB, Belanger AJ, Wolf PA: Independent risk factors for atrial fibrillation in a population-based cohort. The Framingham Heart Study. Jama 1994, 271(11):840-844.
Wang TJ, Parise H, Levy D, D'Agostino RB, Sr., Wolf PA, Vasan RS, Benjamin EJ: Obesity and the risk of new-onset atrial fibrillation. Jama 2004, 292(20):2471-2477.
Alonso A, Lopez FL, Matsushita K, Loehr LR, Agarwal SK, Chen LY, Soliman EZ, Astor BC, Coresh J: Chronic kidney disease is associated with the incidence of atrial fibrillation: the Atherosclerosis Risk in Communities (ARIC) study. Circulation 2011, 123(25):2946-2953.
Chung MK, Martin DO, Sprecher D, Wazni O, Kanderian A, Carnes CA, Bauer JA, Tchou PJ, Niebauer MJ, Natale A et al: C-reactive protein elevation in patients with atrial arrhythmias: inflammatory mechanisms and persistence of atrial fibrillation. Circulation 2001, 104(24):2886-2891.
Aviles RJ, Martin DO, Apperson-Hansen C, Houghtaling PL, Rautaharju P, Kronmal RA, Tracy RP, Van Wagoner DR, Psaty BM, Lauer MS et al: Inflammation as a risk factor for atrial fibrillation. Circulation 2003, 108(24):3006-3010.
Engelmann MD, Svendsen JH: Inflammation in the genesis and perpetuation of atrial fibrillation. Eur Heart J 2005, 26(20):2083-2092.
Barnado A, Carroll RJ, Casey C, Wheless L, Denny JC, Crofford LJ: Phenome-Wide Association Studies Uncover a Novel Association of Increased Atrial Fibrillation in Male Patients With Systemic Lupus Erythematosus. Arthritis Care Res (Hoboken) 2018, 70(11):1630-1636.
Lim SY, Bae EH, Han KD, Jung JH, Choi HS, Kim CS, Ma SK, Kim SW: Systemic lupus erythematosus is a risk factor for atrial fibrillation: a nationwide, population-based study. Clin Exp Rheumatol 2019, 37(6):1019-1025.
Zhou Y, Shi W, Zhao D, Xiao S, Wang K, Wang J: Identification of Immune-Associated Genes in Diagnosing Aortic Valve Calcification With Metabolic Syndrome by Integrated Bioinformatics Analysis and Machine Learning. Front Immunol 2022, 13:937886.
Yang Q, Li B, Tang J, Cui X, Wang Y, Li X, Hu J, Chen Y, Xue W, Lou Y et al: Consistent gene signature of schizophrenia identified by a novel feature selection strategy from comprehensive sets of transcriptomic data. Brief Bioinform 2020, 21(3):1058-1068.
Tang J, Mou M, Wang Y, Luo Y, Zhu F: MetaFS: Performance assessment of biomarker discovery in metaproteomics. Brief Bioinform 2021, 22(3).
Fu J, Tang J, Wang Y, Cui X, Yang Q, Hong J, Li X, Li S, Chen Y, Xue W et al: Discovery of the Consistently Well-Performed Analysis Chain for SWATH-MS Based Pharmacoproteomic Quantification. Front Pharmacol 2018, 9:681.
Barrett T, Wilhite SE, Ledoux P, Evangelista C, Kim IF, Tomashevsky M, Marshall KA, Phillippy KH, Sherman PM, Holko M et al: NCBI GEO: archive for functional genomics data sets--update. Nucleic Acids Res 2013, 41(Database issue):D991-995.
Kennedy WP, Maciuca R, Wolslegel K, Tew W, Abbas AR, Chaivorapol C, Morimoto A, McBride JM, Brunetta P, Richardson BC et al: Association of the interferon signature metric with serological disease manifestations but not global activity scores in multiple cohorts of patients with SLE. Lupus Sci Med 2015, 2(1):e000080.
Yeh YH, Kuo CT, Lee YS, Lin YM, Nattel S, Tsai FC, Chen WJ: Region-specific gene expression profiles in the left atria of patients with valvular atrial fibrillation. Heart Rhythm 2013, 10(3):383-391.
Tsai FC, Lin YC, Chang SH, Chang GJ, Hsu YJ, Lin YM, Lee YS, Wang CL, Yeh YH: Differential left-to-right atria gene expression ratio in human sinus rhythm and atrial fibrillation: Implications for arrhythmogenesis and thrombogenesis. Int J Cardiol 2016, 222:104-112.
Zhu H, Mi W, Luo H, Chen T, Liu S, Raman I, Zuo X, Li QZ: Whole-genome transcription and DNA methylation analysis of peripheral blood mononuclear cells identified aberrant gene regulation pathways in systemic lupus erythematosus. Arthritis Res Ther 2016, 18:162.
Barth AS, Merk S, Arnoldi E, Zwermann L, Kloos P, Gebauer M, Steinmeyer K, Bleich M, Kääb S, Hinterseer M et al: Reprogramming of the human atrial transcriptome in permanent atrial fibrillation: expression of a ventricular-like genomic signature. Circ Res 2005, 96(9):1022-1029.
Yao M, Zhang C, Gao C, Wang Q, Dai M, Yue R, Sun W, Liang W, Zheng Z: Exploration of the Shared Gene Signatures and Molecular Mechanisms Between Systemic Lupus Erythematosus and Pulmonary Arterial Hypertension: Evidence From Transcriptome Data. Front Immunol 2021, 12:658341.
Li F, Zhou Y, Zhang Y, Yin J, Qiu Y, Gao J, Zhu F: POSREG: proteomic signature discovered by simultaneously optimizing its reproducibility and generalizability. Brief Bioinform 2022, 23(2).
Tang J, Fu J, Wang Y, Li B, Li Y, Yang Q, Cui X, Hong J, Li X, Chen Y et al: ANPELA: analysis and performance assessment of the label-free quantification workflow for metaproteomic studies. Brief Bioinform 2020, 21(2):621-636.
Tang J, Fu J, Wang Y, Luo Y, Yang Q, Li B, Tu G, Hong J, Cui X, Chen Y et al: Simultaneous Improvement in the Precision, Accuracy, and Robustness of Label-free Proteome Quantification by Optimizing Data Manipulation Chains. Mol Cell Proteomics 2019, 18(8):1683-1699.
Fu J, Zhang Y, Liu J, Lian X, Tang J, Zhu F: Pharmacometabonomics: data processing and statistical analysis. Brief Bioinform 2021, 22(5).
Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK: limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015, 43(7):e47.
Langfelder P, Horvath S: WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 2008, 9:559.
Kanehisa M, Furumichi M, Tanabe M, Sato Y, Morishima K: KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res 2017, 45(D1):D353-d361.
The Gene Ontology Resource: 20 years and still GOing strong. Nucleic Acids Res 2019, 47(D1):D330-d338.
Yu G, Wang LG, Han Y, He QY: clusterProfiler: an R package for comparing biological themes among gene clusters. Omics 2012, 16(5):284-287.
Warde-Farley D, Donaldson SL, Comes O, Zuberi K, Badrawi R, Chao P, Franz M, Grouios C, Kazi F, Lopes CT et al: The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res 2010, 38(Web Server issue):W214-220.
Huang ML, Hung YH, Lee WM, Li RK, Jiang BR: SVM-RFE based feature selection and Taguchi parameters optimization for multiclass SVM classifier. ScientificWorldJournal 2014, 2014:795624.
Hänzelmann S, Castelo R, Guinney J: GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 2013, 14:7.
Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, Müller M: pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011, 12:77.
Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, Hoang CD, Diehn M, Alizadeh AA: Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 2015, 12(5):453-457.
Hu K: Become Competent within One Day in Generating Boxplots and Violin Plots for a Novice without Prior R Experience. Methods Protoc 2020, 3(4).
Kobak D, Berens P: The art of using t-SNE for single-cell transcriptomics. Nat Commun 2019, 10(1):5416.
Ito K, Murphy D: Application of ggplot2 to Pharmacometric Graphics. CPT Pharmacometrics Syst Pharmacol 2013, 2(10):e79.
Takawale A, Aguilar M, Bouchrit Y, Hiram R: Mechanisms and Management of Thyroid Disease and Atrial Fibrillation: Impact of Atrial Electrical Remodeling and Cardiac Fibrosis. Cells 2022, 11(24).
Hu YF, Chen YJ, Lin YJ, Chen SA: Inflammation and the pathogenesis of atrial fibrillation. Nat Rev Cardiol 2015, 12(4):230-243.
Molaei M, Vandehoef C, Karpac J: NF-κB Shapes Metabolic Adaptation by Attenuating Foxo-Mediated Lipolysis in Drosophila. Dev Cell 2019, 49(5):802-810.e806.
Luo CT, Li MO: Foxo transcription factors in T cell biology and tumor immunity. Semin Cancer Biol 2018, 50:13-20.
Wang C, Li X, Xue B, Yu C, Wang L, Deng R, Liu H, Chen Z, Zhang Y, Fan S et al: RasGRP1 promotes the acute inflammatory response and restricts inflammation-associated cancer cell growth. Nat Commun 2022, 13(1):7001.
Ebinu JO, Bottorff DA, Chan EY, Stang SL, Dunn RJ, Stone JC: RasGRP, a Ras guanyl nucleotide- releasing protein with calcium- and diacylglycerol-binding motifs. Science 1998, 280(5366):1082-1086.
Stone JC: Regulation of Ras in lymphocytes: get a GRP. Biochem Soc Trans 2006, 34(Pt 5):858-861.
Ebinu JO, Stang SL, Teixeira C, Bottorff DA, Hooton J, Blumberg PM, Barry M, Bleakley RC, Ostergaard HL, Stone JC: RasGRP links T-cell receptor signaling to Ras. Blood 2000, 95(10):3199-3203.
Limnander A, Depeille P, Freedman TS, Liou J, Leitges M, Kurosaki T, Roose JP, Weiss A: STIM1, PKC-δ and RasGRP set a threshold for proapoptotic Erk signaling during B cell development. Nat Immunol 2011, 12(5):425-433.
Liu Y, Zhu M, Nishida K, Hirano T, Zhang W: An essential role for RasGRP1 in mast cell function and IgE-mediated allergic response. J Exp Med 2007, 204(1):93-103.
Lee SH, Yun S, Lee J, Kim MJ, Piao ZH, Jeong M, Chung JW, Kim TD, Yoon SR, Greenberg PD et al: RasGRP1 is required for human NK cell function. J Immunol 2009, 183(12):7931-7938.
Toki S, Kawasaki H, Tashiro N, Housman DE, Graybiel AM: Guanine nucleotide exchange factors CalDAG-GEFI and CalDAG-GEFII are colocalized in striatal projection neurons. J Comp Neurol 2001, 437(4):398-407.
Tang S, Chen T, Yu Z, Zhu X, Yang M, Xie B, Li N, Cao X, Wang J: RasGRP3 limits Toll-like receptor-triggered inflammatory response in macrophages by activating Rap1 small GTPase. Nat Commun 2014, 5:4657.
Behl T, Upadhyay T, Singh S, Chigurupati S, Alsubayiel AM, Mani V, Vargas-De-La-Cruz C, Uivarosan D, Bustea C, Sava C et al: Polyphenols Targeting MAPK Mediated Oxidative Stress and Inflammation in Rheumatoid Arthritis. Molecules 2021, 26(21).
Zheng PF, Chen LZ, Liu P, Pan HW, Fan WJ, Liu ZY: Identification of immune-related key genes in the peripheral blood of ischaemic stroke patients using a weighted gene coexpression network analysis and machine learning. J Transl Med 2022, 20(1):361.
Zhang Y, Xia R, Lv M, Li Z, Jin L, Chen X, Han Y, Shi C, Jiang Y, Jin S: Machine-Learning Algorithm-Based Prediction of Diagnostic Gene Biomarkers Related to Immune Infiltration in Patients With Chronic Obstructive Pulmonary Disease. Front Immunol 2022, 13:740513.
Romero MF, Chen AP, Parker MD, Boron WF: The SLC4 family of bicarbonate (HCO₃⁻) transporters. Mol Aspects Med 2013, 34(2-3):159-182.
Cappellesso F, Orban MP, Shirgaonkar N, Berardi E, Serneels J, Neveu MA, Di Molfetta D, Piccapane F, Caroppo R, Debellis L et al: Targeting the bicarbonate transporter SLC4A4 overcomes immunosuppression and immunotherapy resistance in pancreatic cancer. Nat Cancer 2022, 3(12):1464-1483.
Parbin S, Pradhan N, Das L, Saha P, Deb M, Sengupta D, Patra SK: DNA methylation regulates Microtubule-associated tumor suppressor 1 in human non-small cell lung carcinoma. Exp Cell Res 2019, 374(2):323-332.
Bai X, Zhou Y, Ouyang N, Liu L, Huang X, Tian J, Lv T: A de novo Mutation in the MTUS1 Gene Decreases the Risk of Non-compaction of Ventricular Myocardium via the Rac1/Cdc42 Pathway. Front Pediatr 2019, 7:247.
Zuern C, Krenacs L, Starke S, Heimrich J, Palmetshofer A, Holtmann B, Sendtner M, Fischer T, Galle J, Wanner C et al: Microtubule associated tumor suppressor 1 deficient mice develop spontaneous heart hypertrophy and SLE-like lymphoproliferative disease. Int J Oncol 2012, 40(4):1079-1088.
Choi HY, Ruel I, Malina A, Garrod DR, Oda MN, Pelletier J, Schwertani A, Genest J: Desmocollin 1 is abundantly expressed in atherosclerosis and impairs high-density lipoprotein biogenesis. Eur Heart J 2018, 39(14):1194-1202.
Genest J, Choi HY: Novel Approaches for HDL-Directed Therapies. Curr Atheroscler Rep 2017, 19(12):55.
Frieri M, Stampfl H: Systemic lupus erythematosus and atherosclerosis: Review of the literature. Autoimmun Rev 2016, 15(1):16-21.
Lolle S, Greeff C, Petersen K, Roux M, Jensen MK, Bressendorff S, Rodriguez E, Sømark K, Mundy J, Petersen M: Matching NLR Immune Receptors to Autoimmunity in camta3 Mutants Using Antimorphic NLR Alleles. Cell Host Microbe 2017, 21(4):518-529.e514.
Evangelista F, Roth AJ, Prisayanh P, Temple BR, Li N, Qian Y, Culton DA, Liu Z, Harrison OJ, Brasch J et al: Pathogenic IgG4 autoantibodies from endemic pemphigus foliaceus recognize a desmoglein-1 conformational epitope. J Autoimmun 2018, 89:171-185.
Kato K, Oguri M, Hibino T, Yajima K, Matsuo H, Segawa T, Watanabe S, Yoshida H, Satoh K, Nozawa Y et al: Genetic factors for lone atrial fibrillation. Int J Mol Med 2007, 19(6):933-939.
Lin H, Sinner MF, Brody JA, Arking DE, Lunetta KL, Rienstra M, Lubitz SA, Magnani JW, Sotoodehnia N, McKnight B et al: Targeted sequencing in candidate genes for atrial fibrillation: the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Targeted Sequencing Study. Heart Rhythm 2014, 11(3):452-457.
Marcus GM, Whooley MA, Glidden DV, Pawlikowska L, Zaroff JG, Olgin JE: Interleukin-6 and atrial fibrillation in patients with coronary artery disease: data from the Heart and Soul Study. Am Heart J 2008, 155(2):303-309.
Guo Y, Lip GY, Apostolakis S: Inflammation in atrial fibrillation. J Am Coll Cardiol 2012, 60(22):2263-2270.
Hu YF, Yeh HI, Tsao HM, Tai CT, Lin YJ, Chang SL, Lo LW, Tuan TC, Suenari K, Li CH et al: Electrophysiological correlation and prognostic impact of heat shock protein 27 in atrial fibrillation. Circ Arrhythm Electrophysiol 2012, 5(2):334-340.
Jacob KA, Nathoe HM, Dieleman JM, van Osch D, Kluin J, van Dijk D: Inflammation in new-onset atrial fibrillation after cardiac surgery: a systematic review. Eur J Clin Invest 2014, 44(4):402-428.
Smit MD, Maass AH, De Jong AM, Muller Kobold AC, Van Veldhuisen DJ, Van Gelder IC: Role of inflammation in early atrial fibrillation recurrence. Europace 2012, 14(6):810-817.
Li Z, Li X, Jin M, Liu Y, He Y, Jia N, Cui X, Liu Y, Hu G, Yu Q: Identification of potential biomarkers and their correlation with immune infiltration cells in schizophrenia using combinative bioinformatics strategy. Psychiatry Res 2022, 314:114658.

No competing interests reported.

Additionalfile1TableS1.29DEGs.xlsx
Supplementary Information Additional file 1: Table S1. 29 common DEGs of SLE and AF.
Additionalfile2TableS2.SVMRFE.xlsx
Additional file 2: Table S2. The results of Support Vector Machine-Recursive Feature Elimination (SVM-RFE).
Additionalfile3FigureS12.docx
Additional file 3: Figure S1. Immune infiltration. Figure S2. Correlation analysis between immune cells.

Download PDF

Version 1

posted

You are reading this latest preprint version

Differential gene expression analysis reveals common biomarkers for systemic lupus erythematosus and atrial fibrillation

Status:

Version 1

Abstract

Figures

1 Introduction

2 Methods

2.1 Data acquisition

2.2 Data pre-processing

2.3 Detection of differentially expressed genes (DEGs)

2.4 Detection of important module genes using weighted gene co-expression network analysis (WGCNA)

2.5 Identification of common DEGs

2.6 Enrichment analysis of the common DEGs

2.7 Machine learning algorithms

2.8 Validation and Single-sample gene-set enrichment analysis (ssGSEA) of potential biomarkers

2.9 Assessment of diagnostic effectiveness of biomarkers

2.10 Evaluation of immune infiltration

2.11 Consensus clustering analysis for SLE

3 Results

undefined

3.1 Determination of DEGs in SLE/AF training datasets

3.2 WGCNA of DEGs in GSE50772 datasets

3.3 Identification of common DEGs between SLE and AF

3.4 Analysis of functional enrichment for common DEGs

3.5 SVM-RFE algorithm of common DEGs

3.6 Identification of candidate biomarkers

3.7 Single-sample gene-set enrichment analysis (ssGSEA)

3.8 Assessment of the prognostic significance of the 5 biomarkers and nomograms

3.9 Immune infiltration analysis in SLE/AF training datasets

3.10 Consensus clustering analysis

3.11 Analysis of subtypes of SLE

4 Discussion

5 Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1