2.1 Technology roadmap
A flow chart of the comprehensive analysis of FRDEGs is illustrated in Fig. 1. The roadmap provides an overview of the key steps undertaken in the study, starting from the selection of relevant datasets and preprocessing steps, such as data normalization. It then proceeds to identify ferroptosis-related differentially expressed genes (FRDEGs) and explore their biological functions and pathways. Additionally, the roadmap includes the construction of various regulatory networks, such as protein-protein interaction, TF-mRNA, and mRNA-miRNA networks, to elucidate potential gene regulatory mechanisms. The roadmap also emphasizes the assessment of the interaction between the identified FRDEGs and drugs, providing insights into potential therapeutic interventions for AFib.
2.2 Analysis of gene expression differences related to AFib
Data from the AFib datasets GSE79768 and GSE115574 were divided into the AFib and SR groups, respectively. The datasets of two groups were standardized, data cleaning operations such as annotation probes were carried out, and a box diagram (Fig. 2) of the data distribution before and after standardization was drawn.
The GSE79768 and GSE115574 datasets were analyzed using the R package DESeq2. The results were as follows: 5,365 DEGs were identified in the GSE79768 dataset, of which 2,519 genes were upregulated and 2,846 genes downregulated, and a volcano plot (Fig. 3A) was drawn. There were 1,298 DEGs in the GSE115574 dataset, of which 598 DEGs were upregulated and 700 DEGs downregulated, and a volcano plot (Fig. 3B) was drawn.
The Co-DEGs among different subgroups of samples in the GSE79768 (Fig. 3C) and GSE115574 (Fig. 3D) datasets of AFib were analyzed, and the results are displayed in heat maps using the R-pack pheatmap. Co-DEGs and ferroptosis-related genes in the datasets were intersected to produce 14 FRDEGs, and a Venn diagram (Fig. 3E) was drawn. The 14 FRDEGs were SNAI2, FHL2, DIAPH3, ATF3, CYBB, TLR4, GPX2, SMARCA4, HCFC2, METTL14, CAD, CAV1, COPG1, and VIM.
2.3 GO analyse of FRDEGs
Using GO analyse, the BP, MF, CC, biological pathway, and the relationship between 14 FRDEGs and AFib were further explored. The GO analyse (Table 1) showed that the 14 FRDEGs were mainly concentrated in the BP response to steroid hormone, intracellular receptor signaling pathway, and negative regulation of anoikis, in the CC endocytic vesicle, nuclear matrix, and focal adhesion, and in the MF androgen receptor binding, transcription coactivator activity, and steroid hormone receptor binding. The results were visualized using a histogram (Fig. 4A-4C).
Table 1
Results of GO Enrichment Analysis
ONTOLOGY
|
ID
|
Description
|
GeneRatio
|
BgRatio
|
pvalue
|
p.adjust
|
qvalue
|
BP
|
GO:0048545
|
response to steroid hormone
|
5/14
|
385/18670
|
6.24E-06
|
6.73E-03
|
3.50E-03
|
BP
|
GO:0030522
|
intracellular receptor signaling pathway
|
4/14
|
280/18670
|
4.40E-05
|
1.77E-02
|
9.21E-03
|
BP
|
GO:2000811
|
negative regulation of anoikis
|
2/14
|
17/18670
|
7.06E-05
|
1.77E-02
|
9.21E-03
|
BP
|
GO:0071496
|
cellular response to external stimulus
|
4/14
|
339/18670
|
9.26E-05
|
1.77E-02
|
9.21E-03
|
BP
|
GO:1903409
|
reactive oxygen species biosynthetic process
|
3/14
|
122/18670
|
9.40E-05
|
1.77E-02
|
9.21E-03
|
BP
|
GO:0042535
|
positive regulation of tumor necrosis factor biosynthetic process
|
2/14
|
20/18670
|
9.84E-05
|
1.77E-02
|
9.21E-03
|
BP
|
GO:2000209
|
regulation of anoikis
|
2/14
|
24/18670
|
1.43E-04
|
2.20E-02
|
1.14E-02
|
BP
|
GO:2001236
|
regulation of extrinsic apoptotic signaling pathway
|
3/14
|
155/18670
|
1.91E-04
|
2.34E-02
|
1.22E-02
|
CC
|
GO:0030139
|
endocytic vesicle
|
3/14
|
303/19717
|
1.15E-03
|
3.59E-02
|
2.34E-02
|
CC
|
GO:0016363
|
nuclear matrix
|
2/14
|
109/19717
|
2.64E-03
|
3.59E-02
|
2.34E-02
|
CC
|
GO:0005925
|
focal adhesion
|
3/14
|
405/19717
|
2.65E-03
|
3.59E-02
|
2.34E-02
|
CC
|
GO:0005924
|
cell-substrate adherens junction
|
3/14
|
408/19717
|
2.70E-03
|
3.59E-02
|
2.34E-02
|
CC
|
GO:0030055
|
cell-substrate junction
|
3/14
|
412/19717
|
2.78E-03
|
3.59E-02
|
2.34E-02
|
CC
|
GO:0034708
|
methyltransferase complex
|
2/14
|
113/19717
|
2.83E-03
|
3.59E-02
|
2.34E-02
|
CC
|
GO:0034399
|
nuclear periphery
|
2/14
|
131/19717
|
3.78E-03
|
3.65E-02
|
2.37E-02
|
CC
|
GO:0045335
|
phagocytic vesicle
|
2/14
|
132/19717
|
3.84E-03
|
3.65E-02
|
2.37E-02
|
MF
|
GO:0050681
|
androgen receptor binding
|
2/14
|
44/17697
|
5.39E-04
|
5.07E-02
|
2.95E-02
|
MF
|
GO:0003713
|
transcription coactivator activity
|
3/14
|
319/17697
|
1.82E-03
|
6.74E-02
|
3.92E-02
|
MF
|
GO:0035258
|
steroid hormone receptor binding
|
2/14
|
92/17697
|
2.34E-03
|
6.74E-02
|
3.92E-02
|
MF
|
GO:0009055
|
electron transfer activity
|
2/14
|
114/17697
|
3.56E-03
|
6.74E-02
|
3.92E-02
|
MF
|
GO:0035257
|
nuclear hormone receptor binding
|
2/14
|
152/17697
|
6.23E-03
|
6.74E-02
|
3.92E-02
|
MF
|
GO:0016884
|
carbon-nitrogen ligase activity, with glutamine as amido-N-donor
|
1/14
|
10/17697
|
7.88E-03
|
6.74E-02
|
3.92E-02
|
MF
|
GO:0030957
|
Tat protein binding
|
1/14
|
10/17697
|
7.88E-03
|
6.74E-02
|
3.92E-02
|
MF
|
GO:0017048
|
Rho GTPase binding
|
2/14
|
177/17697
|
8.36E-03
|
6.74E-02
|
3.92E-02
|
GO network (Fig. 4D) was drawn. The lines indicate the corresponding molecule and annotation of the corresponding entry, and the larger the node, the more molecules the entry contains. Finally, GO analyse (Fig. 4E) with logFC was performed on 14 FRDEGs. Based on the enrichment analysis, the z-score corresponding to each entry was calculated using the logFC of the molecule and visualized with a circle graph. The results of GO analyse show that negative regulation of anoikis, endocytic vesicles, and focal adhesion pathways were the maximum positive regulatory pathways; steroid hormone receptor binding, androgen receptor binding, and nuclear matrix pathways were the maximum negative regulatory pathways. Finally, the results of GO analyse was visualized in a Sankey diagram (Fig. 4F). The Sankey diagram shows the relationship between GO (which contains BP, CC, and MF) and the corresponding function, pathway number (ID), and the number of genes contained.
Table 3 Results of GSE115574 GSEA
ID
|
setSize
|
enrichmentScore
|
NES
|
pvalue
|
p.adjust
|
qvalues
|
WP_TYROBP_CAUSAL_NETWORK
|
59
|
0.7451
|
2.4565
|
0.0018
|
0.0690
|
0.0597
|
NABA_CORE_MATRISOME
|
256
|
0.5890
|
2.4257
|
0.0015
|
0.0690
|
0.0597
|
WP_MICROGLIA_PATHOGEN_PHAGOCYTOSIS_PATHWAY
|
40
|
0.7655
|
2.3584
|
0.0018
|
0.0690
|
0.0597
|
NABA_ECM_GLYCOPROTEINS
|
177
|
0.5852
|
2.3143
|
0.0016
|
0.0690
|
0.0597
|
REACTOME_ECM_PROTEOGLYCANS
|
74
|
0.6569
|
2.3065
|
0.0017
|
0.0690
|
0.0597
|
PID_AVB3_INTEGRIN_PATHWAY
|
72
|
0.6397
|
2.2351
|
0.0017
|
0.0690
|
0.0597
|
KEGG_ECM_RECEPTOR_INTERACTION
|
81
|
0.6165
|
2.1916
|
0.0017
|
0.0690
|
0.0597
|
REACTOME_SIGNALING_BY_PDGF
|
58
|
0.6624
|
2.1801
|
0.0018
|
0.0690
|
0.0597
|
REACTOME_COLLAGEN_FORMATION
|
86
|
0.5890
|
2.1075
|
0.0017
|
0.0690
|
0.0597
|
PID_RAC1_REG_PATHWAY
|
38
|
0.6833
|
2.0779
|
0.0018
|
0.0690
|
0.0597
|
KEGG_FC_GAMMA_R_MEDIATED_PHAGOCYTOSIS
|
90
|
0.5691
|
2.0413
|
0.0017
|
0.0690
|
0.0597
|
REACTOME_ASSEMBLY_OF_COLLAGEN_FIBRILS_AND_OTHER_MULTIMERIC_STRUCTURES
|
61
|
0.6108
|
2.0296
|
0.0018
|
0.0690
|
0.0597
|
REACTOME_BINDING_AND_UPTAKE_OF_LIGANDS_BY_SCAVENGER_RECEPTORS
|
41
|
0.6547
|
2.0240
|
0.0018
|
0.0690
|
0.0597
|
BIOCARTA_CXCR4_PATHWAY
|
19
|
0.7632
|
2.0066
|
0.0019
|
0.0690
|
0.0597
|
REACTOME_ELASTIC_FIBRE_FORMATION
|
43
|
0.6432
|
2.0057
|
0.0018
|
0.0690
|
0.0597
|
WP_MIR5093P_ALTERATION_OF_YAP1ECM_AXIS
|
17
|
0.7822
|
2.0020
|
0.0019
|
0.0690
|
0.0597
|
WP_INFLAMMATORY_RESPONSE_PATHWAY
|
29
|
0.6608
|
1.9161
|
0.0036
|
0.0831
|
0.0720
|
WP_FOCAL_ADHESIONPI3KAKTMTORSIGNALING_PATHWAY
|
299
|
0.4424
|
1.8417
|
0.0015
|
0.0690
|
0.0597
|
WP_IL9_SIGNALING_PATHWAY
|
17
|
0.7136
|
1.8262
|
0.0057
|
0.1050
|
0.0910
|
BIOCARTA_P53HYPOXIA_PATHWAY
|
20
|
0.6473
|
1.7306
|
0.0111
|
0.1462
|
0.1267
|
2.4 GSEA
GSEA (Table 2–3) was employed to identify enriched pathways among the differential gene sets. The results showed that the dataset GSE79768 (Fig. 5A-5E) significantly affected biologically related functions and signaling pathways such as the IL8 CXCR2 pathway, TGFB signaling in thyroid cells for epithelial-mesenchymal transition, Notch 2 activation and transmission of signal to the nucleus, and FCERI-mediated MAPK activation. In addition, analysis of the GSE115574 dataset (Fig. 5F-5J) showed that signaling pathways such as the inflammatory response pathway, focal adhesion PI3KAKTMTOR signaling pathway, IL9 signaling pathway, and P53 hypoxia pathway) were significantly affected.
2.5 Construction of PPI, TF-mRNA, mRNA-miRNA, and mRNA-drug regulatory networks
First, a PPI network for FRDEGs (Fig. 6A) was constructed using the STRING database. Of these, only 11 FRDEGs were found to be associated, including SNAI2, FHL2, ATF3, CYBB, TLR4, GPX2, SMARCA4, METTL14, CAD, CAV1, and VIM.
Second, the TFs associated with FRDEGs (Fig. 6B) were obtained using the ChIPBase database. There were 14 key mRNA genes (SNAI2, FHL2, DIAPH3, ATF3, CYBB, TLR4, GPX2, SMARCA4, HCFC2, METTL14, CAD, CAV1, COPG1, and VIM) and 132 TFs.
The mRNA-miRNA regulatory network (Fig. 6C) was constructed by retrieving the FRDEGs from the StarBase database. Of these, 12 key mRNAs (SNAI2, FHL2, DIAPH3, ATF3, TLR4, SMARCA4, HCFC2, METTL14, CAD, CAV1, COPG1, and VIM) and 77 miRNAs were included.
Finally, potential drugs or molecular compounds for the FRDEGs (Fig. 6D) were identified using the CTD database. Of these, there were 14 key mRNAs (SNAI2, FHL2, DIAPH3, ATF3, CYBB, TLR4, GPX2, SMARCA4, HCFC2, METTL14, CAD, CAV1, COPG1, and VIM) and 76 pharmaceuticals or molecular compounds.
2.6 Dataset validation and ROC analysis
To validate the FRDEGs in the AFib datasets, the difference analysis results of the 14 FRDEGs in the AFib and SR groups were shown in a group comparison scatter plot with the GSE79768 and GSE115574 datasets.
The results of the GSE79768 dataset (Fig. 7A) showed that 10 FRDEGs were statistically significant (p < 0.01 for FHL2, CYBB, and GPX2; p < 0.05 for SNAI2, DIAPH3, ATF3, TLR4, SMARCA4, CAD, and COP1). The rest four genes (HCFC2, METTL14, CAV1, and VIM) were not statistically significant (p ≥ 0.05).
The difference analysis results (Fig. 8A) of the GSE115574 dataset showed that six FRDEGs were statistically significant (p < 0.01 for ATF3; p < 0.05 for DIAPH3, SMARCA4, CAD, CAV1, and COP1 ). The rest eight genes (SNAI2, FHL2, CYBB, TLR4, GPX3, HCFC2, METTL14, and VIM) were not significantly expressed (p ≥ 0.05).
The ROC curves of FRDEGs in the GSE79768 dataset were plotted, and the results are presented (Fig. 7B-7F). The ROC curve showed that the FRDEGs ATF3 (AUC = 0.905) and CAD (AUC = 0.905) in the GSE79768 dataset were highly correlated with the different groups (AUC > 0.9); DIAPH3 (AUC = 0.881), SMARCA4 (AUC = 0.857), and COPG1 (AUC = 0.857) showed a certain correlation with different groups (0.7<AUC<0.9). Similarly, the ROC curves of the FRDEGs in the GSE115574 dataset were plotted, and the results are presented (Fig. 8B-8F). The ROC curve showed that the expression levels of FRDEGs DIAPH3 (AUC = 0.724), ATF3 (AUC = 0.810), SMARCA4 (AUC = 0.752), CAD (AUC = 0.757), and COPG1 (AUC = 0.733) in GSE115574 dataset were correlated to different groups (0.7 < AUC < 0.9).
2.7 qRT-PCR
We used RT-qPCR to analyze levels of the DEGs in the AF group. The qRT-PCR analyses revealed that the expression levels of COPG1 elevated in atrial tissue of AF rats, and the expression level of ATF3 decreased (Fig. 9). The PCR results were consistent with the results obtained by bioinformatics analysis.