Epithelial-Mesenchymal Transition-Related lncRNAs And SNAI2 Are Potential Biomarkers in Coronary Artery Disease

Increasing evidence suggests that epithelial-mesenchymal transformation (EMT) is critical in the development of inammatory response, atherosclerosis, and coronary artery disease (CAD). However, landscapes of EMT-related lncRNAs and their target genes have not been fully established in CAD. LncRNA and mRNA expression proles obtained from Gene Expression Omnibus (GEO) database were used to identify the differentially expressed mRNAs (DEGs) and lncRNAs (DElncRNAs) between CAD and normal samples. Based on Pearson correlation analysis to identify the EMT-related lncRNAs, the optimal features were identied by receiver operating characteristic (ROC), the least absolute shrinkage and selection operator (LASSO) regression, Support Vector Machine Reverse Feature Elimination (SVM-RFE) algorithms, and logistic regression models were constructed aiming to distinguish CAD from normal samples. The cis and trans-regulatory networks were constructed based on EMT-related lncRNAs. We further estimated the inltration of the immune cells in CAD patients with the CIBERSORT algorithm, and the correlation between key genes and inltrating immune cells was analyzed. our understanding of the molecular mechanism between EMT and CAD.


Introduction
Coronary artery disease (CAD) is a common public health problem, mainly occurring in people over 45 years of age. CAD is now the leading cause of death in the United States, accounting for one in six deaths alone [1]. The American Heart Association has said that cardiovascular disease causes more than 17.3 Therefore, based on previous research, we constructed two machine learning algorithms: LASSO regression algorithm and SVM-RFE algorithm to screen out EMT-related diagnostic lncRNAs in CAD patients. Meanwhile, we construct cis-trans regulatory networks based on EMT-related lncRNAs and explore the potential EMT gene of related molecules in the cis-trans network and the target drugs and structures. We also investigated the correlation between EMT-related diagnostic signatures and immune cell subsets by immune analysis. Through bioinformatics methods, in-depth excavation of the EMT genes in promoting role in coronary atherosclerosis and the potential signal pathways and molecular mechanisms, late for the prevention and treatment of CAD can provide a new train of thought and targets.

Methods And Materials
Data collection In our study, the microarray gene expression pro ling data of CAD was downloaded from the GEO (https://www.ncbi.nlm.nih.gov/geo/) database with accession number GSE113079. 48 normal samples and 93 CAD samples were included in the GSE113079 dataset. 200 EMT-related genes were obtained from the Molecular Signatures Database (MsigDB, http://www.broad.mit.edu/gsea/msigdb/). Besides, 1639 genes related to TFs were acquired from the database of The Human Transcription Factors (TFBS, http://tfbsdb.systemsbiology.net/).

Differentially expressed analysis
The limma package in R was used to identify the DElncRNAs and DEGs between CAD and normal samples (Supplementary Table 1-2). The lncRNAs/mRNAs met the selection standards of |log 2 (Fold change)| > 1.5 and false discovery rate (FDR) < 0.01 were considered as DElncRNAs/DEGs for further study.

Correlation analysis
By mating the listed 200 EMT-related genes in the MsigDB database, differentially expressed EMT genes (DE-EMTs) for CAD were identi ed. Then, Pearson correlation analysis was operated between the harvested DE-EMTs and DElncRNAs expression data in samples to identify the EMT-related lncRNAs according to the correlation coe cient and P values (|Cor| > 0.8 and P < 0.05) (Supplementary Table 3).

Diagnostic value of EMT-lncRNAs and SNAI2
To explore the diagnostic ability of EMT-lncRNAs mentioned above, ROC analysis was rst performed using the R package pROC, and the EMT-related lncRNAs with AUC > 0.95 were screened for further study. After ltration of EMT-related lncRNAs, candidate diagnostic lncRNAs for CAD were selected via integrated analysis of two algorithms consisting of LASSO and SVM-RFE. Logistic regression was performed on diagnostic lncRNAs and the SNAI2 gene, respectively, to construct a logistic regression diagnostic model, and the bias residual diagram was drawn ( Supplementary Fig. 1A-B). 5-fold cross validation was used to evaluate the performance of diagnostic signature. Moreover, the diagnostic value of EMT-related lncRNAs and SNAI2 were assessed by ROC curve analysis using the pROC package in the R language.

EMT-related lncRNAs categorization
Based on modi cations of the previous classi cation [19], we classi ed rat lncRNAs according to their gene positions related to the most proximal protein-coding genes. Firstly, the lncRNA genes were regarded as intergenic and genic based on whether they intersect a protein-coding gene. Furthermore, intergenic lncRNAs were categorized as two groups depended on their transcribed from the same or opposite strand: convergent (IC) and divergent (ID). Genic lncRNAs were separated into genic exonic (genic exonic same strand (GES) and genic exonic antisense (GEAS)), genic intronic (genic intronic same strand (GIS) and genic intronic antisense (GIAS)), or overlapping (genic overlapping same strand (GOS) and genic overlapping antisense (GOAS)) based on them overlapped with the exons or introns of a protein-coding gene.

The regulatory mechanisms of diagnostic EMT-lncRNAs
It is reported that lncRNAs regulated transcription of their nearby genes via acting in cis-and tansmanners. For the cis-regulation manner, we rst selected the genes located on the same chromosome within a 300 kb region upstream or downstream of the lncRNAs. Subsequently, the Pearson analysis method was performed to analyze the correlation between the harvested lncRNAs and their corresponding genes under the selection criteria of |Cor| > 0.3 and P < 0.05.
For trans prediction, we focused on that lncRNAs might regulate the expression levels of TFs via the trans manner. After selecting the genes correlated with lncRNAs by the Pearson method (|Cor| > 0.8 and P < 0.05), we further overlapped these genes with identi ed DEGs and TFs to obtain trans-regulated genes. A lncRNA-mRNA network that included EMT-lncRNAs, cis-and trans-regulated genes was constructed and visualized by the Cytoscape software.
Functional enrichment analysis GO annotation and KEGG pathway analyses were employed on the DE-EMTs, cis-regulated genes, transregulated genes, and SNAI2 regulated genes of CAD, respectively, to explore the latent biological functions and pathways. Besides, the optional pathways related to CAD were predicted by the Comparative Toxicogenomics Database (CTD, http://ctdbase.org). Genes related to CAD were predicted by the DisGeNET database( https://www.disgenet.org/home/). The KEGG pathways both in the CAD database and KEGG analysis were introduced into the lncRNA-mRNA network to establish a lncRNA-mRNA-pathway network for CAD.

Immunity analysis and its correlation with key genes
We used CIBERSORT [20] for immune in ltration. R script downloaded from CIBERSORT website (https://cibersort.stanford.edu/). After obtaining the immune cell expression matrix according to the instructions on the CIBERSORT website, we used the "ggplot2" software package to create a cumulative histogram that visually showed the proportion of 22 immune cell in ltrates in CAD patients. We also used the "vioplot" package to draw violin plots showing differences in expression of 22 in ltrating immune cells. "corrplot" software package in R software was used to calculate Pearson correlation coe cients among immune cells, and the results were displayed by correlation heat map. Pearson correlation coe cients and p-value between identi ed key genes and in ltrating immune cells were calculated by "cor" and "Hmisc" software packages and then visualized by the "ggcorrplot" software package. In addition, single-sample gene set Enrichment Analysis (ssGSEA) [21] and QuanTIseq [22] algorithms were also used compared to assess cellular components between the high SNAI2 gene expression group and low SNAI2 gene expression group. The differences in the immune response under different algorithms were uncovered using a Heatmap.
The drug-gene prediction The genes cis-and trans-regulated by diagnostic lncRNAs were supposed to be the promising drug targets for searching drugs through the Drug-Gene Interaction database (DGIdb, https://dgidb.genome.wustl.edu/) that contained the drug-gene interaction information of several databases [23]. The drug-gene network was visualized by the Cytoscape tool.

Statistical analysis
The subcellular localization of diagnostic EMT-related lncRNAs was predicted by the LncLocator online tool [24]. The clusterPro ler package in R was utilized to perform GO and KEGG analyses. P-value < 0.05 was considered as statistically signi cant.

Identi cation of EMT-related genes
We performed the differentially expressed analysis on the GSE113079 dataset. As shown in Fig. 1A, 5955 DElncRNAs were identi ed between CAD and normal samples under |log 2 (Fold change)| and FDR < 0.01 with 3067 were upregulated and 2888 downregulated. Meanwhile, we screened 2868 DEGs between two groups, including 1540 upregulated and 1328 downregulated DEMs (Fig. 1C). The expressed levels of DElncRNAs and DEGs were shown in the heatmap plot and displayed in Fig. 1B, D, respectively. The 32 DEGs related to EMT were generated by overlapping 200 EMT genes in the MsigDB database and preselected DE-EMTs, in which 21 were upregulated, and 11 were downregulated ( Fig. 1E-F).
Functional enrichment analysis suggested DEGs related to EMT were enriched in 313 GO terms and 24 KEGG pathways. The main GO terms with signi cant enrichment involved with these genes in CAD were the 'response to wounding,' 'wound healing,' 'contractile actin lament bundle,' and ' bronectin binding' (Fig. 1G). In KEGG analysis, the DEGs related to EMT were mostly associated with the 'PI3K-Akt signaling pathway', 'focal adhesion,' and 'ECM-receptor interaction' (Fig. 1H), which indicated that these differential genes played a considerable role in the occurrence of EMT and heart development.
Construction of an EMT-related lncRNAs diagnostic signature for CAD To further detect the diagnostic ability of these EMT-related lncRNAs, the AUC value of each EMT-related lncRNAs was analyzed. 223 EMT-related lncRNAs were screened with an AUC value above 0.95 (Supplementary Table. 4). LASSO regression analysis SVM-RFE algorithm was used to identify the optimal diagnostic lncRNAs in the GSE113079 dataset and establish the risk signature for CAD. 16 EMTrelated lncRNAs were screened via the LASSO analysis, which intersected with 34 EMT-related lncRNAs obtained from the SVM-RFE algorithm to identify 11 diagnostic lncRNAs for CAD (Supplementary Table. 5, Fig. 3A-E). After annotating the diagnostic lncRNAs using the Rsubread package in R, we obtained eight lncRNAs that were used to constructed a diagnostic signature for CAD (Table 1), showing accuracy and speci city for the diagnosis of CAD (AUC = 1) (Fig. 3F). Besides, the AUC value of each diagnostic lncRNAs was greater than 0.95, which exhibited a better ability to distinguish CAD patients from normal (Fig. 3G). Subcellular localization of each lncRNA determines the regulatory models. To investigate the subcellular localization of the diagnostic lncRNAs, we assessed LncLocator online platforms. We uncovered that these diagnostic lncRNAs were mainly located in the cytosol and cytoplasm (Supplementary Fig. 2)

Establishment of cis and trans-regulatory network
Previous studies indicated that lncRNAs regulated gene expression via local (cis) and long-distance (trans) mechanisms [25]. In the present study, we identi ed seven diagnostic lncRNAs regulated their nearby genes via the cis-regulatory manner, except RP11-103H7.3 (Table 2). Among them, only CTD-2089N3.3 were signi cantly correlated with their corresponding gene EMB via Pearson analysis under |Cor| > 0.3 and P-value < 0.05 (Fig. 4A). Based on the median expression level of EMB, we divided the CAD patients in the GSE113079 dataset into the high-expressed EMB group and low-expressed EMB group.
By combining 685 genes correlated with lncRNAs with DEGs and identi ed TFs, 33 genes were identi ed to be regulated by diagnostic lncRNAs via trans manner, in which SNAI2 was founded to be a differentially expressed EMT gene ( Fig. 4B-C). Then, a lncRNA-mRNA regulatory network was constructed that contained diagnostic lncRNAs, cis-and trans-regulated genes, which consisted of 42 nodes and 93 edges (Fig. 4D). These genes in the regulatory network were mainly involved in nervous development and vitamin metabolism by GO analysis (Supplementary Fig. 4A). Combining the pathways related to these genes identi ed with the KEGG analysis and related to CAD development in the CTD database, 'maturity onset diabetes of the young' and 'transcriptional misregulation in cancer' were discovered ( Fig. 4E-F, Table  3). Hence, these two KEGG pathways were introduced into the lncRNA-mRNA regulatory network to establish a lncRNA-mRNA-pathway network for CAD that included 46 nodes and 98 edges ( Supplementary Fig. 4B).

Prediction of regulatory genes of SNAI2
Based on the above results, we uncovered that SNAI2 was found to be a DE-EMT and TFs among all genes regulated by diagnostic lncRNAs. In our study, SNAI2 is obviously higher expressed in CAD groups than normal groups (p = 5.8e-15; Fig. 5A). Considering the importance of SNAI2, we also detect the diagnostic ability of SNAI2 in CAD patients. The AUC value of SNAI2 was 0.902 (Fig. 5B). In addition, we used ve-fold cross validation to evaluate the reliability of the SNAI2 gene. Firstly, we randomly divided the samples into ve parts, of which four parts were used as training sets to build the logistic regression model, and the rest were used to verify the model. This process is then repeated ve times to reduce errors and improve the sensitivity of the model. The AUC values of the ve models were 0.9479, 0.9144, 0.9391, 0.7692, and 0.8766, respectively, indicating that the models had good explanatory power (Fig. 5C). Besides, GSEA was performed to investigate the latent biological functions. 'Calcium signaling pathway,' 'linoleic acid metabolism,' 'neuroactive ligand receptor interaction,' and 'olfactory transduction' were mainly associated with the high-expressed SNAI2 group. 'RNA degradation,' 'splicesome,' 'fatty acid metabolism,' and 'histone metabolism' were involved in the low-expressed SNAI2 group ( Supplementary  Fig. 5). Moreover, we overlapped 234 genes regulated by SNAI2 acquired from the TFBS database and 1576 genes related to CAD acquired from the DisGeNET database to obtain 21 genes regulated by SNAI2 for CAD (Fig. 5D). Functional enrichment analysis determined that the harvested 21 genes were concerted on the 'insulin secretion,' 'peptide hormone secretion,' 'long-chain fatty acid biosynthetic process' (Fig. 5E). There are no pathways detected by KEGG analysis.
Prediction of the target drugs of genes in the cis and transregulatory network Next, the target drugs of genes regulated by diagnostic lncRNAs were predicted by the DGIdb database. Through DGIdb prediction, a total of 483 drug-gene pairs were identi ed, and a target-drug network for CAD was constructed, including ve genes and 476 drugs. 459 drugs interacted with VDR, which might be promising to treat patients with CAD ( Supplementary Fig. 6A). The structures of these drugs were illustrated in Supplementary Fig. 6B-I.

Immune analysis of EMT-lncRNAs and SNAI2
Enrichment analysis showed that DE-EMT gene was enriched in in ammatory response-related pathways. Therefore, we evaluated the type and fraction of immune cell in ltration between CAD patients and normal samples in the dataset using the CIBERSORT algorithm. The relative proportion of immune cell subtypes was shown in the cumulative histogram (Fig. 6A). Our results found an apparent proportion of CD8 T cells, NK cells activated, and monocytes. Moreover, the in ltration of CD8 T cells and NK cell activated were decreased, and the in ltration of monocytes was increased in CAD patients (Fig. 6B). By principal component analysis (PCA), immune cell fractions in CAD patients and normal controls showed intergroup bias and individual differences (Fig. 6C). In the correlation heatmap (Fig. 6D), we found that CD8 T cells were negatively correlated with monocytes and macrophages M0, and positively correlated with NK cells activated. It is consistent with the correlation between seven EMT-related lncRNAs and the immune cells we found, except lncRNA AC109460.4 (Fig. 7A). In addition, we also conducted an immune analysis of SNAI2. Then, we divided the samples into high and low groups according to the expression level of SNAI2. It was found that in the high expression level group, the in ltration of monocytes was decreased. In contrast, the in ltration of NK cell activated and CD8 T cells were increased, which was similar to immune cell in ltration in CAD patients (Fig. 7B). The heatmap of immune cell compositions based on CIBERSORT, quanTIseq, ssGSEA algorithms is shown in Fig. 8. It was found that CD8 T cells, monocytes, and NK cells activated had similar immune cell in ltration trends in the CAD and SNAI2 gene high expression groups.

Discussion
EMT plays a critical physiological and pathological role in developing the cardiovascular system, vascular tissue remodeling, and heart valve disease during the embryonic period[26]. However, more research focused on the impact of the EMT in tumor development and treatment. In contrast, few studies have explored the diagnostic value of EMT-related genes or lncRNAs in CAD. Hence, exploring diagnostic biomarkers of EMT genes/lncRNAs in CAD is urgent.
Our analyses uncovered 32 EMT-related DEGs in CAD. KEGG pathway analysis of these DE-EMTs was mainly enriched in the PI3K/Akt signaling pathway. Several reports have shown that PI3K/Akt pathway is a canonical EMT signaling pathway [27,28]. Meanwhile, we found this signaling pathway plays an essential role in the CAD. A recent study indicated that miRNA-26a-5p activated the PI3K/Akt pathway by targeting Phosphatase and Tensin Homolog (PTEN) and affected the proliferation and apoptosis of endothelial cells isolated from CAD mice [29]. A comparative study also reported that miR-26a-5p could activate the PI3K/Akt signaling pathway through inhibition of PTEN, thereby protecting against myocardial defect/reperfusion injury [30]. These studies have con rmed that activating the PI3K/Akt signaling pathway can prevent myocardial ischemia-reperfusion in animal models. Other studies have also suggested that regulation of the PI3K/Akt signaling pathway plays a vital role in inhibiting myocardial brosis, apoptosis, and the in ammatory response [31,32].
In the present study, we performed a co-expression analysis between EMT genes and DElncRNAs through paired lncRNA and mRNA expression data in CAD patients from GEO. Eight differently expressed EMTrelated lncRNAs were found to be diagnosis factors for CAD patients. After a literature review, we found no research had been conducted about the mechanisms of the eight lncRNAs except LINC02747.
Previous studies have reported that LINC02747 can upregulate the expression of TFE3 by absorbing miR-608 and ultimately promote the proliferation of renal cell cancer cells (ccRCC) [33]. Gu et al. indicated that miR-608 exerts anti-in ammatory effects by targeting ELANE in monocytes [34]. Our results showed that monocytes were more expressed in the CAD group, so whether the regulation of LINC02747-mir608-ELANE might achieve the reversal of in ammatory response in CAD patients. Other seven EMT-related lncRNAs have not been reported in relevant studies, and reports on how lncRNAs interact with EMT genes have been rarer. However, many "cis" and "trans" genes are involved in the formation and development of CAD in the cis-trans regulatory network. For example, EMB, as a "cis" gene, was enriched in the mTOR signaling pathway in our GSEA analysis. This pathway is closely associated with atherosclerosis, and the pro-in ammatory response of monocytes in CAD requires activation of mTOR [35]. Among "trans" genes, some studies have reported that VDR gene polymorphisms lead to the development and formation of CAD by affecting changes in serum levels of 25(OH) vitamin D. [36,37]. Previous study reported VDR in regulating in ammation through inhibiting the NF-ĸB pathway and activating autophagy [38]. EBF4 gene promotes the elevation of Cu and leads to the progression of CAD by affecting copper related DNA methylation sites [39]. CTCF gene is essential for cardiogenesis and inhibit cardiomyocytes apoptosis, and can be applied as a therapeutic target for the treatment of heart failure in future. [40,41]. FLI1 gene is also reported to be closely related to immune dysfunction and platelet disorders [42]. Although the lack of direct support in the literature, we speculated that these cis-trans genes, under the regulation of lncRNA, affect the formation and development of CAD through immune microenvironment, cell apoptosis, platelet dysfunction and other ways. So far, there has been no study on the role of EMT-related lncRNA in CAD diagnosis. These ndings may provide valuable insights into the future diagnosis and treatment of CAD.
The presence of immune cells in the infarct area is vital to initiating the repair process of injured heart tissue. Temporal and spatial regulation of in ammation after infarction is crucial [43,44]. We evaluate the type and fraction of immune cell in ltration between CAD patients and normal samples in the dataset using the CIBERSORT algorithm. Our results found CD8 T cells and NK cells share a decreased in ltration, and the in ltration of monocytes was increased in CAD patients, which was similar to the previous results [45][46][47]. In this GEO dataset, CD8 T cells and NK cells are favorable factors for preventing CAD, and monocytes likely promote the occurrence of CAD. Previous studies have suggested that the imbalance of immune regulation is an essential factor in promoting atherosclerosis, heart failure, and chronic kidney disease by monocytes cells [44]. CD8 T cells play a dual role in atherosclerosis. On the one hand, CD8 T cells can secret many in ammatory cytokines to accelerate the in ammatory response and increase the instability of atherosclerotic plaques. On the other hand, cytotoxic activity against antigenpresenting cells and the presence of regulatory CD8 T cell subsets could suppress immunity and limit atherosclerosis[48]. Ong et al. suggested that NK cells appear to protect the development of cardiac brosis by preventing the accumulation of speci c in ammatory groups in the heart and directly restricting collagen formation in cardiac broblasts [49]. Although the results of our study are similar to these researches, the mechanism of the immune system is still very complex, and some results in the immunotherapy of CAD are not ideal. We need a lot of clinical studies to demonstrate the underlying mechanism. Besides, we also found that except AC109460.4, the other seven lncRNAs related to EMT were signi cantly negatively correlated with CD8 T cell and NK cell and positively correlated with Treg and monocytes. The results of AC109460.4 were just the opposite. The association between these lncRNAs and the innate immune system is still unclear. More in vivo and in vitro studies are needed to explain the interaction mechanism between these lncRNAs and immune cells in CAD.
It is generally believed that lncRNAs can act in "trans" to regulate TFs mediated chromatin remodeling and transcription [50]. These lncRNAs recruit protein factors to enhancer and regulate enhancer activity [51]. We constructed cis and trans-regulatory networks based on these eight signatures. In the trans-regulatory network, we obtained 33 differentially expressed TF genes. The most surprising discovery was the screening of SNAI2, an EMT-TF gene (the gene coding product was the transcription factor Snai2). Our results indicated that SNAI2 was not only signi cantly highly expressed in CAD patients but also strongly positively correlated with LINC01775 and CTD-2089N3.3. The ROC curve showed that the SNAI2 could be a potential biomarker for diagnosing CAD. As a classic EMT-TF gene, SNAI2 has recently been shown to be involved in a broader range of biological processes, including tumor metastasis, heart development, cell differentiation, vascular remodeling, and DNA damage repair [52][53][54].
Previous studies have reported that the deletion of protein arginine methyltransferase 1 (PRMT1) leads to the accumulation of p53, and enhancing the degradation of SNAI2 can limit the formation of cardiac broblasts, coronary smooth muscle cells, and pericytes [55]. Meanwhile, Cooley et al. reported that, by grafting mouse veins to the femoral artery in mice to simulate human coronary artery bypass grafting (CABG), the results showed that TGF-β/Smad2/3-Snai2 mediated EMT plays a crucial role in venous graft vessel remodeling [56]. These studies have indicated that high expression of SNAI2 can promote the formation of vascular endothelium to EMT and vascular remodeling, which is one of the vital factors in the formation of CAD. At present, the role of SNAI2 in CAD has not been reported, several studies have proven that the vascular endothelial EMT process is involved in atherosclerosis, post-stent stenosis, pulmonary hypertension, and coronary artery remodeling [57][58][59]. Additionally, the role of EMT can be seen in a range of cell types involved in immunity, such as lymphocytes, NK cells, and myeloid cells, which contribute to in ammatory responses in diverse pathophysiological processes. Ricciardi et al. have reported a decreased viability and proliferation of NK cells and T cells after co-culture with cancer cell lines in which EMT had been induced[60]. In our study, SNAI2 correlated with in ltration of monocytes, CD8 T cells, and NK cells activated. Previous studies have suggested that SNAI2 deletion in mice leads to impaired development of the T-lymphatic system [61]. Subsequent studies also con rmed that Snai2 plays a vital in uence in regulating CD8 T cells and targets genes with functions for T cells [62]. Furthermore, our results indicated that the difference in these immune cell in ltrations in the SNAI2 high expression group was similar to the results of CAD patients. These immune cells have been researched to play a role in the formation, erosion, and rupture of coronary plaques [63,64]. In summary, we inferred that SNAI2 might have signi cant roles in the occurrence of CAD by regulating innate and adaptive immunity through these immune cells. To con rm our conclusion, more experimental mechanistic research should be carried out in future studies.
Our study should acknowledge some limitations. First, the expression levels of critical lncRNAs in CAD were not veri ed in clinical samples. Secondly, these EMT-related lncRNAs were investigated in datasets with no access to individual patients' characteristics; thus, we cannot adjust the ROC curve for traditional cardiovascular risk factors. A prospective cohort recruiting CAD patients is needed to con rm the predictive value of EMT-related lncRNAs. Thirdly, the molecular function details of SNAI2 and EMT-related lncRNAs in the progression of CAD have not been further studied. Therefore, molecular biological experiments and ow cytometry analysis are required to validate these ndings, and another external validation based on a larger sample is needed.

Conclusions
In conclusion, this comprehensive bioinformatic analysis revealed that SNAI2 and EMT-related lncRNAs could be reliable biomarkers for diagnosing CAD and use decision-making in the treatment of CAD patients. At the same time, based on the eight EMT-related lncRNAs, we constructed the cis and transregulatory networks of CAD. Furthermore, the immune analysis suggested that these biomarkers were closely related to immune cells and CAD. These ndings provide references for clinicians to understand the molecular mechanism of interaction between CAD and EMT and develop individualized treatment for CAD patients.