Comparative molecular analysis of oral submucous fibrosis and other organ fibrosis based on weighted gene co-expression network analysis

Abstract

Objective

There is currently a lack of economic and suitable animal models that can accurately recapitulate the oral submucous fibrosis (OSF) disease state for indepth study. This is one of the primary reasons for the limited therapeutic methods available for OSF. Based on the underlying logic of pan-cancer analysis, this study systematically compares OSF and the other four types of organ fibrosis from the aspects of molecules, signaling pathways, biological processes, etc. A comprehensive analysis of the similarities and differences between OSF and other organ fibrosis is helpful for researchers to discover some general rules of fibrosis disease and may provide new ideas for studying OSF.

Methods

Microarray data of the GSE64216, GSE76882, GSE171294, GSE92592, and GSE90051 datasets were downloaded from GEO. Differentially expressed mRNAs (DEmRNAs) of each type of fibrosis were identified by Limma package. Weighted gene co-expression network analysis (WGCNA) was used to identify each type of fibrosis-related module. The similarities and differences of each fibrosis-related-module genes were analyzed by function and pathway enrichment analysis.

Results

A total of 6 057, 10 910, 27 990, 10 480, and 4 801 DEmRNAs were identified in OSF, kidney intestinal fibrosis (KIF), liver fibrosis (LF), idiopathic pulmonary fibrosis (IPF), and skin fibrosis (SF), respectively. By using WGCNA, each type of fibrosis-related module was identified. The co-expression networks for each type of fibrosis were constructed respectively. Except that KIF and LF have 5 common hub genes, other fibrotic diseases have no common hub genes with each other. The common pathways of OSF, KIF, LF, IPF, and SF mainly focus on immune-related pathways.

Conclusion

OSF and the other 4 types of fibrotic diseases are tissue- and organ-specific at the molecular level, but they share many common signaling pathways and biological processes, mainly in inflammation and immunity.

Fibrosis is well known as a late outcome of many chronic inflammatory diseases. When the tissue injury is severe and persistent or when the regulation of wound healing becomes abnormal, normal tissue repair would evolve into a progressively irreversible fibrotic response^[1]. Fibrosis is characterized by the deposition of extracellular matrix (ECM) components as a result of the tissue repair response. Fibrosis can lead to permanent scarring, organ malfunction, and even death. Moreover, studies have indicated that fibrosis has cancerous potential and can influence tumor invasion and metastasis^[2-3]. It has been estimated that up to 45% of all deaths in developed countries are attributed to fibrotic tissue reactions^[4]. Due to the difficulty and financial burden of the treatment, fibrosis is increasingly recognized as one of today’s major healthcare challenges^[5-6].

Many trigger factors, including genetic factors; persistent infection; chronic inflammation; repeated tissue damage; recurrent exposure to toxins, irritants, or smoke; myocardial infarction; hypertension; poorly controlled diabetes, and minor human leukocyte antigen mismatches in transplants contribute to the development of fibrosis^[7-8]. Fibrosis can occur in almost any tissue or organ in the whole body. However, the trigger factors of fibrosis in different tissues or organs are not the same (Supplementary Table 1, https://doi.org/10.11817/j.issn.1672-7347.2022.220452T1). For example, areca nut chewing is considered as one of the most significant trigger factors for oral submucous fibrosis (OSF). Meanwhile, consumption of spicy foods, human papillomavirus infection, and immune factors may also be trigger factors^[9-10]. Chronic kidney diseases, including diabetic nephropathy and glomerular disease, trigger a complex program of cellular and extracellular matrix responses inducing renal fibrosis^[11]. Hepatitis B virus (HBV), hepatitis C virus (HCV), excess alcohol consumption, and non-alcoholic fatty liver disease (NAFLD) are the primary trigger factors related to liver fibrosis (LF)^[12]. Particulate inhalation, including smoking, metal and wood dust, stone, and silica, is associated with idiopathic pulmonary fibrosis (IPF)^[13]. Tissue injury and chronic inflammation are the primary trigger factors for skin fibrosis (SF)^[14].

The development of anti-fibrosis drugs is minimal and has been limited to a single organ system, and the progress continues slowly. One potential strategy to overcome this challenge is to explore the molecular mechanisms and identify common mechanisms and core pathways that have central pathophysiological relevance across different fibrotic diseases. Therefore, this study focused on the following five fibrotic diseases: OSF, kidney interstitial fibrosis (KIF), LF, IPF, and SF. By performing gene-, pathway-, and module-level comparisons, the common and different molecules and functions in fibrosis arising from different organs were investigated. A thorough understanding of the common principles of molecule mechanisms of fibrous formation in different organs is the basis for developing improved approaches, including biomarkers and novel anti-fibrous therapies.

In addition, different from the mature animal models of KIF, LF, IPF, and SF, there is currently a lack of economic and suitable animal models that can accurately recapitulate the OSF disease state for in-depth study. This is one of the primary reasons for the limited therapeutic methods available for OSF^[15]. A comprehensive analysis of the similarities and differences between OSF and other organ fibrosis may provide new ideas for studying OSF.

1. Materials and methods

1.1. Data collection

We searched fibrosis-related gene expression profile data in Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo). Study types were restricted to “Expression profiling by array” or “Expression profiling by high throughput sequencing”. Species were restricted to “Homo sapiens”. The attribute name was restricted to “tissue”. The deadline was December 2021. Finally, the raw data from the GSE64216, GSE76882, GSE171294, GSE92592, and GSE90051 datasets were downloaded from GEO. The basic information of the five datasets can be seen in Supplementary Table 2 (https://doi.org/10.11817/j.issn.1672-7347.2022.220452T2).

1.2. Identification of differentially expressed genes

The normalized “BetweenArrays” of the “limma” package was used to read the microarray and normalize the expression data^[16]. After the raw data were transformed into the expression files, a t-test (Student’s t-test) was used to screen the differentially expressed genes (DEGs) in “R”. Selection condition: t-test P<0.05.

1.3. Weighted gene co-expression network analysis

The “weighted gene co-expression network analysis (WGCNA)” package in “R”^[17] was used to build a weighted gene co-expression network for each fibrosis type. The gene expression profile matrix was input and preprocessed by normalization and transformation to remove the batch effect.

The adjacency between genes was calculated, and the similarity between genes was calculated based on the adjacency. For each candidate soft threshold “power” (1-30), the scale-free network evaluation parameter R ² and the average connectivity degree of different modules in a given model were calculated by the “pickSoftThreshold” function. When the independence R ² is greater than 0.8 and the average connectivity is high, select an appropriate power value. Secondly, the power function was used to power the correlation between genes, and the weighted correlation coefficient was obtained to form the adjacency matrix. The similarity between genes is calculated based on adjacency degree, and then the dissimilarity between nodes is deduced. The adjacency matrix is then converted to the topological overlap matrix (TOM) by the “blockwiseModules” function. The dynamic cut tree clustering algorithm is used to initially identify network modules. When the difference between modules is less than the threshold, the initially constructed modules are merged. The minimum module size was set to 300. Each module is assigned a unique color label and is displayed as a branch in the cluster tree. After determining the gene modules according to the dynamic shear method, the gene modules were analyzed again, and the characteristic vector values of each module were calculated in turn. Clustering analysis was performed on the modules, and the modules close to each other were merged into new modules.

The correlation between modules and specific traits or phenotypes was analyzed. By conducting WGCNA for each fibrosis type, the module most significantly positively correlated with each fibrosis type was obtained respectively. We defined the module as the fibrosis-related module.

1.4. Construction of co-expression network and hub genes identification

Co-expression gene networks were constructed for the genes in fibrosis-related modules of each fibrosis type. Hub genes were a series of genes with the highest degree of connectivity in a gene network and determined the characteristics of a network. Connectivity degree was used to node analyze the interaction network. In this study, we used the scale-free nature of the network to identify the hub genes. The fibrosis score of each hub gene was performed by using the Comparative Toxicogenomics Database^[18] (CTD;http://ctdbase.org/). The higher the score, the more closely associated with fibrosis.

1.5. Function and pathway enrichment analysis of genes in each fibrosis-related module

To explore the correlation and specificity of different tissue fibrosis, Gene Ontology (GO) and Kyoto Encylopedia of Genes and Genomes (KEGG) analyses were carried out for the genes in the fibrosis-related module corresponding to each fibrosis type. The identified candidate genes were enriched for terms in the gene biological process (BP), molecular function (MF), and cellular component (CC) categories. KEGG analysis was used to illustrate the potential functions and pathways of candidate genes. The above data were analyzed by using Fisher’s exact test (P<0.05). The smaller the P-value was, the greater the association between the item and the gene input; most of the genes in the set have the descriptive function corresponding to the item.

2. Results

2.1. Workflow

The workflow is shown in Supplementary Figure 1 (https://doi.org/10.11817/j.issn.1672-7347.2022.220452F1). DEG identification was conducted. WGCNA for the DEGs of each fibrosis type was performed. Co-expression networks for each fibrosis-related module were constructed, and hub genes for each fibrosis type were identified. GO and KEGG analyses were conducted to mine the correlation and specificity of different tissue/organ fibrosis.

Figure 1. Co-expression network construction and common network components identification A: Co-expression network of OSF. B: Co-expression network of KIF. C: Co-expression network of LF. D: Co-expression network of IPF. E: Co-expression network of SF. F: Identification of common genes in each type of fibrosis. G: These genes were scored for fibrosis by using the Comparative Toxicogenomics Database. OSF: Oral submucous fibrosis; KIF: Kidney intestinal fibrosis; LF: Liver fibrosis; IPF: Idiopathic pulmonary fibrosis; SF: Skin fibrosis.

2.2. Screening of differentially expressed mRNA

Differentially expressed mRNA (DEmRNAs) for each fibrosis type were obtained by statistical tests. In GSE64216, 4 OSF and 2 normal samples were compared, and 6 057 DEmRNAs were obtained. In GSE76882, 135 KIF and 26 normal samples were compared, and 10 910 DEmRNAs were obtained. In GSE171294, 4 LF and 4 normal samples were compared, and 27 990 DEmRNAs were obtained. In GSE92592, 20 IPF and 19 normal samples were compared, and 10 480 DEmRNAs were obtained. In GSE90051, 7 SF and 7 normal samples were compared, and 4 801 DEmRNAs were obtained.

2.3. Weighted gene co-expression network analysis and fibrosis-related modules detection

DEmRNAs of each fibrosis type were subjected to the WGCNA. Eight groups of co-expressed genes (termed “modules”) were identified in OSF. Four, three, six, and six modules were identified in KIF, LF, IPF, and SF, respectively. Each module was marked with a unique color under the cluster tree (Supplementary Figure 2A, 2C, 2E, 2G, 2I, https://doi.org/10.11817/j.issn.1672-7347.2022.220452F2). We regressed the above module genes on fibrosis-related status to identify the module associated with each fibrosis-related status. We identified the most significant positive correlation module for each type of fibrosis and named the OSF-related module, KIF-related module, LF-related module, IPF-related module, and SF-related module, respectively (Supplementary Figure 2B, 2D, 2F, 2H, 2J, https://doi.org/10.11817/j.issn.1672-7347.2022.220452F2).

Figure 2. KEGG pathway analyses for each type of fibrosis A: KEGG bubble chart of OSF; B: KEGG bubble chart of KIF; C: KEGG bubble chart of LF; D: KEGG bubble chart of IPF; E: KEGG bubble chart of SF; F: Identification of common and specific signaling pathways in each type of fibrosis. KEGG: Kyoto Encylopedia of Genes and Genome; OSF: Oral submucous fibrosis; KIF: Kidney intestinal fibrosis; LF: Liver fibrosis; IPF: Idiopathic pulmonary fibrosis; SF: Skin fibrosis.

2.4. Construction of co-expression network and common network components identification

We selected genes in the OSF-related module, KIF-related module, LF-related module, IPF-related module, and SF-related module to construct a co-expression network by using Cytoscape 3.3.0, respectively (Figure 1A-1E). By integrating the results of the co-expression networks of the five fibrotic diseases, we obtained common and specific genes of fibrosis diseases. We can see that OSF shares some DEmRNAs with KIF, LF, and IPF, including CD86, CYBB, FCN1, HLA-DPB1, and IRF8 (Figure 1F). However, we found that SF and the other four types of fibrosis have no DEmRNAs in common. We used CTD to score these DEmRNAs for fibrosis and found that each had a score greater than 30 (Figure 1G), indicating these genes were closely related to fibrosis.

Further, we identified the top 10% of genes in each co-expression network as hub genes. OSF does not share a hub gene with the other four types of fibrosis. In addition, except that KIF and LF have five hub genes in common, including CD2, CORO1A, CD48, CD53, and RAC2, other types of fibrosis do not share hub genes (Supplementary Figure 3, https://doi.org/10.11817/j.issn.1672-7347.2022.220452F3).

Figure 3. GO enrichment analyses for each type of fibrosis A: Gene ontology category of OSF. B: Gene ontology category of KIF. C: Gene ontology category of LF. D: Gene ontology category of IPF. E: Gene ontology category of SF. F: Common biological processes in each type of fibrosis. G: Specific biological process of each type of fibrosis. Many different biological processes have common biological effects. GO: Gene Ontology; OSF: Oral submucous fibrosis; KIF: Kidney intestinal fibrosis; LF: Liver fibrosis; IPF: Idiopathic pulmonary fibrosis; SF: Skin fibrosis.

2.5. Functional annotation of each type of fibrosis-related module genes

To figure out the function of each type of fibrosis-related module, we selected genes in the OSF-related module, KIF-related module, LF-related module, IPF-related module, and SF-related module for KEGG and GO enrichment analysis, respectively.

The detailed results of the KEGG pathway analyses for each type of fibrosis were provided in Figure 2A-2E and GO enrichment analyses were provided in Figure 3A-3E. We found some features of fibrotic diseases by integrating KEGG results. The 5 types of fibrosis share 5 signaling pathways in common, including C-type lectin receptor signaling pathway, NOD-like receptor signaling pathway, programmed death-ligand 1 (PD-L1) expression and PD-1 checkpoint pathway in cancer, VEGF signaling pathway, and apoptosis pathway. In addition to these 5 common signaling pathways, there are 10 common pathways in OSF, KIF, LF, and IPF, mainly in immune and inflammation-related pathways, such as chemokine signaling pathway, NF-kappa B signaling pathway, B cell receptor signaling pathway, Rap1 signaling pathway, Fc epsilon RI signaling pathway, natural killer cell- and ubiquitin-mediated cytotoxicity, TNF signaling pathway, T cell receptor signaling pathway. OSF and SF also have common signaling pathways such as cell cycle, DNA replication, Th17 cell differentiation, mitogenactivated protein kinase (MAPK) signaling pathway, p53 signaling pathway, RNA transport, cellular senescence, and ubiquitin-mediated proteolysis. Collectively, pathways involved in fibrosis diseases mainly fall into 7 categories, including proliferation-related pathways, immune-related pathways, cellular stress response-related pathways, and cancer-related pathways (Figure 2F).

Further, by integrating the results of GO analysis, we found the following characteristics of fibrotic diseases. These 5 fibrotic diseases share some common biological processes, including antigen processing and presentation of exogenous antigen, response to interferon-γ, regulation of innate immune response, regulation of apoptotic signaling pathway, regulation of endopeptidase activity, positive regulation of protein transport, and negative regulation of cysteine-type endopeptidase activity (Figure 3F). These biological processes are primarily associated with inflammation and immunity. Although there are many different biological processes involved in the five types of fibrosis, different biological processes produce some of the same biological effects (Figure 3G).

3. Discussion

Fibrosis is a global health burden as fibrosis can occur in any organ, leading to impairment of organ function or even death^[4]. Therefore, the discovery of therapeutic targets associated with fibrotic diseases and the subsequent development of effective anti-fibrosis drugs targeting these targets are the difficulties and hotspots of current research. The high cost of drug development and the rare incidence of many fibrotic disorders hinder the development of targeted therapies for some fibrotic diseases. So far, only two anti-fibrosis drugs, nintedanib and pirfenidone, have been approved for clinical use, both of which are used to treat IPF^[19-20]. Thus, anti-fibrosis drugs are still very limited and have been limited to a single organ or system, and the progress continues slowly.

OSF is a kind of fibrotic diseases that occurs in the oral mucosa^[21]. Various drugs and surgical treatments have been tried to treat OSF, though the effect is limited^[22]. Researches and development of new OSF-targeted drugs or therapies are still necessary. However, there are still no acknowledged and standardized OSF animal models, which is one of the major factors restricting the development of OSF-targeted drugs^[15]. Study^[23] has even injected arecoline subcutaneously into mice to mimic OSF. However, can skin fibrosis mimic OSF? What are the similarities and differences between OSF and other organ fibrosis at the molecular level?

In recent years, the rapid development of sequencing technology has helped us better understand fibrosis at the molecular level. In this study, we used bioinformatics methods to integrate and analyze several RNA-seq or microarray datasets of different organ fibrosis to analyze the similarities and differences in the molecular mechanisms of OSF and other fibrosis across different organs. This will help researchers better understand OSF from a horizontal perspective and may provide new ideas for OSF research. Meanwhile, exploring the molecular mechanism of fibrosis in different tissues and organs may help lay a foundation for precision medicine in treating fibrotic diseases.

Through WGCNA, we obtained specific module genes related to OSF, KIF, LF, IPF, and SF, respectively. We uncovered that OSF, KIF, LF, and IPF share some common DEmRNAs, especially OSF, KIF, and LF. However, SF and the other four types of fibrosis share few common DEmRNA. Moreover, we found that except KIF and LF had five common hub genes; there were no common hub genes among other types of fibrosis. These results indicated that molecules involved in different organ fibrosis might be tissue-specific, suggesting that researchers may not be advisable to search for new molecules involved in the development of OSF by referring to the scientific literature on fibrosis in other organs. These results also suggested that it may not be appropriate to simulate the animal model of OSF by injecting arecoline into the mouse skin. The key molecules leading to fibrosis in different organs are tissue/organ specific, probably because susceptibility and pathogenesis often differ, with organ-specific risk factors, triggers, and sites of injury^[24]. However, the specific reasons still need to be further investigated.

Signaling pathways involved in the five types of fibrosis mainly fall into the following categories: immune-related pathways, cell stress-related pathways, PI3K/Akt/mTOR pathways, cell proliferation-related pathways, senescence-related pathways, programmed cell death-related pathways, and cancer-related pathways. The five types of fibrosis share only five signaling pathways in common, including C-type lectin receptor signaling pathway, NOD-like receptor signaling pathway, PD-L1 expression and PD-1checkpoint pathway in cancer, VEGF signaling pathway, and apoptosis. In addition to the 5 common signaling pathways mentioned above, there are 10 common pathways in OSF, KIF, LF, and IPF, such as chemokine signaling pathway, NF-kappa B signaling pathway, B cell receptor signaling pathway, Rap1 signaling pathway, Fc epsilon RI signaling pathway, natural killer cell-mediated cytotoxicity, TNF signaling pathway, T cell receptor signaling pathway. These pathways were mainly involved in inflammation and immunity. Inflammation is an essential trigger for fibrosis^[25]. Our study also showed that OSF, KIF, and IPF share one signaling pathway in common, the IL-17 signaling pathway. The existing literature has confirmed that IL-17 can induce fibrosis in different organs, such as the kidney^[26] and lung^[27]. This is consistent with our results. Up to now, no literature has reported changes in IL-17 expression in human specimens or animal models of OSF. However, one study has shown that arecoline can induce human oral mucosal fibroblasts to secrete inflammatory factors such as IL-2, IL-6, and IL-21, which can act on immune cells to increase Th17 and decrease Treg^[28]. The exact mechanism needs to be further explored. Moreover, our study also revealed that the mTOR signaling pathway, PI3K-Akt signaling pathway, ferroptosis, cellular senescence, and the HIF-1 signaling pathway are involved in OSF. The literature on this topic is still limited, and future OSF-related studies could consider this aspect.

Using GO analysis, we found that the five types of fibrotic diseases have seven biological processes in common, primarily associated with inflammation and immunity. In addition, many different biological processes in each type of fibrosis share the same biological functions. This suggested that therapeutics developed for one fibrotic disease may be suitable for a wide range of fibrotic disorders, as this study revealed the shared pathways across different organ fibrosis.

This study was the first to use bioinformatics methods to provide a comprehensive molecular analysis of fibrotic diseases that originated from different organs and tissues. It was also the first time to compare OSF with other organ fibrosis at the level of molecular mechanisms. Previous studies^{[4, 24-25, 29-30]} have analyzed the similarities and differences of fibrosis in different tissues and organs, but they are mainly review articles, and OSF was not included in the discussion scope. The findings of our study coincide with the main points of these reviews, which believed inflammation and immunity are important triggers for fibrosis. Previous studies have also shown that inflammation and fibrosis are two interrelated disease pathologies with several overlapping components. Three cell types, including macrophages, CD4⁺ T cells, and myofibroblasts, play important roles in regulating both inflammation and fibrosis. Inflammatory cytokines secreted by immune cells can promote the proliferation and activation of myofibroblasts. Activated myofibroblasts can secrete inflammatory cytokines, which recruit inflammatory cells to fibrotic foci and amplify the fibrotic response, forming a vicious cycle^[31]. This may explain why our study found that OSF, KIF, LF, IPF, and SF share common pathways in inflammation and immunity.

As this study only analyzed expression profiling data at the transcriptome level, the analysis results have some limitations. With the progress of sequencing technology and bioinformatics technology, it may be possible to analyze the similarities and differences of different fibrotic diseases at the level of multi-omics in the future. Further studies combining transcriptomics, proteomics, and imaging will facilitate a comprehensive understanding of fibrotic diseases. Moreover, the lack of histological expression verification is also a limitation of this paper, which will be considered in future studies.

In this study, we found that the molecule markers involved in different fibrosis are tissue- and organ-specific. Although the triggering factors, tissue and organ origins, and hub genes of OSF, KIF, LF, IPF, and SF are different, they share many common pathways and biological processes, mainly in inflammation and immunity. This suggested that therapies developed for one fibrotic disease may be applicable to a wide range of fibrotic diseases, as the study revealed that fibrosis in different organs has common signaling pathways and biological processes.

Appendix.

Supplementary Table 1 Trigger factors for fibrotic diseases

Oral submucous fibrosis	*Areca nut chewing *Consumption of spicy foods *Infection: HPV *Autoimmunity
Renal fibrosis	*Chronic kidney disease (CKD): diabetic nephropathy, glomerular disease *Autoimmunity
Liver fibrosis	*Infection: Hepatitis B virus, Hepatitis C virus *Excess alcohol consumption *non-alcoholic fatty liver disease (NAFLD) * Autoimmune liver diseases
Idiopathic pulmonary fibrosis	*Smoke *Exposure to the toxin *Particulate inhalation: metal and wood dust, agriculture and farming, stone and silica *Autoimmunity
Skin fibrosis	*Injury *Infection *Autoimmunity

Supplementary Table 2 Basic information of 5 datasets

Year	Series	Platform	Samples
2016	GSE64216	GPL10588 Illumina HumanHT-12 V4.0 expression beadchip	OSF꞉Normal=4꞉2
2016	GSE76882	GPL13158[HT_HG-U133_Plus_PM] Affymetrix HT HG-U133+ PM Array Plate	KIF꞉Normal=135꞉26
2021	GSE171294	GPL24676 Illumina NovaSeq 6000 (Homo sapiens)	LF꞉Normal=4꞉4
2017	GSE92592	GPL11154 Illumina HiSeq 2000 (Homo sapiens)	IPF꞉Normal=20꞉19
2017	GSE90051	GPL6480 Agilent-014850 Whole Human Genome Microarray 4x44K G4112F	SF꞉Normal=7꞉7

OSF: Oral submucous fibrosis; KIF: Kidney intestinal fibrosis; LF: Liver fibrosis; IPF: Idiopathic pulmonary fibrosis; SF: Skin fibrosis.

Supplementary Figure 1 Workflow of this study.

WGCNA: Weighted gene co-expression network analysis; KEGG: Kyoto Encylopedia of Genes and Genome; GO: Gene Ontology.

Supplementary Figure 2 WGCNA analysis was used to screen each type of fibrosis-related module genes.

A and B: Cluster dendrogram and module-trait relationships of OSF. C and D: Cluster dendrogram and module-trait relationships of KIF. E and F: Cluster dendrogram and module-trait relationships of LF. G and H: Cluster Dendrogram and module-trait relationships of IPF. I and J: Cluster dendrogram and module-trait relationships of SF. WGCNA: Weighted gene co-expression network analysis; OSF: Oral submucous fibrosis; KIF: Kidney intestinal fibrosis; LF: Liver fibrosis; IPF: Idiopathic pulmonary fibrosis; SF: skin fibrosis.

Supplementary Figure 3 Identification of common hub genes in each type of fibrosis.

OSF does not share a hub gene with the other four types of fibrosis. Moreover, except that KIF and LF have five hub genes in common, including CD2, CORO1A, CD48, CD53, and RAC2, other types of fibrosis do not share hub genes. OSF: Oral submucous fibrosis; KIF: Kidney intestinal fibrosis; LF: Liver fibrosis.

Funding Statement

This work was supported by the Natural Science Foundation of Hunan Province (2020JJ5404, 2022JJ30871), the Hunan Health Commission Research Grant (202108011054), and the Clinical Medical Boot Technology Innovation Project of Hunan Province (2021SK53602), China.

Conflict of Interest

The authors declare that they have no conflicts of interest to disclose.

AUTHORS’CONTRIBUTIONS

CHEN Jun Designed the research, collected and analyzed the data, wrote the first draft of the manuscript; LIU Binjie, XIE Xiaoli, and LI Wenjie Designed the research. All authors revised the manuscript and approved the manuscript.

Note

http://xbyxb.csu.edu.cn/xbwk/fileup/PDF/2022121663.pdf