Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Integrative weighted molecular network construction from transcriptomics and genome wide association data to identify shared genetic biomarkers for COPD and lung cancer

  • Babajan Banaganapalli ,

    Contributed equally to this work with: Babajan Banaganapalli, Bayan Mallah, Kawthar Saad Alghamdi

    Roles Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing (BB); (NAS)

    Affiliations Department of Genetic Medicine, Faculty of Medicine, King Abdulaziz University, Jeddah, Saudi Arabia, Princess Al-Jawhara Al-Brahim Center of Excellence in Research of Hereditary Disorders, King Abdulaziz University, Jeddah, Saudi Arabia

  • Bayan Mallah ,

    Contributed equally to this work with: Babajan Banaganapalli, Bayan Mallah, Kawthar Saad Alghamdi

    Roles Data curation, Formal analysis, Methodology, Validation, Visualization, Writing – original draft

    Affiliations Department of Genetic Medicine, Faculty of Medicine, King Abdulaziz University, Jeddah, Saudi Arabia, Princess Al-Jawhara Al-Brahim Center of Excellence in Research of Hereditary Disorders, King Abdulaziz University, Jeddah, Saudi Arabia

  • Kawthar Saad Alghamdi ,

    Contributed equally to this work with: Babajan Banaganapalli, Bayan Mallah, Kawthar Saad Alghamdi

    Roles Data curation, Formal analysis, Methodology, Validation

    Affiliation Department of Biology, Faculty of Science, University of Hafr Al Batin, Hafr Al Batin, Saudi Arabia

  • Walaa F. Albaqami,

    Roles Writing – review & editing

    Affiliation Department of Science, Prince Sultan Military College of Health Sciences, Dhahran, Saudi Arabia

  • Dalal Sameer Alshaer,

    Roles Writing – review & editing

    Affiliation Department of Genetic Medicine, Faculty of Medicine, King Abdulaziz University, Jeddah, Saudi Arabia

  • Nuha Alrayes,

    Roles Writing – review & editing

    Affiliations Princess Al-Jawhara Al-Brahim Center of Excellence in Research of Hereditary Disorders, King Abdulaziz University, Jeddah, Saudi Arabia, Department of Medical Laboratory Sciences, Faculty of Applied Medical Sciences, King Abdulaziz University, Jeddah, Saudi Arabia

  • Ramu Elango,

    Roles Conceptualization, Supervision, Writing – review & editing

    Affiliations Department of Genetic Medicine, Faculty of Medicine, King Abdulaziz University, Jeddah, Saudi Arabia, Princess Al-Jawhara Al-Brahim Center of Excellence in Research of Hereditary Disorders, King Abdulaziz University, Jeddah, Saudi Arabia

  • Noor A. Shaik

    Roles Conceptualization, Investigation, Supervision, Writing – original draft, Writing – review & editing (BB); (NAS)

    Affiliations Department of Genetic Medicine, Faculty of Medicine, King Abdulaziz University, Jeddah, Saudi Arabia, Princess Al-Jawhara Al-Brahim Center of Excellence in Research of Hereditary Disorders, King Abdulaziz University, Jeddah, Saudi Arabia


Chronic obstructive pulmonary disease (COPD) is a multifactorial progressive airflow obstruction in the lungs, accounting for high morbidity and mortality across the world. This study aims to identify potential COPD blood-based biomarkers by analyzing the dysregulated gene expression patterns in blood and lung tissues with the help of robust computational approaches. The microarray gene expression datasets from blood (136 COPD and 6 controls) and lung tissues (16 COPD and 19 controls) were analyzed to detect shared differentially expressed genes (DEGs). Then these DEGs were used to construct COPD protein network-clusters and functionally enrich them against gene ontology annotation terms. The hub genes in the COPD network clusters were then queried in GWAS catalog and in several cancer expression databases to explore their pathogenic roles in lung cancers. The comparison of blood and lung tissue datasets revealed 63 shared DEGs. Of these DEGs, 12 COPD hub gene-network clusters (SREK1, TMEM67, IRAK2, MECOM, ASB4, C1QTNF2, CDC42BPA, DPF3, DET1, CCDC74B, KHK, and DDX3Y) connected to dysregulations of protein degradation, inflammatory cytokine production, airway remodeling, and immune cell activity were prioritized with the help of protein interactome and functional enrichment analysis. Interestingly, IRAK2 and MECOM hub genes from these COPD network clusters are known for their involvement in different pulmonary diseases. Additional COPD hub genes like SREK1, TMEM67, CDC42BPA, DPF3, and ASB4 were identified as prognostic markers in lung cancer, which is reported in 1% of COPD patients. This study identified 12 gene network- clusters as potential blood based genetic biomarkers for COPD diagnosis and prognosis.

1. Introduction

Chronic obstructive pulmonary disease (COPD) is a progressive airflow obstruction in the lungs which slowly becomes apparent after the 40th or 50th year of age [1]. With a global prevalence of 251 million, COPD disease is currently the fourth leading cause of global deaths and ranked fifth in terms of disease burden [2, 3]. The primary characteristics of the disease are lung inflammation, breathing difficulties, airflow blockage, emphysema, long term cough with mucus, chronic bronchitis, and refractory asthma [4]. Although cigarette smoking is the most well-known significant risk factor for COPD, other factors such as tuberculosis history and environmental exposure to lung irritants (such as indoor air pollutants and occupational dust) are also known to contribute to modifying disease causality and severity [5, 6]. Chronic inflammation is thought to be responsible for pathologic changes such as narrowing of airways in the lungs and destruction of the lung parenchyma, with an underlying role of genetic, epigenetic, and environmental factors [7].

Genetic studies of twins [8], first degree relatives [9] and sporadic COPD cases [10] have all confirmed the role of heritability, which explains at least 30% of the variation in COPD risk. For so long, the genetic basis of COPD has come from Mendelian syndromes, where rare pathogenic variants in ELN and FBLN5 genes cause cutis laxa and SERPINA1 causes α1-antitrypsin deficiency [11]. Genome-wide association studies have reported the strong association of over 20 genetic loci with COPD and a few additional loci for COPD-related phenotypes like hypoxemia, chronic bronchitis, and emphysema [12]. The molecular basis of COPD, however, could not be fully explained by candidate genetic variants alone, but also by changes in global gene expression. Besides providing an unbiased assessment of thousands of genes in the disease etiology, global gene expression could also potentially help in developing personalized medicine. However, analysis and interpretation of such massive gene expression data is so complex and challenging.

A few studies have attempted to analyze gene expression changes in COPD patients’ blood samples in recent years [1316]. However, the correlation of common gene expression dysregulations between blood and lung tissue samples from COPD patients is not well explored. Recent deployment of advanced statistics and integrative bioinformatics methods, incorporating gene network graphs, unsupervised clustering, and functional annotations of pathways, has provided a new dimension to explore the microarray gene expression datasets to discover the molecular basis of different genetic pathologies [1719]. Therefore, the objective of this study is to expand our current understanding of COPD pathogenesis and to identify potential genetic biomarkers. By involving a series of comprehensive bioinformatics approaches, this study has identified several gene-network clusters involved in cell communication, inflammation, proliferation, and differentiation processes, are dysregulated in blood and lung tissues of COPD patients. Our findings provide an insight into understanding the mechanisms of COPD and its potential link with lung cancer, besides uncovering genetic markers with potential for disease diagnosis and therapeutic modulation.

2. Materials and methods

2.1 Microarray gene expression datasets

The NCBI-GEO and EBL-EBI Array Express databases were used to search for COPD gene expression datasets using the keywords like “COPD”, “COPD blood”, and “COPD tissue”. From the output, we selected two COPD gene expression datasets, i.e. GSE8581 and GSE54837 for our study. The first dataset (GSE8581) consists of gene expression data, from 35 lung tissues, which were collected from 16 COPD subjects (with FEV1 < 70% predicted and FEV1/FVC < 0.7) and 19 controls (with FEV1 > 80% predicted and FEV1/FVC > 0.7), generated on the Affymetrix U133 Plus 2.0 array [20]. The second dataset, GSE54837 includes the expression data generated on GPL570 platform (Affymetrix, Santa Clara, CA, USA) from the blood samples of 136 COPD patients and 6 controls (ex-smokers) [21].

2.2 Data preprocessing and analysis

The microarray gene expression data analysis was performed using R/Bioconductor ( The raw data extracted in.CEL format was normalized into expression values using the Bioconductor-Affy package for the standardization and background correction of the probe data [22]. The limma package was then used to select the statistical significance of the differentially expressed genes between normal and COPD samples by applying the t-test statistical method. The Benjamini-Hochberg method was used to calculate the false discovery rate (FDR) of all the statistically significant genes to enable the removal of false positive ones [24]. The cutoff value for DEGs was set as FDR < 0.01 and |log2 FC| > 1.5. A p-value of less than 0.05 was considered as statistically significant. The expression values of DEGs were divided into up- and down-regulated genes and visualized using the Heatmap online webtool (

2.3 Gene ontology and functional enrichment analyses

Gene Ontology (GO) and KEGG pathway ( enrichment analysis of DEGs was conducted using STRING database ( The significant GO terms and pathways were chosen at a threshold of adjusted p< 0.05 and FDR of 0.05. The GO annotation networks were visualized in the Cytoscape network style plugin (

2.4 Construction of protein-protein interaction (PPI) map

The potential PPI networks from the lung and blood DEGs were constructed using Bisogenet, a cytoscape plugin (version 3.4.0). Furthermore, the network clusters from PPI interactions were identified with the help of network analyzer tool. The cut-off value of input nodes and their neighbors was up to a distance of 1 edge. During the creation of PPIM, only protein-protein interactions were selected, excluding protein-DNA interactions and microRNA silencing interactions. Each node represents a gene connected with edges which are physical or functional between the nodes. Therefore, few nodes have a large number of edges while several nodes have low connectivity [23].

2.5 Hub gene subnetwork construction

PPIM is considered to be a large-scale network. By following the network biology concepts, the PPIM complex was decomposed into significant subnetwork clusters of Significant Protein Interaction Network (SPIN). Based on degree centrality (DC) and betweenness centrality (BC) parameters, several genes were extracted. Each protein captured in the network was incorporated and standardized into Cytoscape 3.2.1 using Network Analyzer to calculate local degree centrality (DC) and global betweenness centrality (BC) parameters of the network [24].

2.6 Genome wide association study analysis

The hub genes from the above gene-network clusters were searched in the GWAS catalog database ( to check their association with COPD risk. Variant details like reference and alternate alleles, population frequency, genome wide association value (P-value of <5 × 10−8), reported trait, and accession number of the study were collected. We have also used another genotype-phenotype association database, PhenoScanner V2 ( to cross reference the association of hub genes with COPD risk. Each hub gene name was searched in the database generated tables, which contain trait specific associations of each gene and genome wide association values for its variants (P-value of <5 × 10−8).

2.7 Lung cancer expression analysis

We used three different databases to investigate the expression status of the COPD-hub genes in lung cancer tissues: Gene Expression Profiling Interactive Analysis (GEPIA2), Gene Expression Database of Normal and Tumor Tissues (GENT2), and Human Protein Atlas (HPA). Gene Expression Profiling Interactive Analysis (GEPIA2) ( was used to provide tumor/normal differential expression analysis. The signature score of hub genes is calculated by mean value of log2 (TPM + 1). The |Log2FC| of 1 and an expression value cutoff of 0.01 (p-value) were determined in Lung Adenocarcinoma (LUAD) and Lung Squamous Carcinoma (LUSC) tissues. The Gene Expression Database of Normal and Tumor Tissues GENT2 ( platform was used to explore the gene expression patterns across normal and tumor tissues generated from public gene expression data sets. The survival rate status of hub genes in lung cancer and its histological subtypes (adenocarcinomas, large and squamous) represented by Kaplan Meier plots at 95% confidence intervals (CI) and computed log rank p-value was determined. The human protein atlas ( database was used to explore the expression status of each hub gene in human non-malignant and lung cancer tissues. This database takes the query gene or protein name and provides the information about that candidate protein expression based on the primary antibody staining data with a series of immunohistochemistry images of the corresponding clinical specimens.

3. Results

3.1 Differently expressed gene (DEGs) identification

A total of 54,675 probes were expressed in both datasets. In the human lung tissue dataset (E-GEOD-8581), 678 DEGs including 247 up- and 431 down-regulated genes were identified, whereas, blood dataset (GSE54837) showed the differential expression of 724 DEGs including 499 up- and 225 down-regulated genes. Comparison of both datasets revealed the shared expression of 63 genes (Fig 1A). The expression level of DEGs of COPD patient samples (both tissue and blood) is shown in the form of heatmaps and volcano plots (Fig 1B and 1C).

Fig 1.

B Analysis of COPD differentially expressed genes (DEGs) in comparison to corresponding controls (A) Volcano plots of log fold changes in gene expression. (B) Identification of 63 common DEGs from blood and lung tissue datasets using VENNY. The overlapped area defines the shared DEGs of lung tissue and blood. (C) Heatmap of DEGs with a LogFC > 1.5. Red: up-regulation; green: down-regulation.

3.2 PPI network analysis and significant genes clusters

Bisogenet, a Cytoscape plugin analysis of DEGs from both datasets generated a complex PPIM network of 1072 nodes (genes) and 20079 edges (interactions). The average edge-node ratio was 18.73 (S1 Fig). In the context of the PPIM network, protein interactions within the same group of clusters are assumed to have similar functions to the less interconnected regions or different cluster groups. Therefore, the Network Analyzer plugin was applied to find significant hub genes with the highest degree of centrality. A total of 12 significant genes and clusters with a degree of centrality of >17 were identified from network analysis (Fig 2; S2 Fig) and chosen as hub proteins (Table 1).

Fig 2. Hub genes TMEM67, IRAK2, ASB4, MECOM, DPF3 and DDX3Y with their clusters identified from the common DEGs between blood and lung tissue datasets.

Their selection is based on degree of centrality in the PPI network with the score >18.

Table 1. A total of 12 significant genes with more than 17 of DC were obtained from network analysis and chosen as hub proteins.

3.3 GO annotation analysis

Gene Ontology annotation is the process by which functional categories of genes are assigned. The GO annotations of 12 COPD gene clusters showed their enrichment in cell-cell communication, cell regulation, immune processes, transcription factors regulation and ubiquitin pathways. Four of these 12 COPD -gene clusters, CCDC74B, MECOM, IRAK2 and DET1 have shown the lowest FDR values (Table 2), which reflects their highest functional enrichment in i molecular function (MF), biological process (BP), cellular components (CC) categories and KEGG pathways. For the CCDC74B cluster, GO enrichment highlights its involvement in ‘ubiquitin pathways and protein modification’, under the biological processes category, ‘Protein Deubiquitination’ (GO:0016579) was the top GO term. The other top GO enriched terms falling into remaining categories are as follows; ‘proteasome-activating ATPase activity’ in MF, ‘Proteasome Regulatory Particle’ (GO:0005838) in CC and ‘Protein degradation’ (hsa03050) in KEGG pathways. IRAK2 cluster was highly involved in signaling pathways and Kinase activity. The GO term in BP highlighted ‘Interleukin-1-Mediated Signaling Pathway’ (GO:0070498), which mediates cytokine responses during inflammation. The MF ontology source showed ‘Protein Kinase Activity’ (GO:0004672) and ‘Catalytic Activity Acting on A Protein’ (GO:0140096) as top GO terms. The CC ontology source was mainly enriched in ‘endosome membrane’ (GO:0010008) and ‘Cytoplasmic Vesicle Part’ (GO:0044433). KEGG underlined GO terms which are responsible for cytokine production and regulating the immune response like ‘NF-Kappa B Signaling Pathway’ (hsa04064) and ‘Toll-Like Receptor Signaling Pathway’ (hsa04620). DET1 cluster was mostly reported in relation to protein degradation processes. The BP ontology source highlighted ‘Regulation of Protein Catabolic Process’ (GO:0042176) as the top GO term. MF ontology source identified ‘Ubiquitin Protein Ligase Binding’ (GO:0031625) as the significant GO term. ‘Cul4-RING E3 Ubiquitin Ligase Complex’ (GO:0080008) are the top CC terms, while ‘Ubiquitin Mediated Proteolysis’ (hsa04120) and ‘Nucleotide Excision Repair’ (hsa03420) was the significant KEGG pathways. (Fig 3A). The MECOM cluster was highly enriched in regulation of transcription by ‘RNA polymerase II’ as top BP GO term (GO:0006357). Top MF term was ‘transcription regulator activity’ (GO:0140110). ‘Nucleoplasm’ (GO:0005654) and ‘Nuclear Lumen’ (GO:0031981) are the top CC terms, and ‘Pathways in cancer’ was the significant KEGG pathway (Fig 3B).

Fig 3.

GO-annotations stacked network view of (A) MECOM and (B) IRAK2 clusters. The size of the circle (left side) represents the number of genes involved in a specific GO-term. The GWAS loci of (C) MECOM (D) IRAK2 genes from the GWAS catalog.

Table 2. Functional enrichment of CCDC74B, MECOM, IRAK2 and DET1 clusters, in which highlights highest functional enrichment in different molecular processes like molecular function (MF), biological process (BP), cellular components (CC) and KEGG pathways based on FDR value.

The functional enrichment values of the remaining 8 gene clusters (CDC42BPA, DPF3, SREK1, TMEM67, ASB4, DDX3Y, KHK, and C1QTNF2) are shown in Fig 3A and 3B. CDC42BPA cluster predicted its participation in ‘Fc Gamma R-Mediated Phagocytosis’ (hsa04666). ‘Legionellosis’ (hsa05134) was mainly enriched by DPF3 cluster. While the TMEM67 and ASB4 clusters were mostly involved in "Proteasome" (hsa03050) and "Ubiquitin-mediated proteolysis" (hsa04120), the SREK1 cluster was mostly involved in ‘RNA binding’ (GO:0003723). The ‘Pathways in cancer’ (hsa05200) was significant GO term in DDX3Y cluster. Lastly, KHK and C1QTNF2 were mainly enriched in ‘metabolic process’ (GO:0044238) and ‘extracellular matrix-receptor interaction’ (hsa04512) respectively.

3.4 Examining the role of hub genes in data from COPD genome wide association studies (GWAS)

Both the GWAS catalog and PhenoScanner databases were used to collate the genetic association data of hub genes with the risk of COPD development. The GWAS data findings of 12 COPD-hub genes, include variant IDs, reference and alternate alleles, significance of association (P = 5 × 10−8), phenotypic traits associated with the query genes, etc. GWAS findings revealed the association of IRAK2 with eosinophil count alterations usually manifested in inflammatory conditions (Table 3). For the MECOM hub gene, the associated traits are pulmonary complications including COPD, asthma and lung function (Fig 4C and 4D No significant GWAS data linking the remaining 10 COPD-hub genes (SREK1, TMEM67, ASB4, C1QTNF2, CDC42BPA, DPF3, DET1, CCDC74B, KHK, and DDX3Y) with any kind of lung disease was found. In PhenoScanner, 6 out of 12 hub genes (IRAK2, MECOM, ASB4, CDC42BPA, DPF3 and TMEM67) have revealed an association with lung related traits and lung cancer (S1 Table). IRAK2 is associated with lung cancer and a high eosinophil count. Genotype-phenotype associations of MECOM highlighted pulmonary function interaction, lung cancer, and COPD with acute exacerbation. Both ASB4 and CDC42BPA showed an association with COPD with acute lower respiratory infection. The DPF3 gene is associated with COPD and squamous cell carcinoma, lung cancer. Lastly, TMEM67 is associated with lung cancer.

Fig 4. Expression levels in lung adenocarcinoma and lung squamous cell carcinoma cells in compression to normal tissues from GEPIA2.

A) SREK1. B) IRAK2. C) DDX3Y. D) C1QTNF2. The signature score is calculated by mean value of log2 (TPM + 1). The |Log2FC| cutoff of the expression of proposed biomarker was 1. The p-value cutoff of the expression of proposed biomarker was 0.01. The red box indicates the tumor samples while the gray one represents the normal tissues. E. Pathological Stage Plot of SREK1, IRAK2, DDX3Y and C1QTNF2 genes in lung cancer.

Table 3. Association of two hub genes with the lung related traits and lung cancer from GWAS catalog database.

3.5 Examining the transcriptional status of COPD hub genes in lung cancer expression

In GPEIA2 analysis, boxplots of 12 hub genes were retrieved. Adenocarcinoma (LUAD) and Squamous Cell Carcinoma (LUSC) were selected with a P-value cutoff of 0.01 using The Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx) database. Out of 12 hub genes, only 4 (IRAK2, SREK1, C1QTNF2 and DDX3Y) have shown significant gene expression in lung cancer compared to the normal tissues (Fig 4A–4E). The boxplots of IRAK2, SREK1 and C1QTNF2 show their significant expression in LUSCs. The DDX3Y gene was significantly expressed in both LAUD and LUSC cells. The GENT2 platform is used to explore gene expression patterns across normal and lung tumor tissues. Fig 5 shows the prognostic value (patient survival in days) of the expression status of 6 COPD-hub genes. Out of the 12-COPD hub genes, 5 genes (SREK1, TMEM67, CDC42BPA, DPF3, and ASB4) showed an improvement in lung cancer survival duration up to 1500 days (P values for all the associations is <0.02) (Fig 5). The correlation of survival status of patients with different lung cancer subtypes to all five gene expression levels reveals that adenocarcinomas have a longer survival rate (0.2–0.4) than those with squamous and large cell lung cancers.

Fig 5. The prognostic values (patient survival in days) of the expression status of 6 COPD-hub genes.

A) SREK1 (P<0.001). B) TMEM67 (P = 0.002). C) CDC42BPA (P = 0.003). D) DPF3 (P = 0.005). E) ASB4 (P = 0.024) F) IRAK2 (P = 0106). The correlation of survival status of patients with different lung cancer subtypes (Squamous, Adeno, Large) to all six genes expression level.

The Human Protein Atlas (HPA) derived protein expression status in normal tissue and lung cancer tissues for COPD- hub genes is illustrated in Fig 6 and Table 4. Abundance of these proteins could be divided into four categories like high, medium, low and not detected by the scoring system based on the intensity of staining, whether strong, moderate, weak, or negative. The macrophage and pneumonocytes staining for ASB4 in normal lung tissues was not detected and was medium in lung cancer. Medium staining detection of CDC42BPA and IRAK2 were found in both normal and cancer lung tissue in addition to SREK1 and C1QTNF2 which was observed in medium in normal tissue but higher in lung cancer tissue. Staining of MECOM was very high in both normal and cancer lung tissue. While low to no protein detection of DET1 and KHK were observed in both normal and lung cancer tissue. Furthermore, DDX3Y and DPF3 staining in normal lung tissues were negative but DDX3Y was higher in lung cancer tissue but not DPF3. Finally, data were not available for TMEM67 and CCDC74B genes in HPA (S3 Fig).

Fig 6. The expression of the two hub genes from the Human Protein Atlas (HPA) in normal and cancer lung tissue.

Table 4. The expression levels of the 10 hub genes in normal lung and cancer tissues: Human Protein Atlas (HPA).

4. Discussion

Massive high throughput genome wide- sequencing and expression studies have been effective in querying the molecular basis of many inherited diseases in humans. However, deciphering the molecular basis of chronic diseases is challenging, owing to the complex interplay of genes and environmental factors. The etiopathogenesis of complex diseases like COPD can be better explained by studying the global gene expression changes. The recent biomarker discoveries in intracranial aneurysm [25], Parkinson disease [26], Diabetes mellitus [27] and cancers [28] once again proves the robustness of bioinformatics methods in analyzing the huge gene expression data. Few studies have attempted to analyze the gene expression changes either in blood samples [29, 30] or tissue samples of COPD patients [31, 32]. But, none of them attempted to identify blood based genetic biomarkers. Therefore, this study tried to explore the shared gene expression changes between both blood and lung tissues to identify potential biomarkers to assist in diagnostics or prognostic aspects of COPD patients.

Chronic diseases are caused by the action of dysregulation of multiple genes at different stages of the disease pathology. Hence, we constructed a protein interactome based on the differentially expressed genes in the COPD patients. The protein-protein interaction networks establish the physical contacts between two or more proteins as a result of biochemical events underlying the disease etiopathogenesis. The characteristic features of PPI is based on various connectivity between nodes and edges, where each node indicates a gene connected to its functional partners [23]. To reduce the complexity of PPI network, highest connected nodes are decomposed into a clusters or modules. The gene with the highest number of edges among group of genes within the same cluster is known as hub gene, which are basically chosen based on its degree of centrality (DC) values in the network [33]. DC also refers to node connectivity, i.e. the number of connections to the node and its interaction [34]. In context of these network principles in identifying the COPD-hub genes, we assessed the essential properties of the genes that are involved in the disease.

Since clusters are characterized by extensive connectivity between a set of genes, GO annotations provides functional interpretation of them under vaiety of biological categories [35]. In the present study we identified 12 COPD hub gene clusters (SREK1, TMEM67, IRAK2, MECOM, ASB4, C1QTNF2, CDC42BPA, DPF3, DET1, CCDC74B, KHK, DDX3Y) from PPI network, which revealed their enrichment in cell regulation, immune process, transcription factors regulations and protein degradation pathways. The upregulated gene (DDX3Y) in both blood and lung tissue were enriched in functions associated with regulation of gene expression, cell cycle, cellular senescence and FoxO signaling pathway which is involved in many cellular physiological events such as apoptosis and cell-cycle control. Moreover, there were two downregulated genes (MECOM and KHK). MECOM were associated with regulation of transcription, pathways in cancers and FoxO signaling pathway. While KHK were involved in Starch and sucrose metabolic processes. However, of those 12 gene clusters, CCDC74B, MECOM, IRAK2 and DET1 clusters had shown the lowest FDR values, which reflects their highest functional enrichment in different molecular processes. The CCDC74B gene cluster was mainly enriched in proteasome pathway, which degrades unneeded proteins within the cell. The activity of proteasome can be impaired by cigarette smoke resulting in reduction of antigen presentation and lead to prolonged lung infections and COPD patients [36].

In lung tissues of COPD patients, accumulation of ubiquitylated proteins and further degradation by proteasome machinery is reported [37]. Protein catabolic processes pathway enriched in DET1 gene cluster also plays an important role in pathogenesis of COPD. The chronic inflammation in COPD contributes to the imbalance of protein degradation resulting in the loss of skeletal muscle protein, one of the characteristic features present in COPD [38]. On the other hand, MECOM gene cluster highlights the regulation of transcriptional pathway which controls the changes in gene transcription of many inflammatory substances that play a key role in the pathogenesis of COPD [39, 40]. The IRAK2 gene cluster showed its involvement in regulation of inflammatory process such as interleukin (IL)-1 pathway activation and Toll like receptor that is directly linked to the pathogenesis of COPD, is characterized by abnormal release of inflammatory cytokines, remodeling of the airways and dysregulated immune cell activity [41, 42].

Genome wide association studies reveals the association of genetic variants with risk of developing common diseases by screening genetic samples from thousands of samples. In this study, 12 hub genes were searched in GWAS databases for their association with COPD, lung function traits as well as lung cancers. The GWAS data confirmed that the variants in the 6 COPD-hub genes (IRAK2, MECOM, CDC42BPA, ASB4, DPF3 and TMEM67) shows genome wide significant association to traits that could potentially modify the risk of COPD pathology development. At least 5 variants in IRAK2 were significantly associated with variety of eosinophil count traits [43]. Eosinophilia (high eosinophil counts) causes inflammation of the lung tissue and exacerbates the lung function in the COPD patients. However, the role of eosinophils in COPD is unclear, as not all COPD patients develop eosinophilic airway inflammation [44, 45]. Interestingly, IL-1 signaling has been shown to be associated with eosinophilic inflammatory profiles in patient with COPD [43]. Moreover, in COPD patients with eosinophilic inflammation have the tendency to respond to steroid therapy. Therefore, eosinophil count is an important point of view to direct biological therapies for COPD [46]. Many variants in MECOM were strongly associated with FEV1and other traits that are directly related to lung function and COPD pathogenesis [47]. Other COPD- hub genes (CDC42BPA, ASB4, DPF3 and TMEM67) are were also associated with lung function related traits and lung cancer [4850].

COPD is one of the significant risk factors for oncogenesis of the lung tissues, which is seen in about 1% of COPD patients every year [51]. Both COPD and lung cancer share many common pathways such as immune dysfunction and regulation of transcription factors [52]. Interestingly, pathways enriched by MECOM and IRAK2 were involved in lung cancer development. For instance, MECOM is an important transcription factor involved in oncogenesis [53, 54]. Aberrant expression of MECOM is one of the characteristic features of many malignancies including leukemia [55] and solid tumors such as breast cancer and hepatocellular carcinoma [53, 56] as well as lung cancer [57]. Moreover, frequent alterations in MECOM have been associated with primary and metastatic lung adenocarcinomas [58]. On the other hand, activation of the TLR pathway has a significant impact on cancer progression regulation including lung cancer [59, 60]. One genetic variant in IRAK2 (rs779901 C > T) in the TLR signaling pathway is suggested to be a prognostic biomarker for non-small cell lung cancer (NSCLC) [61]. Global gene expression profile analysis provides a valuable insight into the normal biological process and to disease pathogenesis [62]. To support the contribution of IRAK2 and MECOM hub genes, significant dysregulation of expression in lung cancer types were observed in HPA, GPEIA2 and GENT2 databases as well. Furthermore, differentially expression of IRAK2 and MECOM genes has been reported in many studies in cancers or COPD [6365].

5. Conclusions

In conclusion, we identified, 12 blood based molecular biomarkers (SREK1, TMEM67, IRAK2, MECOM, ASB4, C1QTNF2, CDC42BPA, DPF3, DET1, CCDC74B, KHK, DDX3Y) for COPD diagnosis, by integrative gene expression and gene network approaches. Out of these 12 hub genes, two (MECOM and IRAK2) were over expressed in lung cancers tissues, which reflects a shared molecular lineage between COPD and lung cancers. Interestingly, we have also identified that the expression status of other COPD hub genes like SREK1, TMEM67, CDC42BPA, DPF3, and ASB4 improves the survival duration of lung cancer patients, hence they may act as potential molecular drug targets and/or biomarkers for both COPD and/or lung cancer. However, biological and clinical relevance of each COPD hub gene can be better understood, when our findings are explored through future in vitro and in vivo validation assays.

Supporting information

S1 Fig. Overview of PPI network constructed from 63 common genes using cytoscape STRING database.

The PPI network at p-value >0.05 consist of 995 nodes interact with 18924 edges.


S2 Fig. The 12 hub gene protein interaction network.

(A) Cluster-1 (SREK1). (B) Cluster-2 (TMEM67). (C) Cluster-3 (IRAK2). (D) Cluster-4 (MECOM). (E) Cluster-5 (ASB4). (F) Cluster-6 (CDC42BPA). (G) Cluster-7 (DPF3). (H) Cluster-8 (DET1). (I) Cluster-9 (KHK). (J) Cluster-10 (DDX3Y), the hub gene selected based on high centrality in the protein network of DEGs.


S3 Fig. Histopathological images of DEGS.

Protein Pathology Atlas of 12 hug genes in normal lung and lung cancer tissues.


S1 Table. Hub genes GWAS association.

Genetic association of hub genes with the lung related traits and lung cancer from Phenoscanner database.



  1. 1. Lareau SC, Fahy B, Meek P, Wang A. Chronic Obstructive Pulmonary Disease (COPD). Am J Respir Crit Care Med. 2019;199(1):P1–p2. Epub 2018/12/29. pmid:30592446.
  2. 2. López-Campos JL, Tan W, Soriano JB. Global burden of COPD. Respirology (Carlton, Vic). 2016;21(1):14–23. Epub 2015/10/24. pmid:26494423.
  3. 3. Woldeamanuel GG, Mingude AB, Geta TG. Prevalence of chronic obstructive pulmonary disease (COPD) and its associated factors among adults in Abeshge District, Ethiopia: a cross sectional study. BMC Pulmonary Medicine. 2019;19(1):181. pmid:31623601
  4. 4. Pauwels RA, Rabe KF. Burden and clinical features of chronic obstructive pulmonary disease (COPD). The Lancet. 2004;364(9434):613–20. pmid:15313363
  5. 5. Rodriguez-Gonzalez E, Ferrer-Sancho J. Occupational exposure and COPD. Current Respiratory Medicine Reviews. 2012;8(6):436–40.
  6. 6. Kurmi OP, Semple S, Simkhada P, Smith WCS, Ayres JG. COPD and chronic bronchitis risk of indoor air pollution from solid fuel: a systematic review and meta-analysis. Thorax. 2010;65(3):221. pmid:20335290
  7. 7. Barnes PJ. Inflammatory mechanisms in patients with chronic obstructive pulmonary disease. Journal of Allergy and Clinical Immunology. 2016;138(1):16–27. pmid:27373322
  8. 8. Ingebrigtsen T, Thomsen SF, Vestbo J, van der Sluis S, Kyvik KO, Silverman EK, et al. Genetic influences on Chronic Obstructive Pulmonary Disease—a twin study. Respiratory medicine. 2010;104(12):1890–5. Epub 2010/06/15. pmid:20541380.
  9. 9. McCloskey SC, Patel BD, Hinchliffe SJ, Reid ED, Wareham NJ, Lomas DA. Siblings of patients with severe chronic obstructive pulmonary disease have a significant risk of airflow obstruction. Am J Respir Crit Care Med. 2001;164(8 Pt 1):1419–24. pmid:11704589.
  10. 10. Regan EA, Hersh CP, Castaldi PJ, DeMeo DL, Silverman EK, Crapo JD, et al. Omics and the Search for Blood Biomarkers in Chronic Obstructive Pulmonary Disease. Insights from COPDGene. American journal of respiratory cell and molecular biology. 2019;61(2):143–9. Epub 2019/03/16. pmid:30874442; PubMed Central PMCID: PMC6670029.
  11. 11. Ragland MF, Benway CJ, Lutz SM, Bowler RP, Hecker J, Hokanson JE, et al. Genetic Advances in Chronic Obstructive Pulmonary Disease. Insights from COPDGene. Am J Respir Crit Care Med. 2019;200(6):677–90. Epub 2019/03/26. pmid:30908940; PubMed Central PMCID: PMC6775891.
  12. 12. Ragland MF, Benway CJ, Lutz SM, Bowler RP, Hecker J, Hokanson JE, et al. Genetic Advances in COPD: Insights from COPDGene. American journal of respiratory and critical care medicine. 2019. Epub 2019/03/26. pmid:30908940.
  13. 13. Bahr TM, Hughes GJ, Armstrong M, Reisdorph R, Coldren CD, Edwards MG, et al. Peripheral blood mononuclear cell gene expression in chronic obstructive pulmonary disease. American journal of respiratory cell and molecular biology. 2013;49(2):316–23. Epub 2013/04/18. pmid:23590301; PubMed Central PMCID: PMC3824029.
  14. 14. Chang Y, Glass K, Liu YY, Silverman EK, Crapo JD, Tal-Singer R, et al. COPD subtypes identified by network-based clustering of blood gene expression. Genomics. 2016;107(2–3):51–8. Epub 2016/01/17. pmid:26773458; PubMed Central PMCID: PMC4761317.
  15. 15. Reinhold D, Morrow JD, Jacobson S, Hu J, Ringel B, Seibold MA, et al. Meta-analysis of peripheral blood gene expression modules for COPD phenotypes. PloS one. 2017;12(10):e0185682. Epub 2017/10/11. pmid:29016655; PubMed Central PMCID: PMC5633174.
  16. 16. Chen L, Appel LJ, Loria C, Lin PH, Champagne CM, Elmer PJ, et al. Reduction in consumption of sugar-sweetened beverages is associated with weight loss: the PREMIER trial. Am J Clin Nutr. 2009;89(5):1299–306. Epub 2009/04/03. pmid:19339405; PubMed Central PMCID: PMC2676995.
  17. 17. Banaganapalli B, Mansour H, Mohammed A, Alharthi AM, Aljuaid NM, Nasser KK, et al. Exploring celiac disease candidate pathways by global gene expression profiling and gene network cluster analysis. Sci Rep. 2020;10(1):16290. Epub 2020/10/03. pmid:33004927; PubMed Central PMCID: PMC7529771.
  18. 18. Sabir JSM, El Omri A, Banaganapalli B, Aljuaid N, Omar AMS, Altaf A, et al. Unraveling the role of salt-sensitivity genes in obesity with integrated network biology and co-expression analysis. PLoS One. 2020;15(2):e0228400. Epub 2020/02/07. pmid:32027667; PubMed Central PMCID: PMC7004317.
  19. 19. Udhaya Kumar S, Thirumal Kumar D, Bithia R, Sankar S, Magesh R, Sidenna M, et al. Analysis of Differentially Expressed Genes and Molecular Pathways in Familial Hypercholesterolemia Involved in Atherosclerosis: A Systematic and Bioinformatics Approach. Front Genet. 2020;11:734. Epub 2020/08/08. pmid:32760426; PubMed Central PMCID: PMC7373787.
  20. 20. Bhattacharya S, Srisuma S, Demeo DL, Shapiro SD, Bueno R, Silverman EK, et al. Molecular biomarkers for quantitative and discrete COPD phenotypes. American journal of respiratory cell and molecular biology. 2009;40(3):359–67. Epub 2008/10/14. pmid:18849563; PubMed Central PMCID: PMC2645534.
  21. 21. Singh D, Fox SM, Tal-Singer R, Bates S, Riley JH, Celli B. Altered gene expression in blood and sputum in COPD frequent exacerbators in the ECLIPSE cohort. PloS one. 2014;9(9):e107381–e. pmid:25265030.
  22. 22. Gautier L, Cope L, Bolstad BM, Irizarry RA. affy—analysis of Affymetrix GeneChip data at the probe level. Bioinformatics (Oxford, England). 2004;20(3):307–15. pmid:14960456
  23. 23. Barabási A-L, Gulbahce N, Loscalzo J. Network medicine: a network-based approach to human disease. Nature Reviews Genetics. 2011;12(1):56–68. pmid:21164525
  24. 24. Assenov Y, Ramírez F, Schelhorn SE, Lengauer T, Albrecht M. Computing topological parameters of biological networks. Bioinformatics (Oxford, England). 2008;24(2):282–4. Epub 2007/11/17. pmid:18006545.
  25. 25. Bo L, Wei B, Wang Z, Li C, Gao Z, Miao Z. Bioinformatic analysis of gene expression profiling of intracranial aneurysm. Molecular medicine reports. 2018;17(3):3473–80. Epub 2017/12/29. pmid:29328431.
  26. 26. Dong W, Qiu C, Gong D, Jiang X, Liu W, Liu W, et al. Proteomics and bioinformatics approaches for the identification of plasma biomarkers to detect Parkinson’s disease. Exp Ther Med. 2019;18(4):2833–42. pmid:31572530
  27. 27. Lu Y, Li Y, Li G, Lu H. Identification of potential markers for type 2 diabetes mellitus via bioinformatics analysis. Mol Med Rep. 2020;22(3):1868–82. pmid:32705173
  28. 28. Liu J, Meng H, Li S, Shen Y, Wang H, Shan W, et al. Identification of Potential Biomarkers in Association With Progression and Prognosis in Epithelial Ovarian Cancer by Integrated Bioinformatics Analysis. Frontiers in Genetics. 2019;10(1031). pmid:31708970
  29. 29. Zhang J, Zhu C, Gao H, Liang X, Fan X, Zheng Y, et al. Identification of biomarkers associated with clinical severity of chronic obstructive pulmonary disease. PeerJ. 2020;8:e10513. pmid:33354437
  30. 30. Liu X, Qu J, Xue W, He L, Wang J, Xi X, et al. Bioinformatics-based identification of potential microRNA biomarkers in frequent and non-frequent exacerbators of COPD. International journal of chronic obstructive pulmonary disease. 2018;13:1217–28. pmid:29713155.
  31. 31. Miao TW, Xiao W, Du LY, Mao B, Huang W, Chen XM, et al. High expression of SPP1 in patients with chronic obstructive pulmonary disease (COPD) is correlated with increased risk of lung cancer. FEBS Open Bio. 2021. Epub 2021/02/25. pmid:33626243.
  32. 32. Yao Y, Gu Y, Yang M, Cao D, Wu F. The Gene Expression Biomarkers for Chronic Obstructive Pulmonary Disease and Interstitial Lung Disease. Frontiers in genetics. 2019;10:1154–. pmid:31824564.
  33. 33. Kuzmanov U, Emili A. Protein-protein interaction networks: probing disease mechanisms using model systems. Genome Medicine. 2013;5(4):37. pmid:23635424
  34. 34. Sharma D, Surolia A. Degree Centrality. In: Dubitzky W, Wolkenhauer O, Cho K-H, Yokota H, editors. Encyclopedia of Systems Biology. New York, NY: Springer New York; 2013. p. 558–.
  35. 35. Zhong S, Xie D. Gene Ontology analysis in multiple gene clusters under multiple hypothesis testing framework. Artificial Intelligence in Medicine. 2007;41(2):105–15. pmid:17913480
  36. 36. Kammerl IE, Dann A, Mossina A, Brech D, Lukas C, Vosyka O, et al. Impairment of Immunoproteasome Function by Cigarette Smoke and in Chronic Obstructive Pulmonary Disease. American journal of respiratory and critical care medicine. 2016;193(11):1230–41. Epub 2016/01/13. pmid:26756824.
  37. 37. Tran I, Ji C, Ni I, Min T, Tang D, Vij N. Role of Cigarette Smoke-Induced Aggresome Formation in Chronic Obstructive Pulmonary Disease-Emphysema Pathogenesis. American journal of respiratory cell and molecular biology. 2015;53(2):159–73. pmid:25490051.
  38. 38. Debigaré R, Marquis K, Côté CH, Tremblay RR, Michaud A, LeBlanc P, et al. Catabolic/anabolic balance and muscle wasting in patients with COPD. Chest. 2003;124(1):83–9. Epub 2003/07/11. pmid:12853506.
  39. 39. Caramori G, Casolari P, Adcock I. Role of transcription factors in the pathogenesis of asthma and COPD. Cell communication & adhesion. 2013;20(1–2):21–40. pmid:23472830
  40. 40. Szulakowski P, Crowther AJL, Jiménez LA, Donaldson K, Mayer R, Leonard TB, et al. The Effect of Smoking on the Transcriptional Regulation of Lung Inflammation in Patients with Chronic Obstructive Pulmonary Disease. American journal of respiratory and critical care medicine. 2006;174(1):41–50. pmid:16574938.
  41. 41. Churg A, Zhou S, Wang X, Wang R, Wright JL. The role of interleukin-1beta in murine cigarette smoke-induced emphysema and small airway remodeling. American journal of respiratory cell and molecular biology. 2009;40(4):482–90. Epub 2008/10/22. pmid:18931327.
  42. 42. Simpson JL, McDonald VM, Baines KJ, Oreo KM, Wang F, Hansbro PM, et al. Influence of age, past smoking, and disease severity on TLR2, neutrophilic inflammation, and MMP-9 levels in COPD. Mediators of inflammation. 2013;2013.
  43. 43. Osei ET, Brandsma C-A, Timens W, Heijink IH, Hackett T-L. Current perspectives on the role of interleukin-1 signalling in the pathogenesis of asthma and COPD. European Respiratory Journal. 2020;55(2):1900563. pmid:31727692
  44. 44. Mycroft K, Krenke R, Górska K. Eosinophils in COPD—Current Concepts and Clinical Implications. The Journal of Allergy and Clinical Immunology: In Practice. 2020;8(8):2565–74. pmid:32251737
  45. 45. Eltboli O, Bafadhel M, Hollins F, Wright A, Hargadon B, Kulkarni N, et al. COPD exacerbation severity and frequency is associated with impaired macrophage efferocytosis of eosinophils. BMC pulmonary medicine. 2014;14(1):112. pmid:25007795
  46. 46. Oliver B, Tonga K, Darley D, Rutting S, Zhang X, Chen H, et al. COPD treatment choices based on blood eosinophils: are we there yet? Breathe (Sheff). 2019;15(4):318–23. pmid:31803266.
  47. 47. Artigas MS, Wain LV, Shrine N, McKeever TM, BiLEVE UK, Sayers I, et al. Targeted Sequencing of Lung Function Loci in Chronic Obstructive Pulmonary Disease Cases and Controls. PloS one. 2017;12(1):e0170222. pmid:28114305
  48. 48. Staley JR, Blackshaw J, Kamat MA, Ellis S, Surendran P, Sun BB, et al. PhenoScanner: a database of human genotype-phenotype associations. Bioinformatics (Oxford, England). 2016;32(20):3207–9. Epub 2016/06/17. pmid:27318201.
  49. 49. Chen W, Brehm JM, Manichaikul A, Cho MH, Boutaoui N, Yan Q, et al. A genome-wide association study of chronic obstructive pulmonary disease in Hispanics. Annals of the American Thoracic Society. 2015;12(3):340–8. pmid:25584925.
  50. 50. Hall R, Hall IP, Sayers I. Genetic risk factors for the development of pulmonary disease identified by genome-wide association. Respirology. 2019;24(3):204–14. pmid:30421854
  51. 51. de Torres JP, Marín JM, Casanova C, Cote C, Carrizo S, Cordoba-Lanus E, et al. Lung cancer in patients with chronic obstructive pulmonary disease—incidence and predicting factors. American journal of respiratory and critical care medicine. 2011;184(8):913–9. Epub 2011/07/30. pmid:21799072.
  52. 52. Parris BA, O’Farrell HE, Fong KM, Yang IA. Chronic obstructive pulmonary disease (COPD) and lung cancer: common pathways for pathogenesis. J Thorac Dis. 2019;11(Suppl 17):S2155–S72. pmid:31737343.
  53. 53. Wu L, Wang T, He D, Li X, Jiang Y. EVI‑1 acts as an oncogene and positively regulates calreticulin in breast cancer. Mol Med Rep. 2019;19(3):1645–53. Epub 2018/12/29. pmid:30592274; PubMed Central PMCID: PMC6390023.
  54. 54. Wieser R. The oncogene and developmental regulator EVI1: expression, biochemical properties, and biological functions. Gene. 2007;396(2):346–57. Epub 2007/05/18. pmid:17507183.
  55. 55. Daghistani M, Marin D, Khorashad JS, Wang L, May PC, Paliompeis C, et al. EVI-1 oncogene expression predicts survival in chronic-phase CML patients resistant to imatinib treated with second-generation tyrosine kinase inhibitors. Blood. 2010;116(26):6014–7. Epub 2010/09/22. pmid:20855863.
  56. 56. Yasui K, Konishi C, Gen Y, Endo M, Dohi O, Tomie A, et al. EVI1, a target gene for amplification at 3q26, antagonizes transforming growth factor-β-mediated growth inhibition in hepatocellular carcinoma. Cancer Sci. 2015;106(7):929–37. Epub 2015/05/12. pmid:25959919; PubMed Central PMCID: PMC4520646.
  57. 57. Choi YW, Choi JS, Zheng LT, Lim YJ, Yoon HK, Kim YH, et al. Comparative genomic hybridization array analysis and real time PCR reveals genomic alterations in squamous cell carcinomas of the lung. Lung Cancer. 2007;55(1):43–51. Epub 2006/11/18. pmid:17109992.
  58. 58. Wu K, Zhang X, Li F, Xiao D, Hou Y, Zhu S, et al. Frequent alterations in cytoskeleton remodelling genes in primary and metastatic lung adenocarcinomas. Nature Communications. 2015;6(1):10131. pmid:26647728
  59. 59. Dajon M, Iribarren K, Cremer I. Toll-like receptor stimulation in cancer: A pro- and anti-tumor double-edged sword. Immunobiology. 2017;222(1):89–100. Epub 2016/06/29. pmid:27349597.
  60. 60. Gu J, Liu Y, Xie B, Ye P, Huang J, Lu Z. Roles of toll-like receptors: From inflammation to lung cancer progression. Biomed Rep. 2018;8(2):126–32. Epub 2017/12/28. pmid:29435270.
  61. 61. Xu Y, Liu H, Liu S, Wang Y, Xie J, Stinchcombe TE, et al. Genetic variant of IRAK2 in the toll-like receptor signaling pathway and survival of non-small cell lung cancer. International Journal of Cancer. 2018;143(10):2400–8. pmid:29978465
  62. 62. Segundo-Val IS, Sanz-Lozano CS. Introduction to the Gene Expression Analysis. Methods Mol Biol. 2016;1434:29–43. Epub 2016/06/15. pmid:27300529.
  63. 63. Zhang X, Dang Y, Li P, Rong M, Chen G. Expression of IRAK1 in lung cancer tissues and its clinicopathological significance: a microarray study. Int J Clin Exp Pathol. 2014;7(11):8096–104. Epub 2015/01/01. pmid:25550857; PubMed Central PMCID: PMC4270603.
  64. 64. Jain A, Kaczanowska S, Davila E. IL-1 Receptor-Associated Kinase Signaling and Its Role in Inflammation, Cancer Progression, and Therapy Resistance. Frontiers in immunology. 2014;5(553). pmid:25452754
  65. 65. Xu X, Liu S, Ji X. Overexpression of ecotropic viral integration site-1 is a prognostic factor of lung squamous cell cancer. Onco Targets Ther. 2017;10:2739–44. pmid:28603423.