Single-cell RNA-seq analysis to identify potential biomarkers for diagnosis, and prognosis of non-small cell lung cancer by using comprehensive bioinformatics approaches

Highlights

• •DEGs were calculated from the publicly available scRNA seq data of NSCLC patients.
• •GO, KEGG, and PPI network analyses led to identification of 12 key genes.
• •Seven key transcription factors and 8 key miRNAs, were also identified.
• •Our finding might play a significant role as candidate biomarkers in NSCLC diagnosis and prognosis.

Abstract

Non-small cell lung cancer (NSCLC) is the most common type of lung cancer and the leading cause of cancer-related deaths worldwide. Identification of gene biomarkers and their regulatory factors and signaling pathways is very essential to reveal the molecular mechanisms of NSCLC initiation and progression. Thus, the goal of this study is to identify gene biomarkers for NSCLC diagnosis and prognosis by using scRNA-seq data through bioinformatics techniques. scRNA-seq data were obtained from the GEO database to identify DEGs. A total of 158 DEGs (including 48 upregulated and 110 downregulated) were detected after gene integration. Gene Ontology enrichment and KEGG pathway analysis of DEGs were performed by FunRich software. A PPI network of DEGs was then constructed using the STRING database and visualized by Cytoscape software. We identified 12 key genes (KGs) including MS4A1, CCL5, and GZMB, by using two topological methods based on the PPI networking results. The diagnostic, expression, and prognostic potentials of the identified 12 key genes were assessed using the receiver operating characteristics (ROC) curve and a web-based tool, SurvExpress. From the regulatory network analysis, we extracted the 7 key transcription factors (TFs) (FOXC1, YY1, CEBPB, TFAP2A, SREBF2, RELA, and GATA2), and 8 key miRNAs (hsa-miR-124-3p, hsa-miR-34a-5p, hsa-miR-21-5p, hsa-miR-155-5p, hsa-miR-449a, hsa-miR-24-3p, hsa-let-7b-5p, and hsa-miR-7-5p) associated with the KGs were evaluated. Functional enrichment and pathway analysis, survival analysis, ROC analysis, and regulatory network analysis highlighted crucial roles of the key genes. Our findings might play a significant role as candidate biomarkers in NSCLC diagnosis and prognosis.

Graphical abstract

Abbreviations

NSCLC

Non-Small Cell Lung Cancer

scRNA-seq

Single-cell RNA sequencing

GEO

Gene Expression Omnibus

DEGs

Differentially expressed genes

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

PPI

Protein-protein interaction

KGs

Key Genes

ROC

Receiver Operating Characteristics

TF

Transcription factors

LUSC

Lung squamous cell carcinoma

LUAD

Lung adenocarcinoma

GAPDH

Glyceraldehyde-3 phosphate dehydrogenase

t-SNE

t-distributed stochastic neighbor embedding

STRING

Search Tool for the Retrieval of Interacting Genes

AUC

Area under curve

logFC

Log of Fold Change

CI

Confidence Interval

SVM

Support Vector Machine

SCLC

Small Cell Lung Cancer

Introduction

Cancer, a heterogeneous disease, poses a serious challenge for precise treatment at the individual level. Both bulk and single-cell RNA sequencing (scRNA-seq) technologies are used for studying transcriptional profiles at the gene expression level. Several articles have identified biomarkers for NSCLC diagnosis using bulk RNAseq technology [1], [2], [3], [4], [5]. scRNA-sequencing categorizes the cell types across multiple tissues, whereas bulk RNA sequencing involves the use of a tissue or cell population [6,7]. scRNA-seq is widely used to determine tumor heterogeneity, cellular identities, novel biomarkers, and molecular and functional strategies [8]. Several scRNA-seq based studies have been performed earlier to explore tumor heterogeneity and to identify novel biomarkers for different cancers [9], [10], [11], [12].

Non-small cell lung cancer (NSCLC) is a highly heterogeneous lung cancer, accounting for approximately 85% of all the types of lung cancers, and it is strongly correlated with smoking habits [13,14]. NSCLC is mainly classified into two groups: lung squamous cell carcinoma (LUSC) and lung adenocarcinoma (LUAD). Most of the current treatment strategies for NSCLC are chemotherapies based on the histology and targeted agents for patients [15,16]. Treatment outcomes of NSCLC are quite insufficient as the post-therapy relapse rate and drug resistance remains high, and the 5-year relative survival rate of the patients is 26% [17,18]. Further exploration into the underlying mechanisms of NSCLC is thus quite urgent, which will impact the discovery of novel diagnostics and provide effective & key targets for NSCLC.

Expanding research recommends that affluent genes, miRNAs, TFs, and/or biological enriched pathways are associated with the development and progression of cancers. The altered expression of miRNAs has a strong correlation with distinct disease and carcinoma [19]. Gene-miRNAs interactions have been generally exhibited to control the convoluted molecular systems’ basic oncogenesis, advancement, intrusion, and tumor metastasis [20]. Several studies have recently shown a link of miRNA in cancer development [21]. Some studies have highlighted the importance of miRNA expression in lung tissues [22]. The deregulation of miRNA might play a role in fatal NSCLC progression [23]. Thus, several studies have identified some important TFs and miRNAs as transcriptional and post-transcriptional factors for cancer through the regulatory interaction networks [24,25].

In the last decade, several biomarkers have been reported as prognostic and predictive markers for NSCLC [2,[26], [27], [28], [29], [30]]. Song et al. identified STAT3,  EGFR, PTEN, KRAS, TP53, RHOA, CTNNB1, and VEGFA, as efficient targeted genes associated with six miRNAs (hsa-miR-21-5p, hsa-miR-31-5p, hsa-miR-708-5p, hsa-miR-30a-5p, hsa-miR-451a, and has-miR-126-3p) by expression analysis and miRNA-hub gene network for NSCLC [2]. Chen et al. discovered four genes viz. CDK1, PLK1, RAD51, and RFC4 as novel biomarkers using microarray gene expression profiles that might be potential therapeutic targets in NSCLC [26]. Valk et al. identified SPAG5, POLH, KIF23, RAD54L, SGCG, NLRC4, MMRN1, and SFTPD as the novel genes that are involved in multiple pathways leading to NSCLC [31]. Puzone et al. showed that the overexpression of glyceraldehyde-3 phosphate dehydrogenase (GAPDH) correlates with poor prognosis in NSCLC patients [32]. These findings are important to understand NSCLC pathogenesis.

Although much research has already been conducted to reveal the molecular mechanism of NSCLC progression, the heterogeneity and complexity of NSCLC still poses a great challenge and need for novel and effective biomarkers. Recently, scRNA-seq technology has been used to detect tumor heterogeneity and explore the gene expression pattern in tissues that can help the researcher to detect the novel biomarkers. Here, we utilized computational models for analyzing the scRNA-seq data to reveal the tumor heterogeneity of NSCLC tissues. We identified differentially expressed genes (DEGs), their associated pathways, and PPI network to screen the key genes (KGs), key miRNAs, and key TFs for personalized diagnosis and prognosis of NSCLC and performed further analysis to validate the result. Thus, the identification of important genes, miRNAs, and TFs as well as the signaling pathways related to cancer via bioinformatics analysis, will provide worthy enlightenment in cancer research.

Materials and methods

scRNA-seq data collection and processing

The publicly available scRNA-seq data (GSE127471, data collected from the peripheral blood of a patient with NSCLC by Newman and the team) were downloaded from the gene expression omnibus (GEO) database [33]. The data were sequenced on Illumina NextSeq 500 (Human). Data processing was performed using the Seurat package V3.1.1 in R V3.6.1 [34]. For quality control check, we extracted genes with a minimum number of features 200 having non zero counts and a minimum number of cells as 3. The filtered data were normalized using log-transformation and was used for further analysis. We used two datasets associated with NSCLC from the GEO database with accession numbers GSE19188 and GSE75037 to assess the diagnostic performance of the identified KGs.

Clustering and DEG identification

For dimensionality reduction, we performed principal component analysis (PCA) on the scaled data. The t-distributed stochastic neighbor embedding (t-SNE) was used to demonstrate two-dimensional data by first 10 principal components. The cell cluster was identified using K-means clustering based on the original Louvain algorithm. We used the Seurat's VlnPlot function to determine the expression of acquainted marker genes to assign clusters. Moreover, we constructed the trajectory analysis to reveal the tendency curve of the eight clusters using "Monocle" package [35]. The R package Seurat was used to analyze DEGs with scRNA-seq data.

Functional enrichment and pathway analysis of DEGs

We used a stand-alone software tool FunRich (version 3.1.4) for gene ontology (GO) functional enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of DEGs [36]. We considered the threshold P-value < 0.01 to obtain significant functional and pathway terms. KEGG and GO were used to annotate the enrichment analysis of bio-term classification processes and genes clusters as well as to impart a fantasy module for inference [37,38].

PPI network analysis of DEGs

Protein-protein interactions (PPIs) networks facilitate the analysis of pathogenic mechanisms and disease progression by providing knowledge on the molecular mechanism underlying cellular activity. In this study, we used the Search Tool for the Retrieval of Interacting Genes (STRING v11.0) database for constructing PPI network of DEGs [39]. Cytoscape (version 3.7.1) was used to discern the PPI networks between the DEGs [40]. We identify 12 KGs using two topological methods Betweenness and Stress in Cytoscape plugin cytoHubba [41].

Survival analysis and expression level of KGs

We used SurvExpress to check the impact of the expression pattern and survival analysis of KGs. SurvExpress (http://bioinformatica.mty.itesm.mx/SurvExpress) is an online tool for analyzing cancer gene expression data for the validation and survival analysis of multi-gene biomarkers [42]. Here we used it to verify and estimate the impact the expression pattern and prognostic value of KGs using the Kaplan-Meier curve and log-rank test.

ROC analysis of the KGs

Receiver operating characteristic (ROC) curve analysis was carried out to assess the true positive rate (Sensitivity) and false-positive rate (1− Specificity) of the identified KGs using the "pROC" package in R [43]. Area under curve (AUC) was determined and used to screen the ROC values.

Regulatory interaction network analysis of KGs

We constructed a regulatory interaction network (KGs-miRNAs and KGs-TFs) of 12 KGs using an online tool NetworkAnalyst 3.0 [44]. It is a bioinformatics tool that visualizes and deciphers the information in the association of network settings. This tool included three gene-miRNAs interaction databases TarBase [45], miRTarbase [46] and miRecords [47] and three gene-TFs interaction databases such as ENCODE [48], JASPAR [49], and ChEA [50]. We used TarBase and JASPAR databases for KGs-miRNAs and KGs-TFs interaction networks, respectively.

Results

Tumor heterogeneity and identification of DEGs

The scRNA-seq technology provided a good transcriptional detailing of cancer cells and gene expression in NSCLC patients. The 1803 cells were adopted for advanced analysis. The 1803 cells were analyzed and then classified into 8 separate clusters based on the identical gene set. Every cluster was separated by identical gene set. To make the segregation clearer among clusters, DEG analysis in cell types based on their covariance patterns and mean expression levels were evolved. We found that the identified marker gene sets were significant to ascertain cell types individually with high potentiality. Heatmap of the top ten genes showing high heterogeneity among clusters based on logFC has been illustrated in Fig. 1(A). The results of trajectory analysis showed the cells of only cluster1 may have a significant difference from other cells in NSCLC Fig. 1(B). We identified DEGs using the Wilcoxon rank sum test, based on the threshold adj.P.Val < 0.01 and logFC > 2 for up-regulated genes and adj.P.Val < 0.01 and logFC < -2 for down-regulated genes. A total of 158 DEGs, of which 48 were up-regulated and 110 down-regulated genes were identified. All DEGs (up- and down-regulated) are listed in Table 1.

Fig. 1.

(A) The heatmap of the top 10 marker genes of each cluster where each row represents genes and column represents clusters. (B) Trajectory plot differentiated by eight clusters.

Table 1.

Differentially expressed genes (DEGs) in NSCLC.

Up-regulated DEGs

Down-regulated DEGs

TREML1, TAGLN2,LTB*,CCL5*, FTL, MNDA, LST1, SPARC, AIF1, ACRBP, AHSP,MS4A1*, CD79B, VCAN, ALAS2,CD74*, CST3, FCN1, PRF1, KLRB1,GZMB*, FGFBP2,TYROBP*, CMC1, CTSS, NRGN, RGS18, MYL9, HBD, CLU, TUBB1, SDPR, HIST1H2AC, S100A12, GNG11, CD79A, RP11-1143G9.4, CA1, LYZ, S100A9, S100A8, HBA2, GNLY, HBB, IGLL5, PF4,HBA1*, PPBP

MALAT1, RPS29, RPL39, MT-ND3, RPS27, RPS28, RPL37, RPL34, RPS26,ACTB*, RPS21, RPL23A, RPL36, MT-ND2, PTMA,EEF1A1*, MT-CO1, RPL28, RPL41, RPL38, RPL26, TMSB4X, HLA-C, RPL10, TMSB10, RPLP2, MT-CO3, MT-ND4, MT-ND1, RPL3,RPS18*, RPL27A, HA2, S100A6, RPS2, RPL27, RPS25, RPL36A, MT-ATP6, RPS8, B2M, RPS15A, HLA-B, RPL17, RPS24, RPL23,RPLP1*, MT-CO2, RPS19, RPL13A, RPL35A, MT-CYB, RPSA, RPS23, RPS3, RPL35, RPS10, RPS16, RPS15, RPS6, VIM, RPL32, RPS13, RPL13, S100A4, RPS4X, RPL30, RPL31, NBEAL1, TXNIP, TPT1, HLA-A, FAU, RPS7, NEAT1,GAPDH*, RPS3A, RPSAP58, RPS9, RPL19, RPL12, RPL15, RPL9, RPL11, RPS14, RPL6, ACTG1, H3F3A, RPL7, RPL7A, GNB2L1, RPS20, RPL5, PFN1, RPL29, RPL18, NKG7, RPL24, RPL22, TOMM7, RPL14, PABPC1, BTG1, RPL21,CFL1, RPS11, SRGN, RPL10A, RPL18A, HNRNPA1

N.B. Bold with star (*) indicates key genes (KGs).

Functional enrichment and pathway analysis of DEGs

For further investigation, the GO functional and KEGG pathway enrichment analyses associated with the DEGs was performed in FunRich software. The top GO (BP, CC, and MF) and KEGG enrichment functions/terms of DEGs are shown in Fig. 2. In BP, 47.3% genes were enriched with protein metabolism, 8.90% genes were enriched in immune response, and 9.5% genes were enriched with cell growth and/or maintenance. In CC, 54.4% of genes were enriched in cytosol and cytoplasm; more than 40% of genes were enriched in exosomes, nucleolus, and ribosome; and approximately 30% of genes were enriched with cytosolic large ribosomal subunit, extracellular, lysosome, etc., where the MF enrichment was mainly correlated with structural constituent of ribosome, MHC class I receptor activity, MHC class II receptor activity, B cell receptor activity, and chemokine activity terms in Table S1. The enriched KEGG pathways for the DEGs included peptide chain elongation, eukaryotic translation elongation, eukaryotic translation termination, viral mRNA translation, 3′ -UTR-mediated translational regulation, and metabolism of proteins pathways, which are associated with lung cancer development (Table S2).

Fig. 2.

Top GO and KEGG terms enriched by DEGs ((A) Biological processes, (B) Cellular components, (C) Molecular functions, and (D) Kyoto encyclopedia of genes and genome (KEGG)).

PPI network of DEGs

Based on the online tool STRING, a total of 158 DEGs were used in the PPI network, involving 156 nodes and 3282 edges with an average node degree 42.2 and PPI enrichment P-value < 1.0e-16 (Fig. 3). The top 15 genes were selected using each of the two network scoring methods Betweeness and Stress in Cytoscape plugin, cytoHubba. We then extracted 12 common genes from the top 15 genes of the two methods and considered them as KGs, such as MS4A1, CCL5, GZMB, HBA1, TYROBP, CD74, LTB, EEF1A1, RPLP1, RPS18, GAPDH, and ACTB, as shown in Table 2. In addition, we assessed the significance of the 12 KGs through survival analysis, cancer prediction model, and regulatory interaction network analysis.

Fig. 3.

Protein-protein interaction (PPI) network of DEGs. The green and yellow colors indicate up- and down-regulated genes respectively. The key genes (KGs) are highlighted in diamond shape.

Table 2.

The summary of the key genes (KGs) identified from the Cytoscape in NSCLC.

S.N	Betweenness		Stress		Key Genes
	Name	Score	Name	Score	MS4A1, CCL5, GZMB, HBA1, TYROBP, CD74, LTB, EEF1A1, RPLP1, RPS18, GAPDH, ACTB
1	GAPDH	6144.813	GAPDH	146026
2	MS4A1	2329.161	MS4A1	43470
3	CCL5	1094.994	CCL5	23206
4	ACTB	1056.283	HBA1	22566
5	GZMB	830.074	CD74	20158
6	HBA1	663.953	GZMB	20044
7	TYROBP	662.6554	ACTB	17804
8	CD74	634.4065	LTB	14866
9	HBA2	551.3868	TYROBP	12728
10	EEF1A1	474.6351	EEF1A1	11602
11	HBB	462.9838	RPS18	11596
12	RPS3	433.276	RPLP1	10574
13	RPLP1	407.0451	RPL13A	10386
14	LTB	399.4331	PPBP	10344
15	RPS18	378.9975	RPL19	10194

Survival analysis and expression level of KGs

The prognostic value of the identified KGs was evaluated by fitting Cox-proportional hazards regression model between high-risk and low risk group patients in Fig. 4(A), where red indicates high-risk group patients and green indicates low-risk group patients. We observed that the overall survival probability for the high-risk group compared to the low-risk group decreased over time based on the expression level of KGs (Hazard Ratio = 1.85 at 1.36-2.53 Confidence Interval (CI), and the log-Rank p-Value = 9.766e-05) This indicates that the proposed KGs have a strong prognostic power for NSCLC. The expression patterns of the KGs by the risk group are shown in the boxplot in Fig. 4(B). The overall results showed that the KGs displayed a significant prognostic performance for NSCLC that support our original results.

Fig. 4.

(A) Kaplan-Meier plot displaying the prognostic effect of the KGs on NSCLC. (B) Boxplot displaying the expression pattern of KGs between risk groups. Red indicates high-risk group and green indicates low-risk group.

Performance measure of the identified KGs using ROC analysis

A supervised machine learning algorithm SVM classifier was considered to develop a Cancer Prediction Model for 12 KGs. We developed the cancer prediction model through the ROC curve for the training dataset with access number GSE19188 (red) and for the test dataset with access number GSE75037 (green) in Fig. 5. We observed that the AUC values range from 0.64 to 0.99 for the training dataset and 0.63 to 0.97 for the test dataset, which indicate the good prediction performance.

Fig. 5.

ROC curve evaluating the diagnostic performance of the KGs in NSCLC. Red color indicates GSE75037 dataset and green color indicates GSE19188.

KGs-miRNAs and KGs-TFs interaction network analysis

To identify the key miRNAs associated with KGs, we constructed the interaction network between the KGs and miRNAs. We uploaded the official symbol in the gene list of human lung tissue and select the TarBase database in NetworkAnalyst to identify miRNAs that targeted the KGs (Fig. 6(A)). On the basis of degree score >= 4, we selected 8 key miRNAs (hsa-miR-124-3p, hsa-miR-34a-5p, hsa-miR-21-5p, hsa-miR-155-5p, hsa-miR-449a, hsa-miR-24-3p, hsa-let-7b-5p, and hsa-miR-7-5p) as post-transcriptional regulatory factors of KGs. From the literature review, we found that these miRNAs have a close relationship with drug resistance of lung cancer, basically NSCLC [51]. Similarly, the interaction network of KGs and TFs was constructed from the JASPAR database in NetworkAnalyst. Seven TFs (FOXC1, YY1, CEBPB, TFAP2A, SREBF2, RELA, and GATA2) with degree >= 4 were selected as transcriptional regulatory factors of KGs (Fig. 6(B)). Key miRNAs and TFs with their degree scores are listed in Table 3.

Fig. 6.

Gene-miRNAs and Gene-TFs interaction network. (A) In the Gene-miRNAs interaction network, the red bigger circles indicate genes and the blue squares indicate miRNAs. (B) In Gene-TFs interaction network, the red circles indicate genes and the blue diamond shape indicates TFs.

Table 3.

Targeted transcription factors (TFs) and miRNAs from regulatory interaction network (Gene-TFs and Gene-miRNAs).

S.N	TFs		S.N	miRNAs
	Label	Degree		Label	Degree
1	FOXC1	6	1	hsa-miR-124-3p	5
2	YY1	4	2	hsa-miR-34a-5p	5
3	CEBPB	4	3	hsa-miR-21-5p	4
4	TFAP2A	4	4	hsa-miR-155-5p	4
5	SREBF2	4	5	hsa-miR-449a	4
6	RELA	4	6	hsa-miR-24-3p	4
7	GATA2	4	7	hsa-let-7b-5p	4
			8	hsa-miR-7-5p	4

Discussion

NSCLC is a disease with a very high concern and is life-threatening for humans. Due to the heterogeneity of NSCLC, its treatment is very challenging. Hence, it will be helpful if NSCLC is managed by targeted treatment; however, patients with NSCLC have a lower prognosis. Hence, the identification of novel biomarkers based on the heterogeneity and the scRNA-seq data are one of the key tasks to improve the personalized and targeted medicine of NSCLC in the future.

In this study, we analyzed scRNA-seq data from the tumor tissue of peripheral blood of NSCLC patients to bioinformatically explore the cellular heterogeneity, DEGs, associated biological pathways, PPI network, key genes, miRNAs and TFs. The 1803 cells were classified into 8 explicit clusters while each cluster was mixed up with variant numbers of cells. By comparing gene expression profiles, a total 158 DEGs containing 48 up- and 110 down-regulated genes were found. To infer the biological functions and pathways associated with NSCLC, GO and KEGG pathway enrichment analyses were performed. We identified 12 KGs (MS4A1, CCL5, GZMB, HBA1, TYROBP, CD74, LTB, EEF1A1, RPLP1, RPS18, GAPDH, and ACTB) based on the two methods of Betweenness and Stress using PPI network results, and the expression pattern and survival analysis of KGs were affirmed on the basis of the TCGA data. Using the TarBase and JASPAR databases in NetworkAnalyst, we identified 8 key miRNAs (hsa-miR-124-3p, hsa-miR-34a-5p, hsa-miR-21-5p, hsa-miR-155-5p, hsa-miR-449a, hsa-miR-24-3p, hsa-let-7b-5p, hsa-miR-7-5p) and 7 key TFs (FOXC1, YY1, CEBPB, TFAP2A, SREBF2, RELA, GATA2).

All the identified 12 KGs have been supported by different studies for lung and other cancers. The deregulation of MS4A1 in lung squamous cell cancers might occur because of the expression of CD20 stromal lymphocytes [52]. Another study showed MS4A1 as a prognostic biomarker for LUAD [53]. hsa-miR-147a can inhibit the outgrowth and metastasis of NSCLC by aiming the CCL5 gene [54]. It has been found that GZMB is significantly associated with poor prognosis in SCLC [55]. Over expression of HBA1 showed low overall survival in non-smoker female lung cancer patients [56]. YAP1 promotes multidrug resistance of SCLC through signaling pathways associated with CD74 [57]. LTB-4 participates in the recruitment of neutrophils in the airways at NSCLC [58]. ACTB, EEF1A1, and RPS18 are reported to be relevant genes for qRT-PCR analysis of lung cancer also EEF1A1 is responsible for lung cancer development in smokers [59], [60], [61]. Aberrant methylation and high expression of GAPDH are associated with poor prognosis in LUAD patients [62]. TYROBP is a novel key gene with prognostic value in gastric cancer by integrated network analysis [63]. The gene biomarker RPLP1 is an anti-metastasis candidate therapeutic target with poor prognosis in triple-negative breast cancer [64]. To our knowledge TYROBP and RPLP1 are not yet been reported for lung cancer progression. Hence, we can say that these two genes (TYROBP and RPLP1) are novel with good prognostic value in our study for NSCLC.

Furthermore, hsa-miR-124-3p is reported as a tumor suppressor and inhibits the progression of several tumors, including NSCLC [65], [66], [67]. hsa-miR-34a-5p resists the brainstem glioma cell invasion [68]. hsa-miR-21-5p is one of the most important prognostic biomarkers for NSCLC [69]. hsa-miR-155-5p, hsa-miR-24-3p, and hsa-let-7a-5p were reported to be up-regulated in LUAD tissues [70]. hsa-miR-449a is reported as a genetic risk factor for gastric cancer [71]. hsa-miR-7-5p is a prognostic biomarker for small cell lung cancer [72]. Among the identified TFs, FOXC1 is one of the pioneer TF and plays an important role in the development of lung, breast, and prostate cancer [73]. The expression of FOXC1 is increased in NSCLC tissues, and it has an adverse relationship with survival [74]. TFAP2C is contributed to NSCLC tumorigenesis by downregulating numerous tumor silencers such as GADD45B, PMAIP1, and XAF1 [75].

The GO functional enrichment and KEGG pathway analysis revealed that the KGs are related to Protein metabolism, MHC class I receptor activity, B cell receptor activity, viral mRNA Translation, 3′ -UTR-mediated translational regulation etc. pathways. Most of the DEGs (47.3%) have enriched with protein metabolism (associated with KGs: EEF1A1, RPS18, RPLP1, and GZMB) term, and several studies have further claimed that NSCLC-causing genes are enriched in protein metabolism term [76,77]. The prognostic effect of the KGs with TCGA datasets in LUAD showed the worst survival rate which indicates that these KGs might be the prognostic biomarkers in LUAD. The differential expression stated the discriminating power of the KGs. Finally, the diagnostic effects of KGs were assessed by ROC analysis. The AUC values in ROC analysis indicated a comparatively good prediction performance of the KGs in NSCLC patients with higher sensitivity and specificity. Therefore, our overall analysis will provide valuable insights into NSCLC progression, KGs, key miRNAs, key TFs might be a novel diagnostic and prognostic biomarkers as well as potential regulators for the progression, diagnosis, and prognosis of NSCLC. In this study, we predicted the results through computational analysis; hence, we cannot recommend for treatment directly. We emphasize for further assessed at the molecular level by the wet-lab experiments in prior to clinical investigation.

Conclusion

The scRNA-seq data of peripheral blood cell allows the identification of distinct cell types and provides a new perspective on the pathogenesis of NSCLC. On the viewpoint of clustering analysis, we conclude that NSCLC is heterogeneous in numerous aspects. Through bioinformatics analysis, we identified 12 KGs (MS4A1, CCL5, GZMB, HBA1, TYROBP, CD74, LTB, EEF1A1, RPLP1, RPS18, GAPDH, and ACTB); among them, there were 2 novel KGs (TYROBP and RPLP1). Their targeted miRNAs and TFs were also identified, which play a significant role in NSCLC. Survival and ROC analysis showed the prognostic and diagnostic effect of KGs. Our overall findings suggested that these KGs might be the prognostic and diagnostic biomarkers for NSCLC.

Ethical Approval and Consent to participate

Not Applicable

Consent for publication

All the authors have provided consent for the publication.

Availability of supporting data

The scRNA-seq data of NSCLC were obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) with accession number GSE127471.

CRediT authorship contribution statement

Adiba Sultana: Data curation, Formal analysis, Writing – original draft, Writing – review & editing. Md Shahin Alam: Data curation, Writing – original draft, Writing – review & editing. Xingyun Liu: Formal analysis, Writing – review & editing. Rohit Sharma: Formal analysis, Writing – review & editing. Rajeev K. Singla: Formal analysis, Writing – original draft, Writing – review & editing. Rohit Gundamaraju: Formal analysis, Writing – review & editing. Bairong Shen: Conceptualization, Funding acquisition, Supervision, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.