RNA-seq data from 442 LUAD patients (including 442 primary tumor samples and 19 adjacent normal tissue samples) together with the associated clinical information was downloaded from the GSE68465 cohort (http://www.ncbi.nlm.nih.gov/geo/). EMT-related genes were downloaded from the EMTome database [21].

The expression of genes between tumor and normal lung tissue in the GSE68465 cohort was compared using the “limma” package in R. Those with a $|$log fold change$|$ $>$2 and adjusted p-value 0.05 were considered to be differentially expressed genes (DEGs). DEGs in the GEO database and EMT-related genes in the EMTome database were intersected to obtain EMT-related DEGs. The DAVID database was used for functional enrichment analysis of these EMT-related DEGs [22].

Genes that were most strongly associated with prognosis were identified using the “survival” and “glement” packages in R. Univariate, multivariate, lasso Cox regression analysis and Kaplan-Meier survival analysis were performed to screen for prognostically significant EMT-associated genes. Correlation analysis of EMT-related DEGs was performed in GEPIA [23].

The genomic characteristics, including genomic and subcellular locations, of EMT-related DEGs with prognostic significance were first examined in GeneCards (https://www.genecards.org/) [24] and COMPARTMENTS (https://compartments.jensenlab.org/Search) [25]. Next, the differential expression of EMT-related DEGs with prognostic value was compared between tumor and normal lung tissue from the GSE68465 cohort using the “limma” package in R. Transcription levels for EMT-related DEGs with prognostic significance were then compared between pan-cancer and normal tissue in the TIMER database (https://cistrome.shinyapps.io/timer/) [26]. Protein expression levels of EMT-related DEGs with prognostic value in LUAD were also investigated in the Human Protein Atlas (HPA) database (http://www.proteinatlas.org). EMT-related genes were grouped according to high or low expression in the GSE68465 cohort, with the “clusterProfiler” package in R then used to perform functional enrichment analysis.

Univariate and multivariate Cox regression analyses were used to evaluate correlations between EMT-related DEG expression and various clinicopathological features in LUAD patients, including age, pathological T stage, TP53 mutation status, and pathological N stage. The prognostic significance of EMT-related DEG expression in clinicopathological subgroups of LUAD patients was also investigated in the UALCAN database [27].

The DNA methylation level and prognostic significance of EMT-related DEGs in LUAD were evaluated in the DNMIVD database [28]. Online databases were used to predict the regulatory relationship between target genes and miRNAs. The TargetScan database (https://www.targetscan.org/vert_80/) [29], miRDB database (http://www.miRDB.org) [30], and miRTarBase database (https://mirtarbase.cuhk.edu.cn/) [31]were used to predict potential miRNAs upstream of EMT-related DEGs. The predicted miRNAs from the three databases were then intersected to identify miRNAs that could regulate the expression of EMT-related DEGs. Finally, the relationship between EMT-related DEGs and the miRNA expression level was further evaluated using the starBase database [32].

The TIMER database was used to correlate the expression of EMT-related DEGs with tumor-infiltrating immune cells including CD4+ T cells, CD8+ T cells, macrophages, dendritic cells, B cells, and neutrophils. Immune cell markers were obtained from the website of R&D Systems (www.rndsystems.com/cn/resources/cell-markers/immune-cells) and analyzed in TIMER and GEPIA. The associations between EMT-associated DEGs and immune cell subtypes, tumor infiltrates (TILs), immune stimulants, immunosuppressants, and major histocompatibility class (MHC) molecules in LUAD were examined in the TISIDB database (cis.hku.hk/TISIDB) [33]. Kaplan–Meier analysis was used to investigate the prognostic significance of immune cell infiltration (https://kmplot.com/analysis/) [34].

The cBioPortal database (http://cbioportal.org) was used to examine the DNA mutation status of EMT-related DEGs in 586 LUAD samples from the TCGA database [35]. We explored mutation sites in EMT-related DEGs and assessed gene mutation co-occurrence patterns between EMT-related DEG signatures and other proteins in LUAD. The top 10 altered genes that co-occurred with mutated genes were selected according to the percentage of alteration. In addition, the mRNA expression level of EMT-related DEGs was confirmed in TCGA-LUAD samples from the UALCAN database.

The OmicStudio tool (https://www.omicstudio.cn/tool) was employed to construct ROC curves in order to assess the diagnostic value of EMT-related genes. A single gene or a combination of EMT-related genes was used to assess diagnostic value.

Human LUAD (H1975) and bronchial epithelial (BEAS-2B) cell lines were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). These cells had been performed by STR analysis and their DNA profiles were consistent with the recorded. Additionally, mycoplasma testing was conducted on the cell lines, and the result is negative. BEAS-2B cells were grown in DMEM (Bio-Channel, Nanjing, Jiangsu, China), while H1975 cells were grown in ERPMI-1640 medium (Bio-Channel, Nanjing, Jiangsu, China). All cells were supplemented with penicillin (100 U/mL), streptomycin (100 U/mL) and 10% heat-inactivated fetal bovine serum (Transgen, Beijing, China) and cultured in a humidified environment at 37 °C with 5% CO${}_2$.

Adherent cells were used in the logarithmic growth phase, which corresponded to a cell density of approximately 80%. TRIzol reagent (Invitrogen, CA, USA) was used to extract total RNA from the cells and transcription reagent (Transgen, Beijing, China) was then used to transcribe the RNA into cDNA. RT-PCR was performed in the 7500 Real-Time PCR System (BIO-RAD) using SYBR qPCR mix (Transgen, Beijing, China) and GAPDH as the endogenous reference [36]. The primer sequences used were: GAPDH, 5${}^{\prime }$-GTCTCCTCTGACTTCAACAGCG-3${}^{\prime }$ (forward) and 5${}^{\prime }$-ACCACCCTGTTGCTGTAGCCAA-3${}^{\prime }$ (reverse); GALNT3, 5${}^{\prime }$-GCTGCAGTTTCATGTTAGGG-3${}^{\prime }$ (forward) and 5${}^{\prime }$-ACAGCCCCAATAACCGTATGAA-3${}^{\prime }$ (reverse); OAS1, 5${}^{\prime }$-GCTGGCTGAAAGCAACAGTG-3${}^{\prime }$ (forward) and 5${}^{\prime }$-TCCAGTCCTCTTCTGCCTGT-3${}^{\prime }$ (reverse). The relative initial copy number of the target gene was calculated using the 2${}^{-△△CT}$ method.

R 4.1.1 (R studio Version 1.4.1717) and GraphPad Prism 8.0 (San Diego, California USA, www.graphpad.com) were used for statistical analysis. The Wilcoxon test was used to screen for genes that were differentially expressed. Prognostic factors were assessed using univariate, Lasso, and multivariate Cox regression analysis. The Kaplan–Meier method was used to evaluate the correlation between gene expression level and overall survival rate. The student’s t-test was used to analyze the RT-PCR results. The 95% confidence interval (CI) and hazard ratio (HR) was used to calculate overall survival. p 0.05 was used as the threshold for statistical significance.

The flow chart for this study is shown in Fig. 1A. Microarray-derived expression database for the GSE68465 LUAD cohort was screened by “limma” in R (adjusted p-value 0.05 and $|$logFC$|$ $>$2). A total of 855 DEGs were obtained, comprising 650 upregulated genes and 205 downregulated genes (Fig. 1B). The EMTome database is a resource that includes pan-cancer analysis of EMT genes and their associated characteristics. EMT signature genes were downloaded from the EMTome database. A total of 320 EMT-related DEGs were obtained by the intersection of 855 DEGs from the GEO database with 3600 EMT signature genes from the EMTome database (Fig. 1C). Subsequently, functional enrichment analysis and enrichment pathway analysis was performed on these EMT-related genes in the DAVID database. As shown in Supplementary Fig. 1A–C, EMT-related DEGs were mainly enriched in extracellular matrix organization via Biological Process (BP), extracellular space via Cellular Component (CC), and heparin-binding via Molecular Function (MF). The significant signaling pathways for EMT-related DEGs were mainly enriched in amoebiasis, the PI3K-Akt signaling pathway, and ECM-receptor interactions (Supplementary Fig. 1).

Fig. 1.

Flowchart for identifying EMT-related genes in LUAD using the GES68465 cohort and the EMTome database. (A) Flowchart used for data collection and the methods used in this study. (B) Volcano plots of DEGs in LUAD from the GSE68465 cohort. (C) Venn diagram of EMT-related genes in LUAD from the GSE68465 cohort and EMTome database. GEO, Gene Expression Omnibus; DEGs, differentially expressed genes; EMT, epithelial-mesenchymal transition; KEGG, Kyoto Encyclopedia of Genes and Genomes; TCGA, The Cancer Genome Atlas; LUAD, lung adenocarcinoma.

The “survival” package in R was used to perform univariate Cox regression analysis on 320 EMT-related DEGs. This identified 76 EMT-related genes with significant prognostic value. Lasso Cox analysis using the “glement” package in R identified 24 genes (Fig. 2A,B). Multivariate Cox analysis was used for further gene screening. A forest plot of hazard ratio (HR) (Fig. 2C) revealed that 5 genes had adverse prognostic significance (GALNT3, MUC5AC, OAS1, TMPRSS11E, VSIG4). The “survival” package in R was then used to screen for the strongest prognostic gene signatures. Kaplan–Meier analysis showed that GALNT3 and OAS1 were associated with significantly worse prognosis of LUAD patients in the GSE68465 cohort (Fig. 2D–H). Correlation analysis using the GEPIA database also revealed that GALNT3 and OAS1 expressions were closely correlated (r = 0.46, p-value = 1.3 $\times$ 10${}^{-44}$) (Fig. 2I).

Fig. 2.

Cox regression and Kaplan-Meier analysis were used to identify EMT-related genes with prognostic significance in LUAD. (A) The coefficient profiles of lasso regression analysis. (B) Minimal criteria were used to choose variables in lasso regression. (C) Forest plot showing the results of multivariate Cox regression analysis. (D–H) Kaplan–Meier survival analysis of GSE68465 patients according to the expression of GALNT3, OAS1, MUC5AC, TMPRSS11E, and VSIG4. (I) Correlation of GALNT3 and OAS1 expression in LUAD.

The genomic location of GALNT3 (via GeneCards) was q24.3 on chromosome 2 (Fig. 3A), while its subcellular location (via COMPARTMENTS) was mainly in the Golgi apparatus and extracellular space (Fig. 3C). The genomic location of OAS1 was q24.13 on chromosome 12 (Fig. 3B), while its subcellular location was centrally in the nucleus and cytosol (Fig. 3D). GALNT3 and OAS1 mRNA were overexpressed in LUAD tumor tissue from the GSE68465 cohort compared to normal tissue (Fig. 3E,H). GALNT3 and OAS1 protein expression were also higher in LUAD tumor tissue from the HPA database compared to normal lung tissue (Fig. 3F,I). The differential expression of GALNT3 and OAS1 between tumor and normal tissue in all TCGA tumors was investigated using the TIMER database (Fig. 3G,J). This analysis indicated that GALNT3 and OAS1 were overexpressed in most tumors. As shown in Supplementary Fig. 2, Gene Ontology (GO) annotations revealed that GALNT3 was mainly enriched in skin development (Supplementary Fig. 2A), while Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis showed that PI3K/Akt signaling was the most enriched pathway (Supplementary Fig. 2B). OAS1 was mainly enriched in response to the virus (Supplementary Fig. 2C), with Hepatitis C being the most enriched pathway according to KEGG pathway analysis (Supplementary Fig. 2D).

Fig. 3.

Genomic characteristics and differential expression of GALNT3 and OAS1 in LUAD. (A,B) Genomic locations of GALNT3 and OAS1 as viewed by GeneCards. (C,D) Subcellular locations of GALNT3 and OAS1 as viewed by COMPARTMENTS. (E) The mRNA expression level of GALNT3 in the GSE68465 cohort. (F) Immunohistochemical staining for GALNT3 in normal lung tissue and LUAD tissue from the HPA database. (G) Differential expression of GALNT3 between all tumors and adjacent normal tissue in TCGA via the TIMER database. (H) The mRNA expression level of OAS1 in the GSE68465 cohort. (I) Immunohistochemical staining for OAS1 in normal lung tissue and in LUAD tissue from the HPA database. (J) Differential expression of OAS1 between all tumors and adjacent normal tissue in TCGA via the TIMER database (*p 0.05, **p 0.01, ***p 0.001). HPA, Human Protein Atlas; TIMER, Tumor Immune Estimation Resource.

Univariate and multivariate Cox regression analyses were used to study the relationships between GALNT3 and OAS1 mRNA expression and the prognosis and clinical characteristics of LUAD. Univariate Cox analysis showed that GALNT3 (HR = 1.230; 95% confidence interval (CI) = 1.097–1.378; p 0.001) and OAS1 (HR = 1.201; 95% CI = 1.048–1.377; p = 0.008) mRNA expression levels were significantly associated with poor prognosis of LUAD in the GSE68465 cohort. The clinicopathological factors of age, gender, T stage and N stage were also associated with poor survival (Fig. 4A,B). Multivariate Cox regression analysis showed that high mRNA expression of GALNT3 was independently associated with poor prognosis (HR = 1.187; 95% CI = 1.055–1.334; p = 0.004), but not OAS1 (HR = 1.094; 95% CI = 0.953–1.2561; p = 0.203) (Fig. 4C,D). These results imply that GALNT3 expression is an independent biomarker of poor prognosis in LUAD. Although OAS1 expression showed no independent prognostic significance in multivariate analysis, this may be due to other confounding factors. Therefore, we comprehensively evaluated OAS1 as a marker of poor prognosis in LUAD by adding a multifaceted analysis method and database.

Fig. 4.

Prognostic significance of GALNT3 and OAS1 expression and associations with clinicopathological features in LUAD. (A) Univariate and (B) multivariate Cox analyses showing independent prognostic significance for GALNT3 expression in LUAD from the GSE68465 cohort. (C) Univariate and (D) multivariate Cox analyses showing the prognostic significance of OAS1 expression in LUAD from the GSE68465 cohort. (E–G) Correlation of GALNT3 expression in LUAD with individual cancer stages, TP53 mutation status and nodal metastasis status. (H–J) Correlation of OAS1 expression in LUAD with individual cancer stages, TP53 mutation status, and nodal metastasis statuses. (K,L) Kaplan-Meier analysis showing the prognostic significance of GALNT3 and OAS1 expression for the survival of LUAD patients (ns, not statistically significant, *p 0.05, **p 0.01, ***p 0.001).

We further explored the prognostic significance of GALNT3 and OAS1 expression levels in the UALCAN database, as well as their correlation with clinical features. The mRNA expression levels for both genes correlated closely with stage, TP53 mutation status and pathological N stage (Fig. 4E–J). Although the association with the pathological N3 stage was not statistically significant, this may have been due to the small number of samples. Patients with high expression of GALNT3 and OAS1 showed poor survival (Fig. 4K,L). Together, the above results indicate that high expression levels of GALNT3 and OAS1 may be associated with poor prognosis of LUAD patients.

DNA methylation status is strongly related to gene dysregulation in cancer cells. We explored the methylation status of promoter regions in GALNT3 and OAS1 in LUAD samples from the DNMIVD database. As shown in Fig. 5A,B, both the GALNT3 (p-value = 1.95 $\times$ 10${}^{-3}$) and OAS1 (p-value = 3.09 $\times$ 10${}^{-25}$) genes were significantly demethylated in LUAD tissue compared with normal lung tissue. Next, the samples were grouped into low, middle, and high methylation groups, with DNA methylation beta values of 0.3 and 0.7 used as cutoffs. The log-rank test was used then to calculate the p-value for prognostic significance of these groups. Kaplan–Meier survival analysis showed that the DNA methylation status of GALNT3 was closely related to prognosis (p-value = 4.02 $\times$ 10${}^{-3}$), but not that of OAS1 (p-value = 0.941) (Fig. 5C,D). Since the expression of oncogenes is regulated by miRNAs, potential upstream miRNAs that regulate GALNT3 and OAS1 mRNA expression were predicted using the TargetScan, miRDB, and miRTarBase databases. These online databases identified 27 overlapping miRNAs that target GALNT3. The StarBase database was used to analyze the coexpression of GALNT3 with the 27 overlapping miRNAs in LUAD. This revealed the mRNA expression level of GALNT3 was negatively associated with the expression of hsa-miR-122-5p (r = –0.092, p value = 3.75 $\times$ 10${}^{-2}$) (Fig. 5E). Only two overlapping miRNAs that target OAS1 were identified, and neither was co-expressed with OAS1 (Fig. 5F).

Fig. 5.

DNA methylation status and miRNA regulation of GALNT3 and OAS1 in LUAD. (A,B) Boxplots showing the methylation status of the GALNT3 and OAS1 promoter regions in LUAD and in normal lung tissue. (C,D) Kaplan–Meier survival curves for LUAD patients according to the DNA methylation status of GALNT3 and OAS1. (E) Potential miRNAs that target GALNT3 mRNA were identified from the intersection of TargetScan, miRDB, and miRarBase databases. The correlation between GALNT3 expression and hsa-miR-122-5p expression in LUAD was analyzed in the StarBase database. (F) Potential miRNAs that target OAS1 mRNA were identified from the intersection of TargetScan, miRDB, and miRarBase databases.

Previous studies have shown that EMT-related genes play a critical role in tumor immunity. Here, we analyzed the correlation of GALNT3 and OAS1 expression with immune cell infiltration in LUAD. High GALNT3 expression was associated with high levels of many immune cell populations, including B cells (partial.cor = –0.129, p = 4.40 $\times$ 10${}^{-3}$), CD8+ T cells (partial.cor = 0.148, p = 1.02 $\times$ 10${}^{-3}$), macrophages (partial.cor = 0.119, p = 8.47 $\times$ 10${}^{-3}$), neutrophils (partial.cor = 0.162, p = 3.59 $\times$ 10${}^{-4}$), and dendritic cells (partial.cor = 0.165, p = 2.59 $\times$ 10${}^{-4}$). Tumor purity (partial.cor = –0.055, p = 2.20 $\times$ 10${}^{-1}$) and CD4+ cells (partial.cor = 0.047, p = 3.05 $\times$ 10${}^{-1}$) were not significantly associated with GALNT3 expression (Fig. 6A). High OAS1 expression was correlated with tumor purity (partial.cor = –0.154, p = 5.92 $\times$ 10${}^{-4}$), neutrophils (partial.cor = 0.235, p = 1.85 $\times$ 10${}^{-7}$), and dendritic cells (partial.cor = 0.159, p = 4.19 $\times$ 10${}^{-4}$), but not with B cells (partial.cor = –0.08, p = 7.90 $\times$ 10${}^{-2}$), CD4+ cells (partial.cor = –0.08, p = 7.90 $\times$ 10${}^{-2}$), CD8+ T cells (partial.cor = 0.067, p = 1.38 $\times$ 10${}^{-1}$) or macrophages (partial.cor = 0.062, p = 1.70 $\times$ 10${}^{-1}$) (Fig. 6B).

Fig. 6.

Correlation analysis of GALNT3 and OAS1 expression in LUAD with infiltrating immune cells. (A,B) Correlation of GALNT3 and OAS1 expression with that of various immune cell markers. (C,D) Correlations between GALNT3 and OAS1 expression and immune subtypes in LUAD. (E–G) Kaplan–Meier survival analysis for LUAD patients with high or low GALNT3 expression and according to the level of infiltration with different immune cell populations. (H–K) Kaplan–Meier survival analysis for LUAD patients with high or low OAS1 expression and according to the level of infiltration with different immune cell populations. HR, hazard ratio.

We further analyzed the correlations between GALNT3 and OAS1 expression in LUAD with immune cell infiltration using the TISIDB database. Samples were divided into TGF-b dominant, immunologically quiet, inflammatory, IFN-gamma dominant, and wound healing groups. GALNT3 expression was highest in the lymphocyte-depleted group and lowest in the inflammatory cell group (Fig. 6C), while OAS1 expression was highest in the IFN-gamma-dominant cell group and weakest in the inflammatory cell group (Fig. 6D). The relationships between GALNT3 and OAS1 expression in LUAD with TILs, MHC molecules, immune inhibitors, and immune stimulators are shown in Supplementary Fig. 3.

To confirm the associations between GALNT3 and OAS1 expression in LUAD with immune infiltrating cells, further analysis was conducted using the TIMER and GEPIA databases (Supplementary Tables 1,2). The results confirmed that GALNT3 and OAS1 expression levels were closely associated with those of Th2 cell immune marker genes, including STAT6, CCR8, and. Numerous factors could modulate the prognostic significance of GALNT3 and OAS1 expression in LUAD. We hypothesized that immune cell infiltration could affect their prognostic impact in LUAD. Therefore, we further analyzed the prognostic significance of GALNT3 and OAS1 in the Kaplan–Meier plotter database according to the presence of immune cell infiltration. This showed that high GALNT3 expression was most strongly associated with poor prognosis in LUAD cases that had high levels of macrophage, regulatory T cell and type 2 T helper cell infiltration (Fig. 6E–G). Similarly, high OAS1 expression was associated with poor prognosis in cases that showed strong infiltration with CD8+ T cells, macrophages, mesenchymal stem cells, and type 2 T helper cells (Fig. 6H–K). These results suggest that LUAD with high GALNT3 and OAS1 expression may potentially respond to immunotherapy through the regulation of immune cell infiltration.

We next examined the frequencies and types of genetic alterations in GALNT3 and OAS1 in TCGA-LUAD using the cBioPortal database. The genetic alterations found in GALNT3 were amplification and mutations. Somatic mutations in GALNT3 were found in 2.6% of cases, with the mutation sites located within amino acids 0 and 633 and in the Glycos-transf-2 and Ricin-B-lectin domains. Four missense mutations were reported, with one being the hotspot mutation E545K (Fig. 7A,C). Gene alterations in OAS1 consisted of amplification, mutations and deletions. Somatic mutations were found in 3% of cases, with mutation sites in the 0 to 400 amino acid NTP-transf-2 and OAS1-C domains. Three missense mutations were reported, with one being the hotspot mutation T251K (Fig. 7B,C). Genetic alterations in GALNT3 and OAS1 correlated with the frequency and pattern of genetic alterations in several other genes (Fig. 7D,E). The top 10 co-occurring genes with GALNT3 alteration were TTN (47% co-alteration frequency), MYO3B (6%), XIRP2 (21%), ADAMTS12 (23%), ZFHX4 (30%), LRP1B (32%), USH2A (35%), MUC16 (40%), TP53 (46%), and ERCC5 (4%) (Fig. 7F). The top 10 co-occurring genes with OAS1 alteration were CIT (6%), TMEM116 (2.6%), CCDC63 (4%), RPH3A (5%), TBX3 (5%), KSR2 (5%), MED13 L (7%), TBX5 (7%), RBM19 (9%), and LRRIQ1 (10%) (Fig. 7G).

Fig. 7.

DNA mutation, confirmation of mRNA expression, and diagnostic value of GALNT3 and OAS1 in LUAD. (A,B) Location of mutations in GALNT3 and OAS1 in LUAD. (C) GALNT3 and OAS1 genetic alteration types in LUAD. (D,E) Bar plots showing the top 10 gene alterations in LUAD with the highest frequency of co-occurrence with GALNT3 and OAS1 alterations. (F,G) Waterfall plots showing the co-occurrence pattern of GALNT3 and OAS1 gene mutations with the most frequent genetic changes in LUAD. (H,I) mRNA expression levels of GALNT3 and OAS1 in LUAD from the TCGA database. (J,K) mRNA expression levels of GALNT3 and OAS1 in LUAD cells (H1975) were higher than those in normal bronchial epithelial cells (BEAS-2B). (L) Receiver operating characteristic (ROC) curve analysis showing the diagnostic value of GALNT3 and OAS1 expression in LUAD, both individually and in combination (**p 0.01, ***p 0.001). BE, bronchial epithelial.

GALNT3 and OAS1 mRNA expression was confirmed in LUAD and normal tissue from the TCGA database. They were again found to be significantly elevated in primary LUAD tissue compared with corresponding adjacent normal tissue (Fig. 7H,I). In addition, RT-PCR was used to examine GALNT3 and OAS1 mRNA expression levels in the normal cell line BEAS-2B and in the LUAD cell line H1975. This confirmed the overexpression of both genes in LUAD cells (Fig. 7J,K). The overexpression of GALNT3 and OAS1 in LUAD was further validated in the GEO and TCGA databases (Figs. 3E,H,7H,I).

ROC curve analysis showed AUC values of 0.9366 and 0.9759 for GALNT3 and OAS1 expression, respectively (Fig. 7L), thus highlighting their diagnostic value for LUAD. Logistic regression analysis for the combination of GALNT3 and OAS1 expression revealed an AUC of 0.9878 (Fig. 7L), indicating a very high diagnostic value for this biomarker combination in LUAD.