Abstract
Objective. The purpose of this study is to screen for microRNAs (miRNAs) associated with the prognosis of lung adenocarcinoma (LUAD) and to explore its prognosis and effects on the tumor microenvironment in patients with LUAD. Methods. Gene expression data, miRNA expression data, and clinical data for two different databases, TCGA-LUAD and CPTAC-3 LUAD, were downloaded from the GDC database. The miRNA prognosis of LUAD was filtered by the Cox proportional hazard model and the Least Absolute Shrinkage and Selection Operator (LASSO) regression model. The performance of the model was validated by time-dependent receiver operating characteristics (ROC) curves. Possible biological processes associated with the miRNAs target gene were analyzed through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG). Finally, the prognostic model was scored by risk, divided into high- and low-risk groups by median, and the differences in the immersion level of 21 immune cells in the high- and low-risk groups were assessed. To gain a deeper understanding of the underlying mechanism behind the model, the two most important miRNAs in the model, miR-195-3p and miR-5571-5p, were selected for HPA database validation and ceRNA network construction. Results. Of the 209 variance expressions identified in the screening analysis, 145 were upregulated and 64 were downregulated by miRNAs. The prognostic models of six miRNA genes were obtained: miR-195-3p, miR-5571-5p, miR-584-3p, miR-494-3p, miR-4664-3p, and miR-1293. These six genes were significantly associated with survival rates in LUAD patients. In particular, miR-1293, miR-195-3p, and miR-5571-5p are highly correlated with OS. The higher expression of miR-195-3p and miR-5571-5p, the better survival of LUAD OS is, and these two miRNA expressions contribute the most to the model. Finally, after sorting the risk scores calculated from low to high using the prognostic model, the patients with higher scores had shorter survival time and higher frequency of death, and there were significant differences in the immersion levels of 21 immune cells in the high- and low-risk groups. ceRNA network analysis found that TM9SF3 was regulated by miR-195-3p and was highly expressed in the tissues of LUAD patients, and the prognosis of the patients was poor. Conclusions. miR-195-3p, miR-5571-5p, miR-584-3p, miR-494-3p, miR-4664-3p, and miR-1293 may be used as new biomarkers for prognosis prediction of LUAD. Our results also identified a lncRNA MEG3/miR-195-3p/RAB1A/TM9SF3 regulatory axis, which may also play an important role in the progression of LUAD. Further study needs to be conducted to verify this result.
1. Introduction
Lung cancer is a major cause of cancer-related mortalities all over the world as well as the most frequent form of cancer in men, and the second most frequent in women [1]. Almost 4/5 of all types of lung cancer are non-small cell lung cancer (NSCLC), and lung adenocarcinoma (LUAD) being the most common NSCLC histological subtype [2]. Despite improvements in molecular diagnosis and therapy, the prognosis of LUAD remains poor and the risk of metastasis and recurrence remains high [3]. Most LUAD patients are identified at a late stage due to a lack of adequate diagnostic techniques, and the 5-year survival rate of patients is identified as poor (approximately 17.4%) [4]. Patients with lung cancer, who receive early surgical resection, have a 5-year survival rate of up to 70% [5]. Hence, it is important to identify the biomarkers and potential therapeutics for the diagnosis and prognosis of LUAD.
MicroRNAs (miRNAs) are a kind of noncoding RNA that has about 19–25 nucleotides and regulates the expression of genes after transcription [6]. It is found to be improperly expressed in many malignancies (including lung cancer) and can be utilized as oncogenes or tumor suppressor genes [7]. There is substantial evidence that miRNAs regulate carcinogenesis processes including cell maturity, growth, invasion, autophagy, motility, invasion, and apoptosis [8, 9]. Therefore, miRNA has great potential as a promising marker for diagnosis, prognosis, and personalized targeted therapy. However, in LUAD, the function of miRNA in the prognosis of cancer patients and tumor microenvironment has not been well elucidated. Therefore, disease management and treatment need to establish a lung cancer risk prediction model by screening biomarkers for lung cancer progression.
Gene expression data, miRNA expression data, and medical information for two different projects, TCGA-LUAD, and CPTAC-3 LUAD, were downloaded from the GDC database. Firstly, the intersection of differential expression of miRNAs in tumor tissues and healthy tissues was screened out in both the data sets, and then univariate Cox regression detected the prognosis-related miRNAs. The prognostic model consisting of six genes (miR-195-3p, miR-5571-5p, miR-584-3p, miR-494-3p, miR-4664-3p, and miR-1293) was obtained and constructed using LASSO. The target genes of the miRNAs were then identified and analyzed using the Gene Ontology (GO) function and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. The relationship of risk score with immune cell infiltration was examined to understand the model’s mechanism. The detailed workflow diagram is shown in Figure 1.
2. Materials and Methods
2.1. Acquisition of Data
The GDC (Global Data Cente) database (https://portal.gdc.cancer.gov/) is the largest collection of cancer gene information that stores data like gene expression, DNA methylation, miRNA expression data, SNP, copy number variation, and others. There are currently two publicly available LUAD tumor data in the GDC data. Among them, TCGA-LUAD (normal group: 45, tumor group: 513) and CPTAC-3 (normal group: 197, tumor group: 219) included miRNA expression data from LUAD patients with sufficient sample size. The level 3 miRNA transcript expression data of these two projects was downloaded and compiled into an arm-level expression matrix. We used TCGA biolinks to download clinical data and post-expression data standardized by FPKM (FragmentsPer Kilobase per Million) at level 3. The FPKM data eliminated the effects of library construction and sequencing depth. In the following analysis, the expression levels of the samples may be directly compared.
2.2. Construction of Model and Prognosis
R software package DESeq2 was utilized to explore the differential expression of TCGA and CPTAC-3, and obtained 209 genes (in tumor tissues upregulated: 145, downregulated: 64) with significant differences ( adj< 0.05 and |log2FoldChange| > 1) in expression in normal tissues and tumor tissues by taking the intersection. A univariate Cox regression analysis of significantly differentially expressed miRNA genes was done, and 13 miRNA genes were shown to have a significant connection to overall survival (OS) in LUAD patients (-value ≤ 0.05). At the same time, to obtain the genes most likely to be related to survival, we used the KM test to further screen. The samples were grouped as high-expression and low-expression categories, with the median serving as the cutoff. By analyzing the impact of each miRNA expression on OS, 8 miRNAs with a high correlation to OS prognosis (-value < 0.05) were obtained. Furthermore, a prognostic-related model was constructed through lasso regression, which was composed of 6 miRNA genes. A risk score formula was created for each patient after integrating the expression value of each particular gene. In the lasso regression analysis, the regression coefficients estimated by this risk score formula were weighted. The median risk score value served as the demarcation point in the risk score calculation to split patients into low-risk and high-risk categories. The differences in the survival rate of the two groups were calculated using Kaplan–Meier and analyzed by log-rank statistical techniques. To investigate the accuracy of the model predictions, the survival ROC in the R package was utilized, and the C-index index was obtained using the SURVCOMP method in the R package. The prognostic model was further validated using the external dataset GSE175462. Given that the dataset did not have complete survival data, we used the support vector machine (SVM) algorithm to construct a classification model consisting of six miRNAs, and performed ROC analysis and confusion matrix visualization at the same time.
2.3. ceRNA Network Construction
In this study, lncRNAs that appeared at least three times were obtained from two databases, and the subsequent ceRNA network construction was carried out. The screened lncRNAs were paired with lncRNA-miRNA in the miRcode database (http://www.mircode.org). Then, mRNA prediction was performed simultaneously in the three miRNA gene prediction databases of miRDB, miRTarBase, and TargetScan. Finally, the matched lncRNA-miRNA pairing and miRNA-mRNA pairing were imported into Cytoscape (Version 3.7.2) software to construct a lncRNA-miRNA-mRNA regulatory network based on the ceRNA mechanism.
2.4. Prediction of RAB1A and TM9SF3 Immunohistochemistry Using the HPA Database
The Human Protein Atlas database (https://www.proteinatlas.org/), the HPA database, is a free public database of more than 26 000 antibodies targeting over 17 000 human genes. Normal tissue and LUAD tissue immunohistochemical samples were obtained by searching for the genes RAB1A and TM9SF3.
2.5. miRNA Target Gene Prediction and GO, KEGG Enrichment
The miRDB was used to predict miRNA target genes. Visualization was provided using the software Cytoscape. To thoroughly investigate the functional significance of these mutant genes, the R package “ClusterProfiler” was used to annotate mutant genes. To analyze associated functional categories, the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were utilized. The GO and KEGG enrichment pathways are considered significant, with -value and q-value <0.05.
2.6. Immune Cell Infiltration Analysis
The CIBERSORT technique is used to determine the types of immune cells present in the tumor microenvironment. This approach uses the support vector regression concept to undergone convolution analysis of the expression matrix of immune cell subtypes. This includes 547 biomarkers and detects 22 human immune cell phenotypes such as T, B, plasma, and myeloid cell subsets. The expression levels of LUAD patients were checked by the CIBERSORT method, to predict the relative infiltration ratio of 22 immune cells. Following the risk score generated by the prognostic model, the subjects were classified into high-risk and low-risk groups based on the median. The difference in the level of infiltration of 21 immune cells was assessed in the high-risk category and the low-risk category (as the infiltration level predicted by T cells CD4 naive in all samples was 0, it was not included in the analysis). The -value was calculated using the Mann–Whitney test method.
2.7. Statistical Analysis
The Kaplan–Meier analysis produced the survival curves, and the log-rank was utilized to compare and detect the -value. For univariate analysis, the Cox proportional hazard model was employed and for correlation analysis, Pearson’s test was utilized. The R programming language was used for all statistical analyses (version 3.6). These statistical tests were two-sided, with a statistical significance of .
3. Results
3.1. Analysis of Differentially Expressed miRNA and Screening of OS-Related Genes
There are two publicly available LUAD tumor data in the GDC database. We downloaded the original miRNA transcript expression data from the TCGA-LUAD and CPTAC-3 datasets. The difference in expression of miRNAs that were upregulated and downregulated in the two datasets was found using differential analysis (Figures 2(a)–2(d)), respectively. Following the intersection, 145 genes were significantly upregulated in tumor tissues (Figure 2(e)) while 64 genes were significantly downregulated in tumor tissues (Figure 2(f)). Next, we performed a univariate Cox regression analysis on these 209 differentially expressed miRNA, and screened out 13 genes related to OS (Figures 3(a)–3(d)). The Kaplan–Meier test was used to further filter out 8 miRNAs significantly related to OS as the included genes for the subsequent prognostic model establishment. These 8 miRNAs were miR-494-3p, miR-1293, miR-4664-3p, miR-5571-5p, miR-5571-3p, miR-7974, miR-195-3p, and miR-584-3p, respectively.
(a)
(b)
(c)
(d)
(e)
(f)
(a)
(b)
(c)
(d)
3.2. Model Construction and Evaluation
The clinical data of TCGA-LUAD patients were gathered. A prognostic model for 8 miRNAs significantly linked to OS () was constructed using univariate Cox regression analysis and lasso regression methods combined. TCGA patients were assigned to the training and validation groups in a 1:1 ratio at random. We utilized lasso regression analysis to get the optimal risk score value for further investigation. Finally, a prognostic model (Table 1) consisting of six genes was obtained: miR-195-3p, miR-5571-5p, miR-584-3p, miR-494-3p, miR-4664-3p, and miR-129. All six genes can be exploited as independent prognostic factors for OS. In particular, miR-1293, mir-195-3p, and mir-5571-5p were highly correlated with OS. The greater the expression of miR-195-3p and miR-5571-5p, the better the LUAD OS (Figures 4(a)–4(f)). The expression of miR-195-3p and miR-5571-5p made the most contribution to the model (Figure 5).
(a)
(b)
(c)
(d)
(e)
(f)
Based on the median risk score, the two groups of patients were high-risk and low-risk. To compare and determine the -value, the Kaplan–Meier curve and log-rank were employed. In all samples, training sets, and test sets, the OS of the high-risk group was significantly lower than the low-risk group (Figures 6(a), 6(c), 6(e)). In addition, the ROC curve findings revealed that the C-index indices of all sample sets, training sets, and test sets were 0.62, 0.66, and 0.58, respectively (Figures 6(b), 6(d), 6(f)), indicating that the model had a better verification performance. The results of the support vector machine model are shown in Figure S1, the AUC of the model composed of six miRNAs reached 0.98, and the prediction accuracy rate was 95.2%, indicating that the construction of the prognosis model has good scalability. EGFR-mutated lung cancer patients receive targeted therapy with significant efficacy and prolonged survival. We wanted to investigate whether riskscore was superior to EGFR expression level as a better survival indicator for lung cancer. GEPIA (http://gepia2.cancer-pku.cn) database was used to analyze whether EGFR expression is related to survival (Figure S1C). The relationship between EGFR expression level and patient survival is not significant, indicating that riskscore is better than EGFR. To determine the prognostic value of the model, this study used univariate and multivariate COX analysis to detect the impact of gender, age, smoking or not, AJCC staging, TNM staging, and other clinical risk factors on the prognosis. According to the findings, AJCC pathological stage was considered an independent marker for prognosis of OS based on the six miRNA prognostic models (Figures 7(a) and 7(b)). After sorting the risk scores using the prognostic model from low to high, the survival time of OS and progression-free interval (PFI) changed accordingly with the increase of the score (Figure 7(c)), the higher the score, the less time the patient survived and the higher the death rate.
(a)
(b)
(c)
(d)
(e)
(f)
(a)
(b)
(c)
3.3. Discussion on Specific Signal Mechanism Related to the Prognosis Model
Tumor-related fibroblasts, immune cells, extracellular matrix, and a range of growth factors, inflammatory agents, unique physical and chemical properties, and cancer cells themselves make up the tumor microenvironment. The tumor microenvironment has a major impact on tumor diagnosis, prognosis, and clinical therapy sensitivity. CIBERSORT is the most widely used technique for determining the degree of immune cell infiltration in tumor tissues. CIBERSORT calculated the infiltration level of 22 immune cells in the LUAD sample. The probable mechanism of risk score regulating tumor immune infiltration was discovered by examining the connection between risk score and tumor immune infiltration. Following risk scoring with a prognostic model, the high-risk and low-risk groups were defined by the median. After excluding T cells CD4 naive whose predicted infiltration level was 0 in all samples, the Mann–Whitney test was used for the two groups to evaluate the difference in the infiltration level of the remaining 21 immune cells and calculate the -value. These findings revealed a significant difference among T cells CD4 memory resting, macrophages M0, mast cells resting, T cells regulatory (Tregs), mast cells activated, plasma cells, and eosinophils in the high-risk group and the low-risk group (; Figure 8).
3.4. Construction of ceRNA Network
A total of six miRNAs were constructed in this model, of which miR-195-3p and miR-5571-5p had the greatest contribution to the model. In this study, these two miRNAs were selected for subsequent ceRNA network construction (Figures 9(a) and 9(b)). The analysis results showed that no associated lncRNA molecules were found for miR-5571-5p, and two lncRNAs, MEG3, and AC016717.2, acted as upstream regulatory elements of miR-195-3p. Further, the miR-195-3p target genes TM9SF3, RAB1A, USP46, and SUB1 were predicted through the database, and the ceRNA network was shown in Figure 9(c). Expression profiling analysis found that MEG3 was downregulated in LUAD patients compared with normal tissues, and survival analysis showed that LUAD patients with high MEG3 expression had a poorer survival trend (Figure 9(d)). We predicted survival for all four target genes of miR-195-3p, demonstrating that patients with high expression of RAB1A and TM9SF3 have a poor prognosis (Figure 9(e)). Thus, the lncRNA MEG3/miR-195-3p/RAB1A/TM9SF3 regulatory axis may play a vital role in the progression of LUAD.
(a)
(b)
(c)
(d)
(e)
3.5. Validation of TM9SF3 Expression Level in HPA Database
Log-rank survival analysis was performed on the four target genes RAB1A, TM9SF3, USP46, and SUB1, and the patients with high expression of RAB1A and TM9SF3 genes had poor prognosis and low survival rate (Figure 9(e)). Further verification of RAB1A and TM9SF3, whether they are highly expressed in lung cancer patients. As shown in Figure 10, the HPA database found that TM9SF3 protein was highly expressed in LUAD tissues. The immunohistochemical staining in the HPA database showed that compared with normal tissues, the immunohistochemical staining of cancer tissues was deeper, suggesting that TM9SF3 protein was highly expressed in LUAD tissues (Figure 10(a)), however, there was no significant difference in RAB1A expression (Figure 10(b)).
(a)
(b)
3.6. miRNA Target Gene Prediction and GO, KEGG Enrichment
The miRDB database was utilized to determine the potential target genes of these six miRNAs. They were as follows: miR-195-3p:273, miR-5571-5p:162, miR-584-3p:261, miR-494-3p:188, miR-4664-3p:4, and miR-1293 : 371 (Figure 11(a)). Using Cytoscape, we visualized the possible link between miRNA and the target gene. To assess the biological activities of these target genes, we used GO enrichment and KEGG biological pathway enrichment to investigate the potential involvement of these essential miRNAs in tumor tissues (Figures 11(b)–11(e)). GO analysis revealed that the biological process (BP) of target genes was enriched in the regulation of ion transmembrane transport and the regulation of metal ion migration. Molecular function (MF) was enhanced in the DNA-binding transcription activator activity and ion channel binding, while cell components (CC) were mainly enriched in synaptic membranes, transcription factor complexes, etc. KEGG biological pathway analysis showed that the MAPK signaling pathway and chemical carcinogenesis-receptor activation pathways were enriched.
(a)
(b)
(c)
(d)
(e)
4. Discussion
miRNA is a highly conserved noncoding RNA with approximately 19–25 nucleotides [10]. They induce posttranscriptional silencing by specifically binding with complementary sites of the target mRNA’s 3′untranslated region (UTR) [11]. miRNA is related to many biological functions, such as proliferation, development, differentiation, apoptosis, and metabolism [12]. Many studies showed that abnormal miRNA expression is significantly linked to tumorigenesis, and it is now a hot research area [13, 14]. The most prevalent subtype of lung cancer is adenocarcinoma, which has a high worldwide morbidity and mortality rate [15]. It is critical to identify precise biomarkers for it. Increasing data suggest that miRNAs have a crucial function in the prevention of LUAD [16, 17]. miRNA has been demonstrated to be a complex combination of gene expression and pathway regulatory systems, as well as prognostic markers and therapeutic targets of different cancers, including lung cancer [18]. Many miRNAs have important roles in the occurrence, progress, and metastasis of lung cancer through regulating a variety of processes, including cancer initiation and progression. Some miRNAs associated with prognostic value have been discovered thus far, such as miR-221 [19], miR-372 [20], miR-429 [21], miR-486 [22], and miR-137 [23]. However, many miRNAs are yet to be discovered in LUAD, and their roles are yet to be clarified.
Here, a genome-wide analysis of miRNAs was conducted for many LUAD patients from TCGA and CPTAC-3 and found that miR-195-3p, miR-5571-5p, miR-584-3p, miR-494-3p, miR-4664-3p, and miR-1293 were differentially expressed in cancer cells than normal cells and were significantly related to OS. Specifically, miR-1293, miR-195-3p, and miR-5571-5p have significantly correlated with soothe higher the expression of miR-195-3p and miR-5571-5p, the longer LUAD survives. Moreover, the ROC analysis results of this study indicated that the AUCs of all six miRNAs in LUAD were more than 0.6, indicating that these six miRNAs had a high diagnostic value for LUAD and may be used as LUAD biomarkers. In addition, these six miRNAs have been found to be abnormally expressed in a variety of tumors. We summarize them in Table 2.
According to a few studies, miR-195-3p is linked to the development of renal cell carcinoma, cervical cancer, and oral squamous cell carcinoma. The miR-195-3p expression is upregulated in renal cell carcinoma and decreased in other cancers [24–26]. Thus, miR-195-3p can be a possible biomarker for a variety of malignancies that can be utilized for detection, targeted treatment, or prognosis prediction. The miR-5571-5p possesses the potential to be a diagnostic biomarker for dilated cardiomyopathy and is related to NYHA classification, but there is no research on the occurrence and development of tumors [27]. Our study found for the first time that miR-5571-5p can be related to the prognosis of LUAD. As a tumor suppressor gene, miR-584-3p is related to the initiation of colon cancer, glioma, gastric cancer, renal cell carcinoma, and malignant melanoma, and is an independent prognostic factor with a good prognosis [28–32]. The miR-494-3p is upregulated in endometrial cancer, glioma, retinoblastoma, and hepatocellular carcinoma, which promotes cancer progression by regulating the PTEN/PI3K/AKT pathway [33–36]. While miR-494-3p is downregulated in synovial sarcoma, prostate cancer, osteosarcoma, and oral squamous cell carcinoma, and acts as a tumor suppressor miRNA [37–40]. Furthermore, miR-494-3p was associated with a new tumor driver of lung cancer. In NSCLC cell lines, miR-494-3p is significantly upregulated. It can improve NSCLC cell proliferation, migration, and invasion by inhibiting WT1-AS overexpression [41]. The miR-4664-3p was found to be associated with postoperative recurrence in patients with small cell carcinoma of the esophagus [42]. According to reports, miR-1293 is upregulated in pancreatic cancer and papillary renal cell carcinoma and acts as an oncogene in tumor development [42]. Moreover, miR-1293 is strongly associated with LUAD patient mortality, can be a potential biomarker for detecting LUAD prognosis, and is significantly enriched in systemic lupus erythematosus pathways [45]. This study supported the previous findings. Our results showed that miR-195-3p and miR-5571-5p were downregulated in LUAD, miR-584-3p, miR-494-3p, miR-4664-3p, and miR-1293 were upregulated in LUAD, which are potential biomarkers for LUAD.
We identified target genes and analyzed associated pathways and GO annotations to obtain insight into the molecular functions of these six miRNAs. The development and progression of lung cancer are heavily reliant on abnormal signaling pathways. It was found that these six miRNAs can regulate several key signal pathways, such as the biological process (BP) of target genes regulates ion transmembrane transport, the regulation of metal ion migration is enriched, the molecular function (MF) to DNA-transcription activator activity is enriched, and ion channel binding is enriched, while the cell component (CC) is mainly enriched in synaptic membranes and transcription factor complexes. KEGG biological pathway analysis shows that the mitogen-activated protein kinase (MAPK) signaling pathway and chemical carcinogenesis-receptor activation pathways are enriched. Furthermore, in this study, patients were grouped as high-risk and low-risk groups based on their median risk score, and the OS of all samples, training sets, and test sets in high-risk groups was considerably lower than that of low-risk groups. Ultimately, after sorting the prognostic model risk scores from low to high, we observed that the higher score had a shorter survival time and higher mortality rate. Moreover, there were significant differences in the infiltration level of 21 immune cells among the high-risk and low-risk groups.
The risk model composed of six miRNAs constructed by lasso can well assess the risk status of LUAD patients. In order to explore the biological issues behind the model, we selected miR-195-3p to construct a ceRNA network, obtained a meaningful regulatory pathway lncRNA MEG3/miR-195-3p/RAB1A/TM9SF3, and finally obtained two genes RAB1A and TM9SF3. Next, expression profiling analysis, survival analysis, and HPA immunohistochemistry found that TM9SF3 was highly expressed in the tissues of LUAD patients, and the patients had a poor prognosis. However, the RAB1A survival analysis results were not significant. Previously, we found that miR-195-3p was less expressed in LUAD compared with normal tissues. Therefore, we can speculate that miRNA-195-3p is regulated by upstream lncRNA, negatively regulates the expression of TM9SF3, and ultimately affects the occurrence and development of cancer cells in LUAD patients.
In summary, these six miRNAs are significantly associated with the survival rate of LUAD patients and may be used as potential markers for predicting the prognosis of LUAD patients. This research also has some limitations. Since the CPTAC-3 sample has fewer deaths, the survival information of this dataset is not used. Another shortcoming of this study is that only internal verification was done, without external verification. Before clinical application, more validated studies in prospective datasets are required to confirm the predictive ability of the diagnosis.
5. Conclusion
In this study, a prognostic model with six miRNAs characteristics was constructed based on the original miRNA transcript expression data of two public datasets TCGA-LUAD and CPTAC-3 in the GDC database, and found that miR-494-3p, miR-195-3p, miR-584-3p, miR-5571-5p, miR-1293, and miR-4664-3p, may be used as a new biomarker for prognostic prediction of LUAD. miR-195-3p in the prognostic model negatively regulates the expression of TM9SF3 through the lncRNA MEG3/miR-195-3p/RAB1A/TM9SF3 regulatory pathway. This study’s findings provide new insights into the precise treatment and prognosis prediction of LUAD.
Abbreviations
| NSCLC: | Non-small cell carcinoma |
| LUAD: | Lung adenocarcinoma |
| miRNA: | microRNA |
| GDC: | Global Data Center |
| TCGA: | The Cancer Genome Atlas |
| CPTAC-3: | Clinical Proteomic Tumor Analysis Consortium-3 |
| FPKM: | Fragments Per Kilobase per Million |
| OS: | Overall survival |
| PFI: | Progression-Free Interval |
| HR: | Hazard ratio |
| KEGG: | Kyoto Encyclopedia of Genes and Genomes |
| GO: | Gene Ontology |
| AUC: | Area under the curve |
| ROC: | Receiver operating characteristic |
| BP: | biological process |
| MF: | molecular function |
| CC: | cell component |
Data Availability
The datasets analyzed during the current study are available from the corresponding author upon reasonable request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
The authors appreciate greatly for the analytic data provided by the TCGA and GEO databases.
Supplementary Materials
Figure S1: Support vector machine model prediction of six miRNAs in GSE175462 dataset and EGFR survival analysis. A: SVM model ROC curve; B: Confusion matrix visualization; C: EGFR log-rank analysis results. (Supplementary Materials)