Role of FBXL19-AS1 in hepatocellular carcinoma by lncRNA–miRNA–mRNA network analysis and its diagnostic and prognostic value

Background: Hepatocellular carcinoma (HCC) is one of the most common neoplastic diseases worldwide. Available biomarkers are not sensitive enough for the diagnosis of HCC, seeking new biomarkers of HCC is urgent and challenging. The purpose of this study was to investigate the role of F-box and leucine-rich repeat protein 19-antisense RNA 1 (FBXL19-AS1) through competing endogenous RNA (ceRNA) network and its diagnostic and prognostic value in HCC. Methods: A comprehensive strategy of genomic data mining, bioinformatics and experimental validation was used to evaluate the clinical value of FBXL19-AS1 in the diagnosis and prognosis of HCC and to identify the pathways that FBXL19-AS1 may be involved in. Results: FBXL19-AS1 was up-regulated in HCC, and its high expression was associated with TNM stage and poor prognosis of HCC patients. The combined use of plasma FBXL19-AS1 and alpha-fetoprotein (AFP) could prominently improve the diagnostic validity for HCC. FBXL19-AS1 might participate in regulating HCC related pathways, including hepatitis C, hepatitis B, microRNAs in cancer, cell cycle, viral carcinogenesis, and proteoglycans in cancer through ceRNA network. Conclusions: Our ndings indicated that FBXL19-AS1 not only serves as a potential biomarker for HCC diagnosis and prognosis, but it may be functionally carcinogenic. The experimental results indicated FBXL19-AS1 was elevated in HCC and its expression was correlated with TNM stage, AFP and GGT. that elevated GGT is associated with the occurrence of acute and chronic hepatitis and alcoholic liver disease, it can be inferred that FBXL19-AS1 involved in the development of related In combination with the GEPIA2 survival analysis results and our follow-up study on 57 patients, we found high expression of FBXL19-AS1 was associated with poor prognosis in HCC. HCC: ceRNA: RNA; AFP: alpha-fetoprotein; ncRNAs: non-coding RNAs; MiRNAs: microRNAs; LncRNAs: long non-coding RNAs; MREs: miRNA response elements; TCGA LIHC: The Cancer Genome Atlas liver hepatocellular carcinoma; GEO: Gene Expression Omnibus; GSEA: Gene Set Enrichment Analysis; AJCC: American Joint Committee on Cancer; CDNA: complementary DNA; qPCR: quantitative real-time PCR; GAPDH: glyceraldehyde 3-phosphate dehydrogenase; ALT: alanine aminotransferase; AST: aspartate aminotransferase; carcinoembryonic antigen; phosphatase; Interactive Analysis2;


Background
Liver cancer is a common malignant cancer globally. According to the latest global cancer statistics, liver cancer ranks sixth and fourth in morbidity and mortality among all types of cancers, respectively [1].
Hepatocellular carcinoma (HCC) accounts for 70%-85% of primary liver cancers, is the most common form of liver cancer. Recent years have witnessed a great gain in treatment methods of HCC, such as surgical resection, liver transplantation, adjuvant therapy, interventional therapy and so on. [2][3]. The 5year survival rate of HCC patients with early diagnosis and appropriate treatment or intervention is more than 50% [4], but for those who are diagnosed and treated after the relevant symptoms appear, the 5-year survival rate is only 14.1% [5]. Currently, although alpha-fetoprotein (AFP) has been reported to be a valid marker for the clinical diagnosis and prognosis of HCC, its value remains unsatis ed in early diagnosis.
To improve the outcome and prognosis of HCC patients, it's essential to nd more effective biomarkers to improve the early diagnosis of HCC [6].
Studies have found that non-coding RNAs (ncRNAs) such as microRNAs (miRNAs) and long non-coding RNAs (lncRNAs) are involved in regulating proliferation, invasion and metastasis of HCC cells, providing a novel perspective for the diagnosis and prognosis of HCC [7]. The competitive endogenous RNA (ceRNA) mechanism is one of the classical mechanisms for the regulation of ncRNAs. Salmena et al rst proposed the ceRNA hypothesis that ceRNAs can bind to miRNAs via miRNA response elements (MREs), thereby affecting miRNA-induced gene silencing and playing an important role in pathological conditions such as cancer [8]. FBXL19-AS1, a lncRNA functions as a ceRNA was widely documented in various cancer studies except for HCC [9][10][11][12][13][14].
In the present study, The Cancer Genome Atlas liver hepatocellular carcinoma (TCGA LIHC) dataset [15] and 6 HCC related microarrays from Gene Expression Omnibus (GEO) database [16] were analyzed combinedly to obtain differentially expressed lncRNAs in HCC tissues. FBXL19-AS1 was ltered to be upregulated in HCC and was predicted to predominantly exist in the cytosol by the LncLocator database [17]. In addition, it was found enriched in in ammation and cancer-related pathways through Gene Set Enrichment Analysis (GSEA). Therefore, our study constructed a rigorous ceRNA network involved in FBXL19-AS1 and demonstrated the possible signi cant value of FBXL19-AS1 in the early diagnosis and prognosis of HCC.

Data source
A part of datasets was obtained from GEO database. We performed comprehensive retrieve on the publicly available HCC non-coding RNA datasets from GEO database (https://www.ncbi.nlm.nih.gov/geo/). Datasets were included in this study based on the following inclusion criteria: (1) datasets contained both normal and HCC samples; (2) HCC subjects were pathologically diagnosed based on clinical and histopathological criteria and without limitation on the clinicopathological stage; (3) type of datasets was non-coding RNA pro ling by array; (4) datasets contained more than 5000 lncRNAs. Based on the criteria above, we downloaded 6 HCC datasets from GEO database, including GSE58043, GSE67260, GSE112613, GSE89186, GSE64631 and GSE70880. In consideration of the limited sample size of single dataset could lead to unreliable results and reduce the effectiveness of bioinformatics analysis, we rst integrated all samples of 6 datasets to signi cantly increase the sample size (39 normal samples vs 44 HCC samples). Heterogeneity and potential variables are generally recognized as the main sources of bias and variability in high-throughput experiments. The merged data was preprocessed by SVA [18] with R software (Version 3.5.3) to remove the batch effect and heterogeneity among virous datasets. If a gene corresponded to multiple probes, we took the average as its expression value. Differentially expressed lncRNAs were screened out using the limma package [19] in R software and the threshold was |log 2 (foldchange) | > 1, adjust P value < 0.05.
The other part of data was from TCGA LIHC dataset (https://portal.gdc.cancer.gov/), which contained 50 normal samples and 374 tumor samples. Differentially expressed lncRNAs were ltered using the edgeR package [20] in R software and the threshold was |log 2 (foldchange)| >1, adjust P value < 0.05. An additional le shows details of each dataset [see Additional le 1].

GSEA
GSEA is a bioinformatics method that inspects the statistical signi cance of a priori de ned set of genes and veri es the differences between two biological states [21]. Samples from TCGA were divided into 2 subgroups on the basis of the median expression of FBXL19-AS1. Genes from each sample were ranked according to the expression difference between the 2 subgroups by GSEA software 4.0. KEGG gene set was analyzed to explore pathways enriched in each subgroup. Gene set permutations were executed for 1000 times in the analysis. Normalized P value < 0.05 was taken as the threshold.

Tissue and Plasma Samples
Surgical specimens were obtained from 57 HCC patients (52 males and 5 females) in Zhongnan Hospital of Wuhan University (Wuhan, China) from 2015 to 2019. None of the patients received preoperative chemotherapy or radiotherapy. The follow-up period ranged from 2 months to 48 months. Whole blood samples were collected during 2017 and 2019 from Zhongnan Hospital of Wuhan University, which contained 79 healthy people, 77 patients with hepatitis B, 80 patients with cirrhosis, and 92 patients with HCC. All whole blood samples were collected into the EDTA anticoagulant tubes and the plasma was isolated at 12000 g for 5min in 4℃. Tissue and plasma samples were stored at -80℃ until use. All patients were diagnosed based on their pathological reports. The tumor stages were identi ed according to the seventh edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual. The detailed clinicopathological information of all patients was shown in Table 1 and Table 2. All experimental schemes were approved by the Ethics Committee of Zhongnan Hospital of Wuhan University.
RNA extraction and quantitative real-time PCR Total RNA was isolated from tissues by TRIZOL reagent (Invitrogen, USA), and RNA from plasm was extracted by RNA Separate Extraction Kit (Bioteke, China). NanoDrop 2000C (Thermo Fisher Scienti c, USA) was applied to evaluate the concentration and purity of extracted RNA. Then ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo, Japan) was used to reversely transcribed RNA into complementary DNA (cDNA) at 37℃ for 15min, 50℃ for 5min and 98℃ for 5min. The quantitative realtime PCR (qPCR) was carried out using SYBR Green I UltraSYBR Mixture (CWBIO, China) on Bio-Rad CFX96 (Bio-Rad Laboratories, USA). We took glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as endogenous reference gene to normalize the expression level among multiple samples. The speci c sequences of each pair of primers were available in an additional le shows [see Additional le 2].
Relative gene expression status was calculated by 2 -ΔCq . All experiments were repeated twice to intensify the credibility.

Survival analysis
Gene Expression Pro ling Interactive Analysis2 (GEPIA2, http://gepia2.cancer-pku.cn/) is an online tool for gene expression and survival analysis based on tumor and normal samples from TCGA and GTEx Databases [22]. Overall survival was evaluated with the Kaplan-Meier method and compared by the logrank test on GEPIA2. We set the dataset as LIHC and retrieved the overall survival of FBXL19-AS1 to obtain the information on the relationship between FBXL19-AS1 expression and prognosis of HCC. To further verify the results, 57 HCC patients from Zhongnan Hospital of Wuhan University were followed up and survival analysis was performed on the basis of FBXL19-AS1 expression status.

Prediction of miRNAs
MiRcode V11 (http://www.mircode.org/) was used for prediction miRNA which would interact with FBXL19-AS1. The highly conserved microRNA families le was downloaded from the miRcode V11 website, and R software was used to predict the complementary miRNAs of FBXL19-AS1. Simultaneously, edgeR package in R software was used to screen out the differentially expressed miRNAs (p<0.05) in HCC based on TCGA. Then we took the intersection of the 2 miRNA lists so as to screened out miRNAs that were both interacted with FBXL19-AS1 and differentially expressed in HCC.

MiRNA expression veri cation
In order to further enhance the credibility of the differential expression of the predicted target miRNAs in HCC, we retrieved studies that contained miRNA expression data from TCGA and GEO for veri cation. All studies included both HCC and normal samples and the sample size of each subgroup was no less than 3. Data were extracted from each study as follows: rst author, year of publication, region, data source, platform, miRNA ID, number of cases, and miRNA expression level. Combined standard mean difference (SMD) and 95% con dence interval (95% CI) were calculated by STATA 15.0 (STATA Corp, USA).
Compared with the normal control, SMD > 0 indicates miRNA is up-regulated in HCC samples, while SMD < 0 indicates miRNA is down-regulated in HCC samples. Statistically signi cant threshold of two-sided P value was set at 0.05.

MiRNA targets prediction
To ensure the miRNA-mRNA interactions conserved in essential cancer pathways, target genes of miRNAs supported simultaneously by miRDB, miRTarBase, and TargetScan were selected by R software.
Meanwhile, edgeR package in R software was used to screen out the differentially expressed mRNAs (p<0.05) in HCC based on TCGA. In addition, genes co-expressed with FBXL19-AS1 were obtained from cBioPortal database (P < 0.05). Finally, the intersection of the 3 lists was taken as the nal target mRNAs.

Establishment of lncRNA-miRNA-mRNA expression network
We constructed a ceRNA network for FBXL19-AS1, target miRNAs and mRNAs. Cytoscape 3.7.2 software was used to visualize the network.

PPI network construction and hub genes selection
Analysis of PPI network is helpful in systematically studying the molecular mechanism of diseases and nding new drug targets. In our study, STRING (V11.0) (https://string-db.org/) [23] was adopted to establish a PPI network, and 0.4 was used as the threshold for interaction score. Subsequently, 12 kinds of algorithms (MCC, DMNC, MNC, Degree, EPC, BottleNeck, EcCentricity, Closeness, Radiality, Betweenness, Stress, ClusteringCoe cient) were jointly used to identify hub genes through Cytoscape 3.7.2 software [24]. We took genes whose sum algorithm scores were more than 10000 to construct hub gene network.
Functional enrichment analysis of hub genes R software and clusterPro ler package [25] were used to execute the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis [26,27]. Subsequently, the ggplot2 package was used to visualize the results, and the cut-off value of statistical signi cance was set to P < 0.05.

Statistical analysis
All statistical analyses of this research were conducted through SPSS version 25.0 (SPSS Inc., USA) and GraphPad Prism 8.0 (GraphPad Inc., USA). Mean ± standard deviation (SD) was used to describe the continuous variables of normal distribution. Median and quartile was used to describe the continuous variable of skewed distribution. The paired sample t test or Kruskal-Wallis test were utilized to compare the differences between the two groups. The chi-square test or Fisher exact test were used to compare the categorical variables between groups. Correlation analysis was performed by Pearson or Spearman test. Receiver operation curve (ROC) was used to evaluate the diagnostic value. The best cutoff point for sensitivity and speci city was selected by using the Jorden index. Overall survival was evaluated by Kaplan-Meier method and compared by log-rank test. The cutoff value of statistical signi cance was set as P < 0.05.

Results
The long noncoding RNA expression pro le of HCC The research ow diagram of this study was shown in Fig. 1. To identify lncRNAs that were differentially expressed in HCC, 6 GEO datasets were integrated into analysis (39 normal samples vs 44 HCC samples), including GSE58043, GSE67260, GSE112613, GSE89186, GSE64631 and GSE70880. We obtained 66 differentially expressed lncRNAs, among which 37 were up-regulated and 29 were down-regulated (Fig.  2a). Then we downloaded the relevant expression pro les of the TCGA LIHC dataset (50 normal samples vs 374 HCC samples) and obtained 2685 differentially expressed lncRNAs, among which 2323 were upregulated and 362 were down-regulated (Fig. 2b). There were 26 lncRNAs that both differentially expressed in GEO joint dataset and TCGA LIHC dataset, among which 15 were up-regulated and 11 were down-regulated (Fig. 2c). An additional le shows these in more detail [see Additional le 3]. The expression heatmaps of GEO joint dataset and TCGA LIHC were displayed in Fig. 2d and Fig. 2e, respectively.
The potential interactive miRNAs of 26 differentially expressed lncRNAs were screened by R software based on the highly conserved microRNA families le downloaded from the miRcode V11 database. R software predicted the target genes of corresponding miRNAs that were simultaneously supported in miRDB, miRTarBase and TargetScan. Only 7 lncRNAs (LINC00221, FAM99B, LINC00355, MAGI2-AS3, CRNDE, PWRN1, FBXL19-AS1) were predicted to have highly conserved targeted miRNAs [see Additional le 4]. We used GEPIA2 to carry out survival analyses for these 7 lncRNAs, and found that only FBXL19-AS1 (Fig. 3a), FAM99B (Fig. 3b) and CRNDE (Fig. 3c) were associated with the prognosis of HCC patients. Among the 3 lncRNAs, the function of FBXL19-AS1 in HCC has not yet been studied and FBXL19-AS1 might serve as a novel HCC biomarker. Therefore, FBXL19-AS1 was selected for further study.
FBXL19-AS1 was signi cantly up-regulated in both GEO joint dataset and TCGA LIHC datasets and mainly enriched on cell cycle, cancer and in ammation-related pathways by GSEA (Fig. 3d). We speculated that FBXL19-AS1 might play an important role in the occurrence, development and prognosis of HCC. In addition, we found that FBXL19-AS1 was located in cytosol according to the LncLocator database [see Additional le 5], which was the basis of establishing a more reliable ceRNA network. Expression of FBXL19-AS1 in HCC tissues and its prognostic value In order to verify the expression status of FBXL19-AS1 in HCC patients and to study its clinical signi cance, we collected 57 pairs of fresh tissues including HCC and adjacent non-tumor tissues. The qPCR results showed that the expression of FBXL19-AS1 in HCC tissue was signi cantly higher than adjacent non-tumor tissue (Fig. 4a).
The results on the study of the correlation between FBXL19-AS1 expression and clinicopathological features showed in Table1 that the over expression of FBXL19-AS1 was signi cantly correlated with higher GGT (P = 0.046), higher AFP (P = 0.020) and worse TNM stage (P = 0.043), which was consistent with what we predicted in GEPIA2 [see Additional le 4].
To further verify the prognostic value of FBXL19-AS1 in HCC, we performed Kaplan-Meier survival analysis and log-rank test in 57 HCC patients with intact prognostic information. The results showed that HCC patients with elevated FBXL19-AS1 had shorter overall survival (P = 0.030) (Fig. 4b), hinting FBXL19-AS1 might be an important prognostic factor for HCC.
Expression of FBXL19-AS1 in plasma and its diagnostic value FBXL19-AS1 expression level was also checked through qPCR in plasma samples, which were taken from 79 healthy subjects, 77 patients with hepatitis B, 80 patients with cirrhosis, and 92 patients with HCC (Fig.  4c). The results showed that the expression of FBXL19-AS1 in plasma was signi cantly higher in hepatitis B, cirrhosis, and HCC patients than in healthy subjects (P < 0.001). While the expression of FBXL19-AS1 in HCC patients was higher than that of hepatitis B patients (P = 0.016) and cirrhosis patients (P = 0.004). The main clinical features of the plasma samples were shown in Table 2. From which the expression of FBXL19-AS1 in each group was correlated with the plasma alanine aminotransferase (ALT) (P < 0.001), aspartate aminotransferase (AST) (P < 0.001), albumin (ALB) (P < 0.001), alpha fetoprotein (AFP) (P < 0.001), but not with gender, age, CEA or other biochemical indicators.
ROCs of FBXL19-AS1 in HCC, drawn to evaluate the diagnostic value, indicated FBXL19-AS1 was with moderate diagnostic ability to distinguish HCC patients from healthy people (AUC = 0.875, P < 0.001).
LncRNA-miRNA-mRNA network construction To further explore the potential downstream targets of these 3 miRNAs, three online bioinformatics servers (miRDB, miRTarBase and TargetScan) were used. There were 399 target genes of these 3 miRNAs simultaneously supported by all three databases. Then, we screened out 12,841 differentially expressed mRNAs in TCGA LIHC (P < 0.05). In addition, 12,194 mRNAs were predicted to be co-expressed with FBXL19-AS1 in HCC by cBioportal database (P < 0.05). Finally, 205 mRNAs were selected as targets through the intersection of the above three gene sets.
A new ceRNA network was formed among lncRNA (FBXL19-AS1), three miRNAs (hsa-miR-22-3p, hsa-miR-20b-5p, hsa-miR-142-3p) and 205 mRNAs (Fig. 8a). The diamond in the middle represented FBXL19-AS1, the gray triangles were miRNAs, and the circles were mRNAs. The circles in red meant the corresponding mRNAs were elevated in HCC, blue circles represented the decreased expression of related mRNA in HCC, deeper color indicated increased logFC, and larger size indicated smaller P value.

PPI network construction and screening of hub genes
We constructed a PPI network of these 205 mRNAs based on the STRING database and then visualized by Cytoscape 3.7.2. After removing the free nodes, the PPI network containing 158 nodes and 272 edges (Fig. 8b). Thereafter, hub genes identi ed by 12 algorithms (MCC DMNC MNC Degree EPC BottleNeck EcCentricity Closeness Radiality Betweenness Stress ClusteringCoe cient) constituted a subnetwork with 9 nodes and 9 edges (Fig. 8c), which revealed the 9 hub genes (STAT3, CNOT7 BTG3, E2F1, TRIM37, YWHAZ, RBBP7, KIF23, ESR1) played important roles in the pathogenesis of HCC.
Functional analysis of 9 hub genes GO and KEGG enrichment analyses were performed on the 9 hub genes (P < 0.05). Top 15 terms and pathways were selected for demonstration by P value. GO functional enrichment analysis revealed hub genes mainly enriched in transcription factor activity and transcriptional activator activity (Fig. 9a, Table  4). KEGG pathway analysis indicated the 9 hub genes might in uence the occurrence and progression of HCC by participating in pathways such as hepatitis C, hepatitis B, microRNAs in cancer, cell cycle, viral carcinogenesis, and proteoglycans in cancer (Fig. 9b, Table 5). In addition, the 9 hub genes were also involved in pathways associated with non-small cell lung cancer, pancreatic cancer, breast cancer and other diseases.

Veri cation of ceRNA network
In order to establish a more reliable ceRNA network, we performed correlation analyses among FBXL19-AS1, 2 miRNAs and 9 mRNAs based on 370 HCC tissues and 50 normal tissues from TCGA LIHC dataset.

Discussion
In our study, FBXL19-AS1 was identi ed as a potential oncogene through integrated analysis of 6 GEO microarray datasets and TCGA LIHC dataset. Previous studies have shown that FBXL19-AS1 is signi cantly increased in breast cancer [9][10]28], lung cancer [11][12]29], osteosarcoma [13] and colorectal cancer [14], and participates in the migration, proliferation and survival of tumor cells. The underlying function of FBXL19-AS1 in HCC has not been studied yet. To verify the bioinformatics results and explore the role of FBXL19-AS1 played in HCC, qPCR was used to assess that the expression status of FBXL19-AS1. The experimental results indicated FBXL19-AS1 was elevated in HCC and its expression was correlated with TNM stage, AFP and GGT. Given that elevated GGT is associated with the occurrence of acute and chronic hepatitis and alcoholic liver disease, it can be inferred that FBXL19-AS1 may be involved in the development of related diseases. In combination with the GEPIA2 survival analysis results and our follow-up study on 57 patients, we found high expression of FBXL19-AS1 was associated with poor prognosis in HCC.
Biomarkers screened from liver tissues are not suitable for early diagnosis of HCC, and the speci city and sensitivity of AFP, the most widely used plasma biomarker in HCC diagnosis, are quite limited. In order to make up for the de ciency of early diagnosis of HCC, we evaluated the diagnostic value of plasma FBXL19-AS1. The results showed that the plasma FBXL19-AS1 in patients with hepatitis B, cirrhosis and HCC was signi cantly higher than that of healthy subjects. ROC analysis revealed plasma FBXL19-AS1 was with satisfactory diagnostic value in differentiating healthy controls from patients with hepatitis B, cirrhosis and especially HCC. Whereas, the discernibility ability of FBXL19-AS1 in hepatitis B patients and cirrhosis patients was unsatisfactory, which could be partially explained by the pathological similarity of the patients. It should be noted that the combination of plasma FBXL19-AS1 and AFP could signi cantly improve the diagnosis for HCC, suggesting that FBXL19-AS1 could serve as a biomarker for the auxiliary diagnosis of HCC. it is also important to note that FBXL19-AS1 has been reported to be associated with a variety of cancers, and therefore the need to combine FBXL19-AS1 and AFP to enhance the diagnostic speci city of HCC should be emphasized.
We analyzed the correlation between FBXL19-AS1 and these 7 miRNAs based on TCGA LIHC dataset, and nally selected 3 miRNAs (hsa-miR-22-3p, hsa-miR-20b-5p, hsa-miR-142-3p) that met our standards for subsequent analysis. Further studies revealed that FBXL19-AS1 may act as ceRNA to competitively bind to the above 3 miRNAs and regulate the expression of 205 mRNAs. In order to elucidate the ceRNA regulatory mechanism, we established a PPI network and obtained 9 hub genes (STAT3, CNOT7 BTG3, E2F1, TRIM37, YWHAZ, RBBP7, KIF23, ESR1). GO functional annotation and KEGG pathway analysis revealed the 9 hub genes were enriched in hepatitis B associated pathways and the important roles of these 9 hub genes in HCC have also been con rmed [37][38][39][40][41][42][43][44][45]. However, the relationship between the 9 hub genes and FBXL19-AS1 has not been reported. Through correlation analysis based on 420 TCGA LIHC tissue samples, STAT3 and CNOT7 were excluded, and the remaining 7 mRNAs were all associated with the prognosis of HCC. Finally, we constructed a more reliable ceRNA network consisting of FBXL19-AS1, hsa-miR-22-3p, hsa-miR-20b-5p, 7 hub genes (BTG3, E2F1, TRIM37, YWHAZ, RBBP7, KIF23, ESR1), and 7 lncRNA-miRNA-mRNA regulatory axes. A study reported miR-22 and miR-20b were involved in the progression of HBV-related HCC which further improved the credibility of our research [46]. Intriguingly, we found FBXL19, the complementary gene of FBXL19-AS1, was associated with tumor immune invasion according to TIMER database. Thus, we speculated FBXL19-AS1 might form RNA-RNA dimer with FBXL19 through the classical pattern, which might affect the expression of FBXL19 and affected the immune in ltration in HCC.

Conclusions
In summary, FBXL19-AS1 was identi ed as an oncogenic lncRNA that may serve as a diagnostic and prognostic agent for HCC potential biomarkers. We also established a FBXL19-AS1-miRNA-mRNA network, and demonstrated that FBXL19-AS1 might participate in the pathological progression of HCC as a ceRNA. Our study elucidates the potential oncogenic pathways involved in FBXL19-AS1 and recognizes the role of FBXL19-AS1 in the possible target genes in HCC. Our ndings might provide new perspectives on the pathogenesis of HCC, thus broadening the therapeutic options for HCC.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.   Tables   Table 1 Relationship between FBXL19-AS1 expression in tissues and clinical characteristics of HCC patients. ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, γ-Glutamyl Transferase; AFP, alphafetoprotein; CEA, carcinoembryonic antigen; HBV, hepatitis B virus.   File format: .pdf Title: FBXL19 is associated with HCC immune in ltration.
Description: Additional le 6. FBXL19 is associated with HCC immune in ltration.
File name: Additional le 7 File format: .pdf Title: Violin diagram of the relationship between FBXL19-AS1 and clinical stage.
Description: Additional le 7. Violin diagram of the relationship between FBXL19-AS1 and clinical stage.

Page 22/33
File name: Additional le 8 File format: .pdf Title: Essential information of the studies for the 7 miRNAs derived from GEO and TCGA database.
Description: Additional le 8. Essential information of the studies for the 7 miRNAs derived from GEO and TCGA database. Figure 1 Flow diagram of the analysis process.    Heatmap for the signaling pathways from DIANA-miRPath in which the 7 miRNAs are involved.

Figure 9
Functional analysis of 9 hub genes. a GO functional enrichment analysis of 9 hub genes. b KEGG pathway enrichment analysis of 9 hub genes.

Figure 10
The correlation analysis of FBXL19-AS1, 2 miRNAs and 9 hub genes. a Correlation heatmap of the whole network. b-o The correlation analyses which were statistically signi cant.

Figure 11
Overall survival and disease-free survival of 7 hub genes. a-g Overall survival of 7 hub genes. h-n Diseasefree survival of 7 hub genes.