A Novel Analysis of the Differentiated Expression of Long Noncoding RNAs Proles in Human Prostate Cancer

Background: Long noncoding RNAs (lncRNAs) have crucial roles in cancer biology. Increasing numbers of evidences have indicated that lncRNAs play an important role in the pathogenesis, invasion, and metastasis in almost all kinds of cancers. But, compared with the large amounts of patients, there is rare reports that showed the differential expression of lncRNAs in prostate cancer. Methods: In this study, lncRNA expression proles were screened in PCa, by using 5 pairs of clinical specimens in PCa and matched non-PCa tissues with lncRNA chip. To arm further clinical value, we extended the samples consisting of another 5 tumor specimens and 7 para-cancerous/benign contrasts by qRT-PCR in top 10 up-regulated and down-regulated lncRNAs. Results: A total of 817 lncRNAs were differentially expressed between PCa tumor and para-cancerous tissues (Fold Change ≥ 2.0, p < 0.05): 422 were upregulated, whereas 395 were downregulated in PCa tissues. Gene ontology and KEGG pathway analyses showed that many lncRNAs were implicated in carcinogenesis. Among differentially expressed lncRNAs, lnc-MYL2-4:1 (FC = 0.00141, p = 0.01909) and NR_125857 (FC = 59.27658, p = 0.00128) had the highest magnitude of change. The subsequent qPCR conrmed the expression of NR_125857 accorded with the clinical samples. Conclusions: Our study detected a relatively novel complicated map of lncRNAs in PCa, which may have the potential to investigate for diagnosis, treatment and follow-up in PCa. Our study revealed the expression of NR_125857 in human PCa tissues was most up-regulated. Further study of these meaningful candidates are need to research deep mechanisms.

the frequency of AR-null CRPC is increasing, because of the application of more effective AR antagonists such as enzalutamide and abiraterone [6]. Since the effort of urologists for the patient in the end stage of this disease is limited, it is imperative for the scientists to progress effective biomarkers for very early detection and active target for clinical treatment.
Accumulating genomic and transcriptomic sequencing results have revealed that only small proportion of the human genome is transcribed into protein-coding mRNAs, whereas the majority of the genome is transcribed into ncRNAs [8,9]. Amongst the classes of ncRNAs, long noncoding RNAs (lncRNAs) are a class of transcripts longer than 200 nucleotides with limited protein coding potential [10]. Unlike microRNAs or proteins, lncRNAs function cannot be currently inferred from sequence or structure, with the diversity of long ncRNAs described to date precluding simple generalizations. The broad functional repertoire of long ncRNAs includes roles in high-order chromosomal dynamics, telomere biology and subcellular structural organization [11]. LncRNAs regulate local protein-coding gene expression at the level of chromatin remodeling, transcriptional control and post-transcriptional processing, which suggests that RNA has continued to evolve and expand alongside proteins and DNA and indicate they have multiple functions in a wide range of biological processes, such as proliferation, apoptosis, or cell migration [10,12]. Unlike miRNAs, lncRNAs are able to fold into secondary and tertiary structures by which they carry out their function [13]. Upward high-throughput transcriptomics studies propose long noncoding RNA (lncRNA) dysregulation in various diseases, including neuroblastoma, pancreatic ductal adenocarcinoma, lung cancer and other cancers through corresponding miRNAs [14][15][16][17]. Moreover, they are detected in circulating blood and/or urine [18][19][20]. LncRNAs are a novel class of potential biomarkers and therapeutic targets for the treatment of cancer [21].
Nevertheless, the function of most lncRNAs is still unknown. A growing amount of evidence has showed that lncRNAs play a vital role in the progression of PCa [22]. Especially, the expression levels and potential roles of lncRNAs in PCa are need to be further studied [23]. Through additional discovery of molecular mechanisms in lncRNAs to broaden the eld of PCa, patients can pro t from developing more effective healing innovations. Generally, the consequences of lncRNA chip vary from samples to samples.
In this study, we considered the differential expression of lncRNAs in PCa tissues to nd out several vital and potential biomarkers so that we could construct an independent bank of lncRNAs in PCa for research among the target cohorts in southeast coastal areas of China. Therefore, to provide research subjects in the further study, we rstly gured out the expression difference of lncRNAs in PCa through chip analysis, and then we carried out the validation of the distensible clinical samples.

Material And Method
Tissue samples A group of ve pairs of PCa and matched non-tumor normal tissues were collected from Huashan Hospital, Fudan University. To deep con rm, another cohort of prostate tissues were obtained from prostate needle biopsies in Huashan Hospital, Fudan University. Our study was permitted by the ethics committee of Huashan Hospital, Fudan University and written informed consent was obtained from all patients. All tissue was histologically identi ed by pathological section. If diagnosed as prostate adenocarcinoma, the Gleason score, PSA value, TNM stage and recurrence were according to the NCCN guideline. Otherwise, the tissues are recognized as normal contrast. A subset of patients had matched PCa tissues and normal tissues available in the qPCR. The initial screening step (Table 1) was conducted with microarray chip assay. Another cohort screening information, which was considered as the validation of the expanded clinical samples (Table 2), was listed with the qPCR.

Bioinformatics Analysis
LncRNA targets correlated with mRNAs were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses using GO (http://www.geneongoloty.org/) and KOBAS software (KEGG Orthology-Based Annotation System, https://www.kegg.jp/). The differentially expressed lncRNAs-targeted miRNAs were sought and predicted by miRanda software (http://miranda.org.uk/) coupled with statistical analysis. The lncRNAs expression pro le for microarray chip assay, besides data and bioinformatics analysis were carried out by Shanghai Biotechnology Corporation (Shanghai, China).

Qpcr Analysis
Total RNA from another normal tissues (9 samples) and PCa tissues (7 samples) sustained by pathology after perineal prostate biopsy guided by ultrasound was prepared by Trizol Isolation Reagent (Invitrogen). Dimethylcarbinol, ethanol and trichloromethane were of analytic grade. DNase I, SYBR Green Realtime PCR Master Mix Plus and the ReverTra Ace qPCR RT Kit are from Toyobo Co. Japan. The reverse transcription kit steps are strictly followed to transcribe to cDNA. cDNA was used as template, and hGAPDH as internal parameter. The primer concentration was set as 0.4 µmol/L. Three parallel samples were set for each sample, tested as 15 µl system used for ampli cation. For qPCR solution, THUNDERBIRD SYBR qPCR Mix (Toyobo, Osaka, Japan) was utilized. qPCR was performed on the LightCycler 96 (Roche, Indianapolis, IN, USA) following the instruction. The reaction conditions of qPCR were: pre-denaturation at 95 °C for 3 min, denaturation at 95 °C for 15 s, annealing at 60 °C for 15 s, extension at 72 °C for 20 s and totally 40 cycles. Assay numbers got involved in the top 10 up-regulated and down-regulated expression of lncRNAs and GAPDH, respectively. The sequences of primer are listed in the Table 1. Differentiated gene expression was calculated by the comparative Ct method.

The Co-expression Network Of Lncrna-mirna-mrna
To supplementary achieve perceptions of the lncRNAs' biological functions in the complex biological processes and cellular regulation, the lncRNA-miRNA-mRNA co-expression network was constructed to investigate the potential interaction between miRNAs, mRNAs and lncRNAs. As shown in the Fig. 5, the co-expression network of lncRNA-miRNA-mRNA included 20 nodes of miRNAs and 84 connections consisting of various lncRNAs and mRNAs. Among the 17 networks, one of the most known coexpression networks was miR-17-5p because it had been proven that miR-17-5p repressed metallopeptidase inhibitor 3 expression in PCa while in this study we found the network of miR-17-5p also got involved in the gene EIF3H, HELLS and DNAL1, which was regulated by the same lncRNA URS000048C392 (also named ENST00000555037.1) [24]. With one edge networks like URS00008B6496(ENST00000547292.1), URS00000B8AF9(ENST00000482003.1), URS0000EEB1F2(ENST00000436764) and URS00007CEE5E(lnc-DHX38-3:6), it should be simple to con rm their roles in PCa by further experiment. Some complicated networks like URS00008C2FEF(ENST00000591956), URS00008BBA94(ENST00000452731) and URS00009BE037(ENST00000492250) were associated with two diverse miRNA signal pathways, which indicated their might have different in uence on PCa. URS00005D043E(ENST00000464382), URS000046AFA0(ENST00000534169) and URS0000EF6BD5(ENST00000435802) were connected to the same miR-375,and URS0000DB7AD5(ENST00000580175) and URS000032BFFB(ENST00000558749) were affected by miR-582-5p in the meanwhile. Although with several edges in mRNAs, the rest of lncRNA and miRNAs had the relationship of one-one correspondence. As demonstrated, those lncRNAs, miRNAs and mRNAs were vastly linked as the key hub of the co-expression network, which implied their vitally potential impact on lncRNAs in the progress of regulating particular target genes in PCa.

Survival Curve Analysis
We used the GEPIA (http://gepia.cancer-pku.cn/) as tool to calculate the survival curve of the top 10 upregulated and downregulated lncRNAs original gene.

Statistical analysis
All data are shown as mean ± standard deviation (SD). Statistical signi cance was determined using Student's t-test by SPSS 13.0 and Graphpad Prism 5. was considered statistically signi cant. Result

LncRNAs expression pro les in PCa
The microarray screening identi ed 68,424 lncRNAs in PCa, non-PCa or both tissues. As illustrated in Fig. 1, totally, 817 lncRNAs were differentially expressed between PCa tumor and paracancerous tissues ( ) (Table S2): among which 422 were upregulated, and the remaining 395 were downregulated in PCa tissues. The magnitude of FC was the highest for NR_125857 for upregulated lncRNAs ( ) while it was the lowest for lnc-MYL2-4:1 in downregulated lncRNAs ( ). Hierarchical clustering (Fig. 2D), volcano plot (Fig. 3), and scatter plots (Fig. 3) shown that the different expression pro les of lncRNAs between PCa and non-PCa tissues were diverse. The top each twenty up-and down-regulated lncRNAs were listed in Table 4. Moreover, the upregulation of the neuron formation shows the nerve paracrine factor involving in the tumorigenesis. Except for the famous pathways, such as TGF-, Wnt, MARK and mTOR that have been proven to be closely correlated to proliferation, invasion and metastasis in PCa, astonishingly, the pathway of aldosterone-regulated sodium reabsorption, dilated cardiomyopathy, hypertrophic cardiomyopathy, pathogenic Escherichia coli infection and vascular smooth muscle contraction also implies the revegetation of smooth muscle may interfere with the microenvironment of PCa. Additionally, the pathogenic Escherichia coli infection may link to the common urinary disease, prostatitis, which also causes the tissue recovery (Fig. 4).

The Results Of Qpcr
The outcome of qPCR showed signi cant statistic differences in NR_125857, NR_015342, NR_109832, ENST00000412654, lnc-AC110080.1-5:1, ENST00000415820, ENST00000558010 ( ) in PCa tissues (n = 5) compared to normal prostate tissues (n = 7) while there was no statistic difference of all the top ten downregulated lncRNA expression (Fig. 6A, B). These Compared with the original outcomes of RNA-seq array, the relative expression of qPCR in the additional samples showed consistency in gure (Fig. 6C).

The Analysis Of Survival Curve
As the top seven upregulated lncRNAs in our study revealed the coherence of bioinformatics analysis and qPCR analysis, we further analyzed their survival curves of original gene in PCa by the tool of GEPIA (http://gepia.cancer-pku.cn/) (Fig. 7). Higher expression of prostate-speci c DD3(PCa3) in patients of PCa showed lower survival rate after about 80 months while the higher expression of PCa associated transcript-14 (PCAT14) demonstrated higher survival rate since approximately 60 months. The high expression of AP001610.9 led to a dramatic decline of survival rate after 110 months despite the phenomenon that it revealed moderately higher survival rate from the 80th to 110th month. Moreover, differentiated expression of RP11-279F6.2 showed a subtle difference that the high expression would result in lower survival rate in the duration of 80th and 105th month. Nevertheless, there was no recorded data of NR_125857, which was the most upregulated lncRNA in our study.

Discussion
Traditionally, by means of measuring the assay of PSA protein in the blood, folk with high risks in PCa screen for that illness. However, in the United States, the US Preventive Services Task Force (USPSTF) reviewed the evidence on the bene ts and harms of PSA-based screening for PCa and subsequent treatment of screen-detected PCa [25]. For men aged 55 to 69 years, the decision to undergo periodic PSA-based screening for PCa should be various from person to person and should take the balance of the possible bene ts and troubles of screening PSA with their clinician [26]. Through screening PSA, for some men, it just gives the patients a limited potential pro t of the reductions in risk of dying of PCa and metastatic disease in the end stage [27]. Many men will suffer potential harms of screening, including false-positive results that entail extra testing and even, invasive prostate biopsy, to separate them from the real patients; overdiagnosis and overtreatment; and it may arouse a lot of treatment complications, such as incontinence and erectile dysfunction, which wastes unnecessary time, in uence the normal life and lowers the quality of life [28][29][30]. Clinicians should not screen men who do not express a preference for screening (C recommendation) [31][32][33]. The USPSTF recommends against PSA-based screening for PCa in men 70 years and older (D recommendation) [25,34,35]. As there is progressively amounts of argument and distrust about the speci city and sensitivity of PSA, it is essential to develop more dependable biomarkers for early screening of PCa. Recent developments in the detection of lncRNAs have acknowledged lineage-and cancer-speci c biomarkers that may be applicable in the clinical utilization of PCa. Herein, we will combine our analysis of RNA-seq datasets, from 5 patient samples, including PCa and adjacent benign prostate tissue with the other investigation to exploit and corroborate differentially expressed lncRNA connected with PCa. After we concluded the lncRNA candidates, which had characteristics of the most signi cant changes, we proved the consistency of RNA-seq results and qRT-PCR outcomes by extending the clinical samples that consisted of another 5 tumor specimens and 7 para-cancerous/benign contrasts through prostate biopsy. The top each twenty up-regulated and downregulated lncRNAs was listed in Table 4.
We noticed that NR_125857, related to the gene EVADR, ranked the rst line of upregulation in our database. EVADR is the written abbreviation of Endogenous retroViral-associated ADenocarcinoma RNA (EVADR), by analyzing RNA-seq data derived from colorectal tumors and matched normal control tissues [36]. This lncRNA demonstrated nominal to low expression in normal tissue, but is signi cantly upregulated in cancer, particularly in colon, rectal, lung, stomach and pancreas adenocarcinomas. It was reported the EVADR lncRNA determined the promoter activity of the MER48 long terminal repeat (LTR) in vitro, mapped the genome-wide MER48 LTR expression [36,37]. Regardless of a biological function, the speci city of EVADR activation in adenocarcinomas coupled with the poorer survival probability that tracks with elevated EVADR expression suggested that further characterization of EVADR as a candidate adenocarcinoma biomarker is warranted [36].Nevertheless, the original article did not mention any details about the EVADR in PCa. In our study, it was totally clear that the expression of NR_125857 is upregulated in PCa by RNA-seq and qPCR. Since it was described as the highest upregulated lncRNA in our research, it seemed to be promising in the PCa research, for, without any doubt, PCa is also a kind of adenocarcinoma. The mechanism mediated by the high expression of NR_125857 in PCa requires further cavernous research and investigation.
In the top ve of the upregulation in lncRNA, NR_015342 and ENST00000412654 are associated with the PCa3, accounting for a large proportion. PCa3 was located on chromosome 9q21-22 [38]. PCa3, as one of the oldest identi ed lncRNAs, is an accepted diagnostic urinary biomarker for PCa [39]. Because PCa3 is over-expressed in 95% of PCa, with up to 100-fold up-regulation compared to adjacent non-neoplastic cells [40,41]. Highly overexpression of PCa3 in PCa tissue was found to be a potential non-invasively prediction of prostate biopsy which might be a promising biomarker in clinical diagnosis [42]. However, PCa3 assays also have limited utility in detecting men with higher grade diseases due to low PCa3 levels [43]. For instance, a patient was observed to be negative for PCa as assessed by urinary PCa3, but was later diagnosed to have very high-grade disease (Gleason Score 9) and high Decipher metastasis risk [44]. Thus, it warns us that single usage of PCa3 as a stand-alone marker for PCa may deliver false negative outcomes for patients with higher grade tumor.
Ranking at the third up-regulation of genes, NR_109832 suggests the gene PCAT14 also play an important role in PCa tumorigenesis. PCAT-14 is commonly up-regulated in primary tumors. PCAT14 is an AR-regulated transcript while PCAT14 is highly expressed in low grade disease and loss of PCAT14 predicts for disease aggressiveness and recurrence, and its overexpression suppresses invasion of PCa cells [45]. PCAT14 lower expression is signi cantly prognostic for multiple clinical endpoints supporting its signi cance for predicting metastatic disease that could be used to improve patient management [46].
The sixth up-regulated gene symbol is related with AP001610.9, and ENST00000415820 may links to LOC111099027, LOC105372809, TMPRSS2 and MX1. TMPRSS2, also named as PP9284 or PRSS10, is transmembrane serine protease 2, which is a member of the membrane-anchored serine proteases family [47]. It has been gured out that TMPRSS2 mediates a proteolytic cascade regulated by androgen signaling, which promotes the progression, invasion, and metastasis of PCa cells by activating the matriptase and disordering the extracellular matrix [48][49][50]. TMPRSS2 mainly affects degradation of extracellular matrix nidogen-1 and laminin β1 [48]. Therefore, it indicates an innovative approach for targeting these two proteases in treatment development, and the intimate connection between tumor cells and extracellular matrix in the PCa. Moreover, MX1 has many aliases, such as IFI-78K, IFI78, MX, MxA and lncMX1-215. It belongs to the class of dynamin-like large guanosine triphosphatases (GTPases) acknowledged to be involved in intracellular vesicle tra cking and organelle homeostasis, which chie y participates in the cellular antiviral response against a wide range of RNA viruses, including in uenza viruses and members of the bunyavirus family [51]. It is an interferon stimulated antiviral protein that is required for a complete antiviral response [52]. Preceding study found that down-regulation of MxA in LNCaP cells by dihydrotestosterone suggests that MxA appears to be meaningfully associated with cell cycle and further cancer development while the loss of MxA expression leads to increased metastasis and decreased sensitivity to Docetaxel, which shows that MxA expression could regulate the outcome of chemotherapy [53,54].
Although the ENST00000365110 ranked the eighth contender, its interrelated gene SNORA62 has the full name as small nucleolar RNA, H/ACA box 62. snoRNAs are one of the most ancient and numerous families of non-protein-coding RNAs. Eukaryotic snoRNAs are conserved from archaeal sRNAs in both function and structure [55]. Therefore, the main function of snoRNAs -to guide site-speci c rRNA modi cation -is the same in archaea and all eukaryotic lineages [56]. Owing to the presence of a conserved H box (5′-ANANNA-3′) and an ACA sequence, H/ACA box snoRNAs form a double hairpin structure [55]. The box H/ACA snoRNAs direct the site-speci c pseudouridylation of pre-rRNAs [57].Recent ndings have proven that deregulation of the pseudouridinylation process is connected with the progression of PCa [58]. Another research demonstrates SNORA62 are encoded from the host gene RPSA or Laminin receptor (LAMR) while it is observed that the mutations in the LAMR/RPSA gene may be related to congenital asplenia, the inborn absence of spleen [59,60].
The lowest down-regulation lncRNA is the anonymous lnc-MYL2-4:1. In our study, it suggests this lncRNA is interrelated to myosins, which are a large and diverse family of molecular motors important for cell migration and motility [61]. In PCa, Myo1b, Myo6, Myo9b, Myo10, and Myo18a were expressed at higher levels in high metastatic potential cells, and especially Myo1b and Myo10 were expressed at higher levels in metastatic tumors [62][63][64]. Changes in expression of several myosin isoforms may contribute to metastasis in PCa [62]. Though the outcome of qPCR in this study was no signi cant different in PCa tissues and normal tissues, the exact interaction between our candidate lncRNA and myosin is still needed to research.
The second down-regulation lncRNA lnc-C19orf73-1:1 is related to histidine rich calcium binding protein (HRC). The HRC is a novel regulator of sarcoplasmic reticulum (SR) Ca2+-uptake, storage and release, so the HRC plays a pivotal role in Ca2+-homeostasis.2 Calcium (Ca2+) is an essential intracellular signaling molecule involved in the regulation of cancer progression, including cell proliferation, apoptosis, invasion and migration [65,66]. It has been proved that HRC promotes growth of hepatocellular carcinoma in vitro and in vivo [67]. Furthermore, HRC also plays a signi cant role in myocyte differentiation and in antiapoptotic cardioprotection against ischemia/reperfusion induced cardiac injury [68].
Lnc-MID1-4:1, located on the chromosome X, is associated with Rho GTPase activating protein 6. Rho GTPases have been gured out to be critical signal transducers, which mediate growth factor-induced changes to the actin cytoskeleton and activating the phagocyte NADPH oxidase [69]. As a result, they get involved in abundant cellular processes. For example, cell migration, cell survival, transcriptional regulation and vesicle transferring [70]. The deleted in liver cancer 1 (DLC-1) gene encodes a GTPase activating protein that acts as a negative regulator of the Rho family of small GTPases, and DLC-1 is assumed as a bona de tumor suppressor gene in different types of human cancer [71]. It hints that the down-regulation of Lnc-MID1-4:1 may in uence on the particular cellular functions in PCa.
lnc-PDCD11-5:1 is connected to neutralized E3 ubiquitin protein ligase 1. The E3 ubiquitin ligase NEDD4 negatively regulates HER3/ErbB3 level and signaling [72]. Many preceding studies reveals NEDD4 has been acknowledged to play a critical role in the regulation of a number of membrane receptors, endocytic machinery components and the tumor suppressor PTEN [73]. The loss of PTEN expression was associated with worse survival and shorter time on abiraterone treatment [74]. Ubiquitin Ligases are also involved in the regulation of Wnt, TGF-β, and Notch Signaling Pathways [75].
There is scarcely any information about the rest of up-regulation lncRNAs, such as ENST00000558010, ENST00000365110, NONHSAT072254, NONHSAT072236, lnc-PTEN-11:1, and ENST00000439575. More experiments are needed to prove their impression and function.
In our analysis, there are ten quali ed samples, so our study still has boundedness in the number of samples. Yet the feature of our study was that our patients are typical Mongoloid men and our results exhibited very speci city in east Asia area. Moreover, to highlight the coherence of our outcomes and practical issues and value, we further extended the clinical samples for qPCR and drew the survival curves of meaningful genes of lncRNAs after the con rmation of qPCR. The top seven upregulation lncRNAs, like NR_125857, NR_015342, NR_109832, ENST00000412654, lnc-AC110080.1-5:1, ENST00000415820 and ENST00000558010 are promising research candidates for extra investigation. The present study of lncRNAs in PCa tissues is a proof-of-principle that lncRNAs have a possible character in PCa formation and progression. As demonstrated in the tables, there are so many lncRNAs has the relationship with PCa, the information, even, the name of their majority is blank. Lots of veri cation test are need to be completed. Our current study on the potential link between lncRNAs and PCa presents a novel analysis for further investigations into the biomarker and target genes of such lncRNAs, leading to clinical research for the disease.
The treatment paradigm of PCa has progressed rapidly in the last decade due to wider availability and choice of therapy. Thus, we have the reason to believe in, with the deep investigation, the potential mechanism of lncRNA will be disclosed stepwise, which provides new breakthroughs in the early diagnosis, prognosis, and therapy targets of PCa.

Conclusion
Our study detected a relatively novel complicated map of lncRNAs in PCa, which may have the potent to gure out a completely new and useful biomarkers or molecule for diagnosis, treatment and follow-up in PCa. As a candidate, we found that NR_125857 expression in human PCa tissues was up-regulated.
Further researches with numerous sample sizes of PCa tissues are needed to evaluate the relationships among the expression of NR_125857, clinicopathological features, and prognosis of PCa patients. In addition, more comprehensive in vivo study is essential.

Declarations
Availability of data and materials The datasets used and analyzed in the current study are available from the corresponding author on reasonable request. The diagram of data processing. In this study, lncRNA expression pro les were screened in PCa, by using ve pairs of clinical specimens in PCa and matched non-PCa tissues with lncRNA chip GPL22120. The abundance of each lncRNA against each miRNA was calculated using Spearman correlation, and then ltered by comparison with the theoretical databases. The theoretical databases included ENCORI, lncBase, miRcode for the relations of lncRNA-miRNA and miRcode, ENCORI, TarBase, miRTarBase, miRDB, miRanda, miRecords for the relations of miRNA-mRNA.

Figure 2
Global view of all lncRNAs expression in PCa tissues compared to paired non-PCa tissues. (A) Boxplot.
The medians between samples were roughly at, and the ranges of expressions were similar. (B) Sample correlation matrix. The correlation coe cient within the groups was signi cantly higher than that between the groups, which indicated the larger differences between PCa tissues and paired non-PCa    The co-expression network of lncRNA-miRNA-mRNA. In this gure, the red squares represented the miRNAs, the blue circle represented the lncRNAs and the green circles represented the mRNAs.