Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Comprehensive analysis of microRNA expression patterns in hepatocellular carcinoma and non-tumorous tissues

Abstract

MicroRNAs (miRNAs) are a non-coding family of genes involved in post-transcriptional gene regulation. These transcripts are associated with cell proliferation, cell differentiation, cell death and carcinogenesis. We analysed the miRNA expression profiles in 25 pairs of hepatocellular carcinoma (HCC) and adjacent non-tumorous tissue (NT) and nine additional chronic hepatitis (CH) specimens using a human miRNA microarray. Targets and references samples were co-hybridized to a microarray containing whole human mature and precursor miRNA sequences. Whereas three miRNAs exhibited higher expression in the HCC samples than that in the NT samples, five miRNAs demonstrated lower expression in the HCC samples than in the NT samples (P<0.0001). Classification of samples as HCC or NT by using support vector machine algorithms based on these data provided an overall prediction accuracy of 97.8% (45/46). In addition, the expression levels of four miRNAs were inversely correlated with the degree of HCC differentiation (P<0.01). A comparison of CH and liver cirrhosis samples revealed significantly different pattern of miRNA expression (P<0.01). There were no differences, however, between hepatitis B-positive and hepatitis C-positive samples. This information may help clarify the molecular mechanisms involved in the progression of liver disease, potentially serving as a diagnostic tool of HCC.

Introduction

MicroRNAs (miRNAs), a non-coding RNA family, are 19- to 25-nt transcripts that are cleaved from 70- to 100-nt hairpin-shaped precursors. The sequences of many miRNAs are conserved between distantly related organisms, suggesting that these molecules participate in essential processes (Pasquinelli et al., 2000; Ke et al., 2003; Moss, 2003). Although the precise biological functions of miRNAs are not yet fully understood, they have diverse expression patterns and may regulate various developmental and physiological processes. Moreover, mis-regulation of miRNA expression might contribute to human disease (He and Hannon, 2004). Therefore, profiling miRNA expression patterns should help to identify the biological functions of miRNAs (Xu et al., 2003). The significance of non-coding RNAs in processes such as chromatin dynamics and gene silencing has received increased attention over the last few years, especially following the unmasking of the large group of small regulatory miRNAs (Bartel, 2004). Using mammalian microarray technology, comprehensive analysis of miRNA expression will help us better understand the relationship between aberrant gene expression and diseases (Liu et al., 2004; Liang et al., 2005).

The biological functions and targets of some miRNAs have been recently reported. miR-143 has been associated with adipocyte differentiation and the target of this miRNA is ERK5 (Esau et al., 2004). miR-375 regulates insulin secretion by modulating the expression of myotropin (Poy et al., 2004). Moreover, some reports have touched on the relationship between miRNA expression and carcinogenesis. For example, aberrant expression of the precursors of miR-155, and mature miR-15 and miR-16 were reported in malignant lymphomas (Calin et al., 2002; van den Berg et al., 2003). Furthermore, the downregulation of miR-143 and miR-145 was observed in colorectal neoplasia (Michael et al., 2003).

Hepatocellular carcinoma (HCC) is one of the most common malignant tumors. Major features of these carcinomas include chronic inflammation and the effects of cytokines on the development of fibrosis and liver cell proliferation. In addition, some viral genes can cause malignant transformation. Understanding the pathogenesis of viral-associated HCC is important in developing effective means of prevention and treatment of this highly malignant form of cancer (Brechot et al., 2000; Thorgeirsson and Grisham, 2002). It is also necessary to identify biological markers that can be used to screen high-risk patients in order to allow better HCC detection, earlier intervention and increase the likelihood of successful treatment. In this study, we investigate the miRNA expression profiles of various liver diseases by microarray analysis.

Results and discussion

Microarray construction and preparation

We initially tested the specificity and the sensitivity of this microarray-based approach. Plotting the un-normalized signal intensity of three miRNAs (miR-16, miR-22 and miR-29) against the quantity of the miRNA-rich fractionated RNA revealed a linear relationship from 10 to 1000 ng in the sample RNA (Figure 1a). To validate the reproducibility of this microarray method, we compared the miRNA distribution by using miRNA-rich RNA from Huh-7 cell line that had prepared at separate time. This result indicated that most miRNA included in twofold relative intensity line differences between two experiments (R2=0.995) (Figure 1b). Furthermore, we measured comparable fold changes in each miRNA within single dye-swap experiments, indicating a high reproducibility of the dye-swap arrays. To verify the accuracy of these data, Northern blot analysis on total RNA samples using two miRNA-specific probes was performed (Figure 1c).

Figure 1
figure1

Sensitivity and specificity of microRNA microarray. (a) Plot of the un-normalized signal intensity of three miRNAs against the amount of target RNA. (b) Reproducibility test of our microarray. Scatter plots comparing the normalized signal intensities of two independent experiments by using target RNA from Huh-7 cells. Red, yellow and green denotes high, moderate and low expression levels, respectively. (c) Validation of the microarray data with Northern blots. RNA levels of miR-16 and miR-122a from HeLa, Huh-7, HL60, HEK293 and HepG2 cells were analysed by the Northern blot. Numbers below the cell lines correspond to the normalized expression values of each miRNA measured using the microarray. The intensity of 5.8S rRNA stained with ethidium bromide was used as a loading control.

miRNA expression in HCC and adjacent non-tumor tissues

We analysed miRNA expression in 24 HCC samples, 22 adjacent non-tumor tissues (NT) tissue samples, a number of liver tumor-derived cell lines, nine chronic hepatitis samples derived from fine-needle biopsy and primary cultured hepatocytes (Table 1). We identified 30 miRNA genes that were significantly differentially expressed in the HCC and corresponding NT specimens (P<0.01) (Figure 2A). When we performed the comparison under stricter conditions, we uncovered seven mature and one precursor miRNAs that exhibited significantly differentially expression pattern between the HCC and NT samples (HCC>NT: miR-18, precursor miR-18, and miR-224, HCC<NT: miR-199a*, miR-195, miR-199a, miR-200a, and miR-125a) (P<0.001) (Figure 2B(a) and Table 2a). To confirm results obtained by microarray analysis, we carried out Northern blot analysis of seven mature and one precursor miRNAs that were expressed at significantly different level between tumor and non-tumor (P<0.001) (Figure 2B(a)). We prepared total RNA from one tumor (122T) and non-tumor pair (122N). The result of Northern blot analysis was a good validation to the result of microarray analysis (Figure 2B(b) and 2B(c).

Table 1 Clinical background of the patients
Figure 2
figure2figure2

(A) miRNA expression data from 24 HCC specimens, 22 NT specimens, primary cultured hepatocytes and two control cell lines (HeLa and Huh-7). The median values from four normalized log-ratio data sets were clustered hierarchically clustered in both dimensions and plotted in a heat map. Red denotes high expression levels, whereas green and gray depict low expression levels and no detectable signal, respectively. Data from each miRNA row were median centered. Dendrograms indicate the correlation between groups of tissues and genes. The expression values ranged from +5-fold to 0. (B) 1: miRNA array expression results for HCC patients (P<0.0001). The bars indicate the relative abundance of each miRNA (ratio of the signal intensities of the tumors to the signal intensities of the non-tumor tissues). (2 and 3): Comparison between microarray and Northern blot. B(b) shows Northern blot analysis of HCC (case 122T) and tumor adjacent tissue (122N) with probes miR-125a, miR-195, miR-199a, miR-199a*, miR-200a, miR-224 and miR-18. In the part of miR-18, the larger RNA bands that represented the precursor miRNA molecules (p18) were only shown for miR-18. The 5.8S ribosomal RNA was used for normalization of expression levels in the different lanes. B(c) shows signal intensity of seven mature and one precursor miRNAs in tissue of 122T and 122N from microarray. (C) miRNA expression data from five well-differentiated HCC samples, 14 moderately differentiated HCC samples and three poorly differentiated HCC samples (P<0.01). (D) Comparison of miRNA expression between 12 CH and 14 LC samples. Each miRNA listed is significantly differentially expressed (P<0.01) between the three histological groups.

Table 2a Comparison of the expression level of miRNA between hepatocellular carcinoma and non-tumor tissue adjacent tumor

The use of expression data to predict disease state is quickly becoming standard practice. Support vector machines (SVMs) techniques allow the use of training sets to uncover patterns that discriminate between classes. SVMs have been shown to perform well in multiple areas of biological analysis including evaluating microarray date (Brown et al., 2000). We attempted to use this algorithm to predict the classification of samples into cancer and non-cancer groups. With the exception of 85T, miRNA profiling allowed the accurate prediction of these groups with an overall cross-validation accuracy of 97.8% (45/46) by an SVM (Table 3). Although 85T did not conform to the prediction, there was no significant difference in the clinical features of another tumor samples. Owing to the severity of the disease, the lack of good diagnostic molecular markers and the absence of effective treatment strategies, HCC remains as a major medical challenge. Our measurement of miRNA levels accurately diagnosed the malignancy based on the gene expression profiling, indicating that such a method may be a powerful tool for diagnosing and treating cancer. Whereas a number of miRNAs were upregulated or unchanged, the majority of examined miRNAs exhibited lower expression levels in tumors than those seen in normal tissues, irrespective of cell type (Lu et al., 2005). Correlating with this observation, we determined that the miRNA expression pattern in primary hepatocytes was different than that seen in both the tumor and the non-tumor groups.

Table 3 Summary of prediction between tumor and non-tumor by support vector machine in detail

Several algorithms have been used to address the participation of these specific miRNAs in the development of cancer through the dysregulation of oncogene (John et al., 2004; Kiriakidou et al., 2004). Using these algorithms, we listed cancer-related mRNA targets of miRNA (Table 4). Especially, searching a candidate gene for miR-18, we picked up two candidates; connective tissue growth factor (CTGF) and receptor activator nuclear factor kappa B ligand (RANKL). Generally, higher expression of miRNA should downregulate a target gene post transcriptionally. We prepared two pairs of tumor and non-tumor portion (70T and 70N, and 87T and 87N) of which miR-18 expression has good correlation of our microarray analysis (Figures 2B and 3). Gene expressions of both CTGF and RANKL in tumor portion were lower than in non-tumor portion. Therefore, these two genes are likely to be a target of miR-18. Moreover, in our preliminary functional analysis of miRNA, overexpression of miR-199a* can introduce cell cycle arrest in G2/M stage (date not shown). Recently it is reported that one cancer-related gene candidate that was regulated by a miRNA was RAS. let-7 family members have been mapped to human chromosomal sites that are implicated in a variety of cancers. In particular, let-7a-2, let-7c and let-7g are located in small chromosomal intervals that are deleted in lung cancers (Takamizawa et al., 2004), a malignancy in which RAS mis-regulation is known to be a key oncogenic event (Johnson et al., 2005). Although the exact mechanism by which overproduction of mir-17-19b might promote oncogenesis is unclear, recent studies of tumor pathology have suggested that increased expression of the mir-17-19 cluster mitigates the proapoptotic response that occurred in response to elevated myc expression levels in vivo. miRNA itself can be oncogene (He et al., 2005).

Table 4 Hypothetical target of several miRNA that have different expression between HCC and adjacent non-tumor
Figure 3
figure3

Validation of candidate target genes of miR-18 by Western blot analysis: (a) Western blot analysis of whole lysate from two pairs of tumorous tissue (70T and 87T) and NT (70N and 87N) (see Table 1) by using the anti-CTGF, RANKL and tubulin antibodies was shown. (b) Signal intensity of miR-18 expression of these two tumor and non-tumor pairs was shown by bar.

Our microarray included both mature miRNAs and their corresponding precursor miRNAs, allowing us to detect independently the expression levels of mature active and precursor miRNAs during tumor biogenesis. Whereas primary transcript profiling offers some benefits, such as the ability to study the regulation of miRNA transcript processing, it does not provide an exact active, mature miRNA expression profile. As miRNA primary transcript processing and RNA-induced silencing complex assembly occur in several steps, detection of similar levels of primary transcripts may be misleading. Instead, the levels of the mature, active miRNA should be the most relevant indicator of biological function (Liu et al., 2004).

Gene expression profiles of various liver diseases and their clinical features

In addition, our analysis of a small number of HCC samples compared miRNA expression in tumors differing in differentiation state (well, moderately and poorly differentiated HCC). The expression levels of four miRNAs (miR-92, miR-20, miR-18 and precursor miR-18) were significantly higher in poorly differentiated HCC samples, with moderate expression in moderately differentiated HCC and low expression in well-differentiated HCC (P<0.01) (Figure 2C and Table 2b). In contrast, miR-99a expression exhibited a positive correlation between expression levels and the degree of tumor differentiation. The expression levels of miR-18, miR-20 and precursor miR-18 were high in the tumor samples in comparison to the non-tumor samples. The degree of tumor differentiation was inversely related to the expression levels of these three miRNAs, suggesting that these miRNAs contribute to both tumorigenesis and the loss of tumor differentiation.

Table 2b Comparison of the expression level of miRNA among tumor differentiation

To identify genes that were abnormally expressed in cases of both chronic hepatitis (CH) and liver cirrhosis (LC), we examined the gene expression profiles of 12 CH and 14 LC samples. The expression levels of 12 miRNAs (CH>LC: miR-182, precursor miR-199b, miR-224 and miR-15b; CH<LC: miR-28, miR-342, miR-126, miR-199a, miR-145b, miR-143, miR-368 and precursor miR-372) were significantly different between these two groups (P<0.01) (Figure 2D and Table 2c). Of the 10 miRNAs and two precursor miRNAs identified during the comparison of CH with LC, a number of these miRNAs are likely to target genes that regulate inflammation and fibrosis.

Table 2c Comparison of the expression level of miRNA between chronic hepatitis and liver cirrhosis

HCC samples were also obtained from patients with hepatitis B (HBV; n=6) or hepatitis C (HCV; n=17) infections. The miRNA expression profiles of tumors with HBV infection were not significantly different from the miRNA expression profiles of tumors with HCV infection (data not shown).

As microarray technology provides us with a more specific and sensitive analysis of miRNA expression, we can examine small sample amounts obtained from fine-needle biopsy. The combination of the SVM and our array may allow a more accurate method for the early detection of HCC. The use of our microarray may enable a variety of clinical applications. A bioinformatic study has now investigated that there are approximately 321 miRNAs in humans in the miRNA registry (www.sanger.ac.uk/Software/Rfam/mirna). This miRNA profiling microarray could easily be expanded to profile miRNAs in a variety of clinical situations and may facilitate the analysis of other non-coding RNA families from several species. The use of this universal microarray for trans-species miRNA expression profiling for each known miRNA under various conditions will make miRNome deciphering more efficient and contribute to miRNA target identification, miRNA expression regulation and studies of pathological diseases.

Materials and methods

Samples

Twenty-five pairs of samples were surgically resected from HCC and adjacent NT tissue. An additional nine samples of CH type C were obtained by fine-needle biopsy (Table 1). All of the patients or their guardians provided written informed consent, and the Ethics Committee from the Kyoto University Graduate School and Faculty of Medicine approved all aspects of this study.

miRNA oligo probe design

Probes corresponding to 180 mature human miRNAs and 206 precursor human miRNAs found in the miRNA registry were synthesized. tRNAs were also printed on the microchip, providing an internal, relatively stable positive control for specific hybridization. The oligonucleotide sequences and the arrangement of the probes are depicted in Supplementary Table 1a and b and Figure 1, respectively.

miRNA microarray Fabrication

Sixty-mer 5′-amine modified oligonucleotides were resuspended in Spotting solution 6 (Toyo Kohan, Tokyo, Japan) at 100 pmol/μl. Individual oligonucleotides were printed in duplicate on Geneslides (Toyo Kohan) by an OmniGrid 100 microarrayer (Genomic Solution, Ann Arbor, MI, USA) in the 1 × 4 pin configuration and the 12 × 20 spot configuration for each subarray. After spotting the oligonucleotides, the DNA was fixed overnight at 80°C, washed with 2 × SSC-0.1% SDS, and then incubated at 95°C with 2 × SSC-0.1% SDS to prevent from non-specific hybridization.

Target preparation and array hybridization

miRNA-rich fractionated RNA (Ambion, Austin, TX, USA) was prepared from tissue samples or cell lines following manufacturer's instructions. miRNA-rich RNA from five cell lines (HeLa, HL60, HepG2, HEK293 and Saos-2) was mixed and used as a reference. An Ulysis labeling kit (Invitrogen, Carlsbad, CA, USA) was used to label 1 μg of target RNA with Alexa Flour 647 and same amount of reference RNA with Alexa Flour 546 (Th Tsangaris et al., 2002). To determine validity of linear RNA amplification, target and reference RNA were prepared using this alternative labeling method and these probes were again hybridized to our custom arrays in a dye-swap experiment. The microarray was hybridized to 100 ng of labeled RNA in 2 × hybridization buffer (Agilent Technologies, Palo Alto, CA, USA) at 55°C for 16 h. Bound RNA was washed first with 6 × SSC-0.005% Triton X-100 at room temperature then with 0.1 × SSC-0.005% Triton X-100 at 4°C for 5 min. Arrays were scanned and imaged on an Agilent Dual-Laser DNA Microarray Scanner with 100% of Red and Green PMT power (G2565BA, Agilent Technologies).

Data analysis

Images were quantified with Feature Extraction Software (Agilent Technologies). Signal intensities for each spot were calculated by subtracting the local background (the median intensity of the area surrounding each spot) from the total intensities. For each miRNA, the perfect match signal (PM, the signal intensity from a probe corresponding to the wild-type sequence) and the mismatch signal (MM, the signal intensity from a probe with a mutated sequence) were measured. True signal intensity was calculated by subtracting the MM from the PM. If this value was negative (PM<MM), the miRNA signal intensity was estimated to be 0.01. Raw data were normalized and analysed using the GeneSpring software (Version 6.1.1, Silicon Genetics). GeneSpring generated an average value from the three spot replicates of each miRNA. After data transformation to convert any negative values to 0.01, data normalization was performed using positive control RNA spots (tRNA(G), tRNA(L), tRNA(T), tRNA(H) and 5S rRNA) to allow comparisons among chips.

Northern blot analysis

Total RNA was extracted from tissues by using the Sepasol Reagent (Nacalai tesque, Kyoto, Japan) Total RNA (20 μg) was separated on a 15% denaturing polyacrylamide gel containing 8 M urea. Loadings were visualized by ethidium bromide staining. The RNA was then transferred to Hybond N nylon membrane (Amarsham, Backinghamshire, UK) by semidry blotting (OWL Separation Systems, Portsmouth, NH, USA), and then the filter was fixed by baking at 80°C for 2 h. Probe was generated by T4 Polynucleotide Kinase (New England Biolabs, Beverly, MA, USA)-mediated end-labeling of DNA oligonucleotides complementary to the mature miRNA with [α-32P]ATP. Each probe sequences was followed (miR-16: 5′-IndexTermcgccaatatttacgtgctgcta-3′, miR-18: 5′-IndexTermtaaggtgcatctagtgcagata-3′, miR-122a: 5′-IndexTermacaaacaccattgtcacactcca-3′, miR-125: 5′-IndexTermcacaggttaaagggtctcaggga-3′, miR-195: 5′-IndexTermgccaatatttctgtgctgcta-3′, miR-199a: 5′-IndexTermgaacaggtagtctgaacactggg-3′, miR-199a*: 5′-IndexTermaaccaatgtgcagactactgta-3′, miR-200a: 5′-IndexTermacatcgttaccagacagtgtta-3′, miR-224: 5′-IndexTermtaaacggaaccactagtgacttg-3′). Filter was hybridized in Ultra-hyb oligo (Ambion, Austin, TX, USA) at 39°C for 16 h and then washed with 2 × SSC-0.5% SDS following manufacturer's instructions.

Western blot analysis

Tumor and non-tumor tissues were homogenized in RIPA buffer (150 mM Sodium chloride, 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 50 mM Tris-HCl (pH 8.0)) and protease inhibitor CompleteMini (Roche-Boehringer) was included. After the debris was removed, supernatants were boiled and mixed with an equal volume of 20% glycerol containing 0.02% bromophenol blue. Proteins were separated by SDS–10% polyacrylamide gel electrophoresis (PAGE) or SDS–7% PAGE and transferred to a polyvinylidene difluoride membrane (Millipore). The membranes were blocked with 5% skim milk in TBST (10 mM Tris (pH 7.5), 100 mM NaCl and 0.1% Tween 20) and incubated with primary antibodies in TBST with 0.5% skim milk overnight at 4°C. The membrane was treated with anti-CTGF (1:500, sc-14939; Santa Cruz Biotechnology), RANKL (1:100, sc-9073) or tubulin antibodies (1:2000, Oncogene) as primary antibodies and horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse immunoglobulin G antibody (1:3000) (Amersham Biosciences) as the secondary antibody, and immunoreactive bands were visualized by Western lightning™ (Perkin Elmer, Boston, MA, USA).

References

  1. Bartel DP . (2004). Cell 116: 281–297.

  2. Brechot C, Gozuacik D, Murakami Y, Paterlini-Brechot P . (2000). Semin Cancer Biol 10: 211–231.

  3. Brown MP, Grundy WN, Lin D, Cristianini N, Sugnet CW, Furey TS et al. (2000). Proc Natl Acad Sci USA 97: 262–267.

  4. Calin GA, Dumitru CD, Shimizu M, Bichi R, Zupo S, Noch E et al. (2002). Proc Natl Acad Sci USA 99: 15524–15529.

  5. Esau C, Kang X, Peralta E, Hanson E, Marcusson EG, Ravichandran LV et al. (2004). J Biol Chem 279: 52361–52365.

  6. He L, Hannon GJ . (2004). Nat Rev Genet 5: 522–531.

  7. He L, Thomson JM, Hemann MT, Hernando-Monge E, Mu D, Goodson S et al. (2005). Nature 435: 828–833.

  8. John B, Enright AJ, Aravin A, Tuschl T, Sander C, Marks DS . (2004). PLoS Biol 2: e363.

  9. Johnson SM, Grosshans H, Shingara J, Byrom M, Jarvis R, Cheng A et al. (2005). Cell 120: 635–647.

  10. Ke XS, Liu CM, Liu DP, Liang CC . (2003). Curr Opin Chem Biol 7: 516–523.

  11. Kiriakidou M, Nelson PT, Kouranov A, Fitziev P, Bouyioukos C, Mourelatos Z et al. (2004). Genes Dev 18: 1165–1178.

  12. Liang RQ, Li W, Li Y, Tan CY, Li JX, Jin YX et al. (2005). Nucleic Acids Res 33: e17.

  13. Liu CG, Calin GA, Meloon B, Gamliel N, Sevignani C, Ferracin M et al. (2004). Proc Natl Acad Sci USA 101: 9740–9744.

  14. Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D et al. (2005). Nature 435: 834–838.

  15. Michael MZ, O'Connor SM, van Holst Pellekaan NG, Young GP, James RJ . (2003). Mol Cancer Res 1: 882–891.

  16. Moss EG . (2003). Trends Biotechnol 21: 185–187.

  17. Pasquinelli AE, Reinhart BJ, Slack F, Martindale MQ, Kuroda MI, Maller B et al. (2000). Nature 408: 86–89.

  18. Poy MN, Eliasson L, Krutzfeldt J, Kuwajima S, Ma X, Macdonald PE et al. (2004). Nature 432: 226–230.

  19. Takamizawa J, Konishi H, Yanagisawa K, Tomida S, Osada H, Endoh H et al. (2004). Cancer Res 64: 3753–3756.

  20. Th Tsangaris G, Botsonis A, Politis I, Tzortzatou-Stathopoulou F . (2002). Toxicology 178: 135–160.

  21. Thorgeirsson SS, Grisham JW . (2002). Nat Genet 31: 339–346.

  22. van den Berg A, Kroesen BJ, Kooistra K, de Jong D, Briggs J, Blokzijl T et al. (2003). Genes Chromosomes Cancer 37: 20–28.

  23. Xu P, Vernooy SY, Guo M, Hay BA . (2003). Curr Biol 13: 790–795.

Download references

Acknowledgements

We thank Itsuro Inoue from the Division of Genetic Diagnosis at the Institute of Medical Science, University of Tokyo, for his helpful advice and comments.

Author information

Affiliations

Authors

Corresponding author

Correspondence to K Shimotohno.

Additional information

Supplementary information accompanies the paper on Oncogene website (http://www.nature.com/onc)

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Murakami, Y., Yasuda, T., Saigo, K. et al. Comprehensive analysis of microRNA expression patterns in hepatocellular carcinoma and non-tumorous tissues. Oncogene 25, 2537–2545 (2006). https://doi.org/10.1038/sj.onc.1209283

Download citation

Keywords

  • hepatocellular carcinoma
  • micro-RNA
  • support vector machine

Further reading

Search

Quick links