Interdisciplinary Program in Bioinformatics, Seoul National University, Seoul 151-742, Korea

Center for Biointelligence Technology (CBIT), Seoul National University, Seoul 151-742, Korea

School of Computer Science and Engineering, Seoul National University, Seoul 151-742, Korea

Abstract

Background

Dysregulation of genetic factors such as microRNAs (miRNAs) and mRNAs has been widely shown to be associated with cancer progression and development. In particular, miRNAs and mRNAs cooperate to affect biological processes, including tumorigenesis. The complexity of miRNA-mRNA interactions presents a major barrier to identifying their co-regulatory roles and functional effects. Thus, by computationally modeling these complex relationships, it may be possible to infer the gene interaction networks underlying complicated biological processes.

Results

We propose a data-driven, hypergraph structural method for constructing higher-order miRNA-mRNA interaction networks from cancer genomic profiles. The proposed model explicitly characterizes higher-order relationships among genetic factors, from which cooperative gene activities in biological processes may be identified. The proposed model is learned by iteration of structure and parameter learning. The structure learning efficiently constructs a hypergraph structure by generating putative hyperedges representing complex miRNA-mRNA modules. It adopts an evolutionary method based on information-theoretic criteria. In the parameter learning phase, the constructed hypergraph is refined by updating the hyperedge weights using the gradient descent method. From the model, we produce biologically relevant higher-order interaction networks showing the properties of primary and metastatic prostate cancer, as candidates of potential miRNA-mRNA regulatory circuits.

Conclusions

Our approach focuses on potential cancer-specific interactions reflecting higher-order relationships between miRNAs and mRNAs from expression profiles. The constructed miRNA-mRNA interaction networks show oncogenic or tumor suppression characteristics, which are known to be directly associated with prostate cancer progression. Therefore, the hypergraph-based model can assist hypothesis formulation for the molecular pathogenesis of cancer.

Background

Prostate cancer is a common disease in the male population, induced by complex interactions among various genetic factors

Recently, miRNAs have caused great excitement as diagnostic and therapeutic signatures of prostate cancer

Modern cancer research has progressed from identifying biomarkers to systemically exploring gene interactions

Here we introduce a data-driven model for identifying cancer stage-specific interactions that reflects the high-order relationships between miRNAs and mRNAs (Figure

Overview of the hypergraph-based model for constructing higher-order miRNA-mRNA interaction networks at a specific cancer stage

**Overview of the hypergraph-based model for constructing higher-order miRNA-mRNA interaction networks at a specific cancer stage.** Solid and dotted circles denote miRNAs and mRNAs, respectively. Closed curves denote hyperedges (i.e. modules). In the conventional graph representation (two graphs in the right-bottom of the central box of the figure), ellipses and boxes denote miRNAs and mRNAs, respectively. Grey and white indicate respective high and low gene expression levels.

The learning process involves the iteration of two learning phases; structure and parameter. The structure learning phase constructs a hypergraph of putative hyperedges for discovering potential gene interactions, from a huge feature space represented by the combinations of many miRNAs and mRNAs. Because the miRNA-mRNA interactions are intractably complex, we adopt an evolutionary strategy based on an information theoretic co-regulatory measure, called mutual information. This strategy is used to select genetic variables for generating hyperedges. During the parameter learning phase, the hypergraph is refined by updating the weights of the hyperedges (representing higher-order miRNA-mRNA modules). To this end, we employ a gradient descent method similar to the back-propagation algorithm for learning artificial neural networks. The learned model is then converted into a network structure reflecting the cooperative higher-order gene activities by connecting the extracted hyperedges. Data-driven learning allows the model to build new miRNA-mRNA interaction networks which display the hidden properties of primary and metastatic prostate cancers from a given dataset, which are not known

We construct cancer stage-specific miRNA-mRNA interaction networks reflecting their higher-order relationships using the MSKCC Prostate Oncogenome Project dataset

Results

Data and experimental settings

In this study, miRNA and mRNA expression profiles obtained from the MSKCC Prostate Oncogenome Project

**Parameters**

**Values**

**Parameters**

**Values**

# of miRNA

**3**

# of mRNA

**5**

# of modules

**variable**

**1.0**

Epochs of structure learning

**100**

Epochs of parameter learning

**20**

**1.0**

**1.0**

γ in (13)

**1.0**

_{max} , _{min}

**0.9, 0.5**

Classification performance

Classification performance was evaluated using three standard classification models; support vector machines (SVMs) with the 2nd polynomial kernel and sequential minimal optimization (SMO),

Boxplots of classification accuracy on the test set

**Boxplots of classification accuracy on the test set.**

Model evaluation

The proposed hypergraph-based learning method is evaluated on simulation data for verifying whether the method finds true solutions. The data consist of 500 instances with 7 variables whose mean is zero and the class label of each instance is determined as follows:

where _{
i
} and ^{(n)} denote the _{
2
}, _{
3
}, _{
4
}) and 2 (_{
5
}, _{
6
}, _{
7
}) in a learned hypergraph. Because we conducted 10-fold cross validation, the maximum values of Module 1 and 2 are ten. Therefore, we indicate that our method can find true solutions from small combinatorial spaces, considering the accuracy and the number of found variable modules.

**Models**

**SVM**

**DT**

**kNN**

**HG**

**Module 1**

**Module 2**

Accuracy

**0.956**

**0.886**

**0.93**

**0.956**

**10**

**10**

±SD

**±0.002**

**±0.004**

**±0.006**

**±0.003**

**-**

**-**

Figure

Learning curves in the structure and the parameter learning phases

**Learning curves in the structure and the parameter learning phases.** As the performance measure, we used mean multivariate mutual information (MMI) of all hyperedges in the model for the structure learning and accuracy on 10 fold cross validation for the parameter learning. **(a)** and **(b)**. All results are averaged on 10 experiments of 10- fold cross validation.

Moreover, Figure

Classification accuracy according to the number of miRNA and mRNA in the hyperedges

**Classification accuracy according to the number of miRNA and mRNA in the hyperedges.** The classification accuracy is the best when a hypergraph consists of hyperedges with three miRNAs and five mRNAs. All results are averaged on 10 experiments of 10- fold cross validation.

Table _{i}), for verifying the stability of the model as follows:

where _{
i
} denotes the _{
m
} is the _{
i
}, _{
m
}) is an indicator function and it returns one when _{
i
} appears at least once in _{
m
}, otherwise zero. The proposed method is compared to randomly generated hypergraphs each comprising 200 hyperedges involving three miRNAs and five mRNAs. The results are derived from 100 models learned by 10 experiments of 10-fold cross validations, and 100 randomly generated hypergraphs. According to Figure

**Our method**

**Random**

**Our method**

**Random**

Frequent

# of

Frequent

# of

Rare

# of

Rare

# of

miRNAs

appearances

miRNAs

appearances

miRNAs

appearances

miRNAs

appearances

miR-1

100/100

miR-152

97/100

miR-95

0/100

miR-30a

58/100

miR-100

100/100

miR-1

95/100

miR-937

0/100

miR-134

60/100

miR-133a

100/100

miR-486-5p

95/100

miR-933

0/100

miR-106a

60/100

miR-143

100/100

miR-199b-5p

94/100

miR-887

0/100

miR-362-5p

63/100

miR-145

100/100

miR-377

94/100

miR-744

0/100

miR-200b

63/100

**Our method**

**Random**

Frequent

# of

Frequent

# of

Frequent

# of

Frequent

# of

mRNAs

appearances

mRNAs

appearances

mRNAs

appearances

mRNAs

appearances

ACTA2

67/100

ILK

60/100

AIPL1

10/100

CACNA1D

9/100

SVIL

64/100

CSRP1

59/100

CBY3

10/100

CDC25C

9/100

ACTN1

63/100

TPM1

59/100

SHKBP1

10/100

DHRS7C

9/100

CAV1

63/100

FRMD6

58/100

ADCY5

9/100

FAT3

9/100

CCND2

60/100

LOC645954

58/100

C17orf58

9/100

FOXN3

9/100

Reproducibility of decisive miRNAs (a) and mRNAs (b) influencing on classification

**Reproducibility of decisive miRNAs (a) and mRNAs (b) influencing on classification.** 100 hypergraphs are generated by randomly selecting miRNAs and genes, while another 100 hypergraphs are generated by our learning method (10 experiments with 10-fold cross validation). Each hypergraph includes 200 hyperedges consisting of three miRNAs and five mRNAs. The x-axis denotes the rank of the appearance of miRNAs or mRNAs, and y-axis is the number of miRNA or mRNA appearances. Both axes are log-scaled.

Constructed higher-order miRNA-mRNA interaction networks in prostate cancer

The miRNA-mRNA interaction network constructed from the proposed model is illustrated in Figure

Constructed (a) primary prostate cancer-specific and (b) metastatic prostate cancer-specific miRNA-mRNA interaction networks

**Constructed (a) primary prostate cancer-specific and (b) metastatic prostate cancer-specific miRNA-mRNA interaction networks.** The primary-specific network includes 67 miRNAs and 233 mRNAs, while the metastatic network involves 65 miRNAs and 180 mRNAs. Both networks include 500 bi-relational edges which are selected based on their summed weight (among all edges converted from 20000 hyperedges of 100 hypergraphs). Up- and down-expressed miRNAs and genes are determined by the mean of each stage class. The red boxed miRNAs and genes have been reported to be associated with the particular stage of prostate cancer. The triangles, rectangles, diamonds and circles denote miRNAs, oncogenes or tumor suppressor genes, transcription factors, and other genes in the network, respectively.

Many of the miRNAs in the constructed networks have been significantly associated with prostate cancer in the literature, and are thus termed prostate cancer-related miRNAs

The miRNAs and mRNAs in the constructed networks are enriched in cancer-related genes with a significant

**The miRNAs and mRNAs in the constructed networks are enriched in cancer-related genes with a significant **
**
p
**

Interestingly, the enriched hyperedges, and the expression levels of the miRNAs and mRNAs, differ considerably between the primary and metastatic networks. Up- and down-expressed miRNAs and genes are determined by their means at each stage. The red boxed miRNAs and genes are known to be associated with the various stages of prostate cancer

Functional analysis of the constructed interaction networks

The constructed miRNA-mRNA interaction networks were validated by functional analyses based on a literature review and gene set analysis. As mentioned above, many of the miRNAs and mRNAs involved in the identified interactions are known indicators of prostate cancer

Especially, hsa-miR-143 and hsa-miR-145 play a crucial role in metastatic prostate cancer, and are recognized as a clinicopathological signature of prostate cancer

Moreover, hsa-miR-200c emerges as a distinct miRNA in the network of primary prostate cancer. According to several studies, hsa-miR-200c overexpression inhibits metastasis prostate cancer, while aberrant regulation triggers the invasion and migration of prostate cancer at the post-transcriptional level

Our model identified several transcription factors associated with prostate cancer metastasis, such as ETS2, HOXC4, STAT3, STAT5B, SOX4 and ZEB2. Among these, SOX4, STAT3 and STAT5B are known regulators of metastatic prostate cancer through the regulation of genes involved in miRNA processing, transcriptional regulation, and developmental pathways

Interactions involving hsa-miR-29b/MMP2 and hsa-miR-335/SOX4 appear concurrently in the constructed metastatic network (Table

**miRNAs [exp. levels: up (+), down (−)]**

**mRNAs [exp. levels: up (+), down (−)]**

miRNAs and their predicted targets are given in bold font. The underlined genes are the cancer genes archived in the Memorial Sloan-Kettering Cancer Center.

**Primary prostate cancer**

hsa-miR-330-3p(−)

**hsa-miR-133b**(+)

**hsa-miR-222**(−)

**
**

**WWC3**(−)

- CAV1

DHX35(−)

TSHZ3(−)

hsa-miR-143(+)

hsa-miR-502-5p(−)

**hsa-miR-548c-3p**(+)

**ZZEF1**(−)

**C20orf194**(−)

**
**

MBD3(+)

- GPR132

**hsa-miR-19a**(+)

**hsa-miR-133a**(+)

hsa-miR-153(+)

**
**

**WWC3**(−)

- PCBP4

TCEAL4(−)

CUL4A(+)

**hsa-miR-130a**(+)

**hsa-miR-375**(+)

**hsa-miR-19a**(+)

**
**

- CYLD

SNORA71D(+)

NDUFA6(−)

RGS9BP(−)

**hsa-miR-221**(−)

**hsa-miR-106b**(+)

**hsa-miR-222**(−)

**ARSJ**(−)

- SSPN

C3orf58(+)

PTGDS(−)

- RARB

**hsa-miR-130a**(+)

**hsa-miR-133a**(+)

**hsa-miR-19a**(+)

**VNN1**(−)

- FGF5

ELOVL7(+)

PHPT1(−)

- RND3

**hsa-miR-133a**(+)

**hsa-miR-222**(−)

**hsa-miR-130a**(+)

**C10orf137**(+)

FAM108C1(+)

- SCRIB

- PRKAR1A

MOXD1(−)

**hsa-miR-130a**(+)

**hsa-miR-149***(−)

**hsa-miR-26a**(+)

**
**

- TPM1

CRB2(−)

TMEM132A(+)

LIX1L(−)

**hsa-miR-133b**(+)

**hsa-miR-23b**(+)

**hsa-miR-106b**(+)

**PFAS**(+)

- UNC5C

HLF(−)

PSEN1(+)

- EZH2

hsa-miR-145(+)

hsa-miR-200c(+)

hsa-miR-23b(+)

TTC23(−)

PARM1 (−)

TOPORS(+)

NEBL(−)

RCAN2(−)

**Metastatic prostate cancer**

**hsa-miR-221**(−)

**hsa-miR-29b**(−)

**hsa-miR-143**(−)

**
**

**
**

**
**

**
**

SCN9A(+)

hsa-miR-29b(−)

**hsa-miR-335**(−)

**hsa-miR-143**(−)

**
**

**MPPED1**(+)

**
**

- HOXC4

SMTN(−)

hsa-miR-143(−)

**hsa-miR-22***(−)

hsa-miR-23b(−)

**
**

**
**

**PELO**(−)

**
**

TMEM150(+)

**hsa-miR-125b**(−)

hsa-miR-616(+)

**hsa-miR-143**(−)

**
**

**ERBB3**(+)

**ACAD8**(−)

**PHF15**(+)

TMEM16G(−)

hsa-miR-19a(−)

**hsa-miR-141**(+)

**hsa-miR-145**(−)

**PCDH20**(+)

**DNAJC3**(−)

- STAT3

ZNF385(+)

- ACTA2

hsa-miR-133b(−)

**hsa-miR-145**(−)

**hsa-miR-218**(−)

**IRF2**(−)

**
**

**
**

- RAB2B

- WFDC1

**hsa-miR-143**(−)

**hsa-miR-145**(−)

hsa-miR-222(−)

**
**

**MAPK7**(+)

MAP3K2(−)

- RAB34

S100A1(+)

**hsa-miR-143**(−)

**hsa-miR-145**(−)

**hsa-miR-214**(−)

**FEM1A**(+)

**
**

**NAGPA**(+)

C1orf142(+)

- ERAS

**hsa-miR-143**(−)

**hsa-miR-193b**(−)

**hsa-miR-145**(−)

**CLINT1**(−)

**
**

**MAPK7**(+)

RARRES2(−)

IL28A(+)

hsa-miR-221(−)

**hsa-miR-1**(−)

hsa-miR-133b(−)

**
**

**NDFIP2**(−)

**
**

**VPS28**(+)

**INPPd5E**(+)

To confirm the biological relevance of the constructed interaction networks, we analyzed the functional correlations among the network genes by canonical pathway analysis

**Canonical pathway analysis**

**
p
**

**Primary prostate cancer**

Pathways in cancer

1.70e-03

Rb1 pathway

5.95e-03

Retinoic acid pathway

6.61e-03

Aurora A pathway

7.44e-03

Beta-catenin degradation pathway

9.95e-03

Wnt/beta-catenin pathway

1.03e-02

Wnt canonical signaling pathway

1.34e-02

Met pathway (signaling of HGF receptor)

1.39e-02

P38-alpha/beta downstream pathway

1.52e-02

Beta-catenin nuclear pathway

1.58e-02

Aurora B pathway

1.66e-02

EPHB forward pathway

1.81e-02

IFN-gamma pathway

1.81e-02

P53 hypoxia pathway

1.97e-02

MYC repress pathway

2.15e-02

Progesterone mediated oocyte maturation

2.19e-02

Rac CycD pathway (Ras and Rho protein on G1/S transition)

2.73e-02

PLK1 pathway

2.88e-02

IL-6 (interleukin-6) pathway

3.08e-02

FGFR2C ligand binding and activation

3.58e-02

Cell cycle

4.43e-02

PDGFR-beta signaling pathway

4.59e-02

**Metastatic prostate cancer**

MYC activate pathway

1.41e-04

ErbB network pathway

2.78e-03

KIT receptor signaling pathway

3.28e-03

IL-10 pathway

4.40e-03

Pathways in cancer

4.76e-03

ErbB4 pathway

6.12e-03

Her2 pathway (ErbB2 in signal transduction and oncology)

8.51e-03

Yap1 and Wwtr1/Taz stimulated gene expression

1.09e-02

Smooth Muscle Contraction

1.22e-02

Barrestin pathway

1.53e-02

IL-6 signaling pathway

1.85e-02

STAT3 pathway

1.85e-02

IL-2/STAT5 pathway

2.00e-02

RAS pathway

2.00e-02

ErbB2/ErbB3 signaling pathway

2.19e-02

Syndecan4 pathway

2.38e-02

PPAR-alpha pathway

2.61e-02

Integrin signaling pathway

3.72e-02

Rela pathway

3.78e-02

HDAC class I pathway

3.94e-02

FOXM1 pathway

4.24e-02

IL-7 pathway

4.23e-02

EGFR pathway

4.70e-02

Discussion

The proposed hypergraph-based model characterizes higher-order interactions among heterogeneous genetic factors from archived data. Human cancers are typically caused by the modular control of multiple genetic factors. By analyzing gene relationships at higher-order levels, thus, we can better understand the behavior of complex cancer mechanisms. Moreover, the cooperative activities and the combinatorial regulations governed by miRNAs and mRNAs are largely unknown. We have demonstrated that higher-order relationships discriminate between specific cancer stages more precisely than pair-wise analyzes of single miRNA and mRNA interactions. From this viewpoint, we can construct a more complete interaction network consisting of putative biologically significant miRNA-mRNA modules.

In addition, our method focuses on discovering potential interactions in unknown miRNA-mRNA regulatory circuits related to specific cancer stages without the known biological information

The proposed hypergraph-based model is similar to Bonnet’s

Furthermore, the proposed model finds the true solution in a small subset of the features, because the problem space is small enough to search exhaustively. Also, unlike other models, our model can efficiently handle the very high-dimensional data required for complex higher-order interactions among features. However, the limitation of the proposed hypergraph-based model emerges at small sample sizes. If the data are few, the reliability of the mean and covariance defined in a hyperedge is reduced.

Conclusions

We have proposed a hypergraph-based model consisting of higher-order miRNA-mRNA modules, which allows the construction of biologically meaningful interaction networks associated with specific cancer stages. For identifying potential significant interactions and refining model performance, we introduced a two-phase learning approach comprising structure and parameter learning. Finally, we constructed cancer stage-specific interaction networks reflecting higher-order miRNA and mRNA relationships by converting the hypergraph structure into an ordinary graph.

We constructed higher-order miRNA-mRNA interaction networks associated with the specific stage of prostate cancer from a matched dataset using the proposed model. The performance of the proposed model is similar to that of SVMs and superior to other classification models (outperforming them by approximately 6–10%). More importantly, our model can construct carcinogenic miRNA-hubbed networks that characterize primary and metastatic prostate cancer. Furthermore, we demonstrated that a large proportion of the miRNAs and mRNAs identified in the constructed interaction networks are indeed involved in prostate cancer progression and development. The proposed hypergraph-based model therefore presents as an alternative method for discovering potential gene regulatory circuits. Such discoveries will greatly assist our understanding of cancer pathogenesis.

Methods

Hypergraph-based models

A hypergraph-based model characterizes complex interactions among many genetic factors using hypergraph structures. A hypergraph generalizes the edge concept to a hyperedge by which more than two variables can be connected simultaneously

A hypergraph-based model **x**
^{(n)} and **z**
^{(n)} are real-valued vectors of miRNA and mRNA expressions in the _{
i
} contains the mean vectors and the covariance of its miRNAs and mRNAs for the given cancer stage:

where **(x**, **z**), the cancer stage of the profile is classified as **(x**, **z**) matching the hyperedge), is highest among the elements of **(x**, **z**) matches _{
i|y
}” means that **(x**, **z**) has similar expression values to ones of the _{
i|y
} at cancer stage **(x**, **z**) and _{
i|y
}, **x**, **z**, _{
i|y
}). The matching probability is calculated by the normalized subdimensional distance between _{
i|y
} and **(x**, **z**):

where **(x**, **z**) matches _{
i|y
} , **x**
_{
ij
} and **z**
_{
ik
} (the **(x**, **z**) is computed as follows:

1. Calculate _{
y '}, the sum of the expected values for each

where |_{
i|y
}) is the weight of _{
i|y
}, explained in the next subsection.

2. Predict the cancer stage as

Biological meaning of mean and variance used in representing a hyperedge

**Biological meaning of mean and variance used in representing a hyperedge.** Panels **(a)** and **(b)** illustrate how the means and variances differ between low and high discriminative genetic factors. A gene is low-discriminative when the means are similar at each disease stage but the variances are large (where **(c)** illustrates the enhanced discriminative capability of a hyperedge involving two genetic factors. By comparing the discriminative capability of each miRNA or mRNA, the discrimination capability of the hyperedge is enhanced.

In terms of distance-based connectionist models, our model is related to radial basis function networks (RBFNs)

Learning hypergraph-based models

The proposed model learns by finding a hypergraph structure with high discriminative capability at a specific cancer stage. This is achieved by maximizing the conditional likelihood for a model _{
D,H
} , the log conditional likelihood is maximized by least mean square criteria using (7) and a sigmoidal function:

s.t.

where (^{(n)}, ^{(n)}) denotes the ^{(n)} is the cancer stage of the example.

To meet these requirements, the learning iterates two phases: structure learning and parameter learning. The structure learning constructs a hypergraph from hyperedges that identify potential miRNA-mRNA modules. The weights of the hyperedges are updated to minimize the classification error of the generated gene module population during the parameter learning phase. Because the hypergraph-based model represents a huge combinatorial feature space (size 2^{|x|+|z|}) of many miRNAs and mRNAs, exhaustively searching for the optimal population is infeasible. Instead we adopt an evolutionary learning method based on information-theoretic criteria to generate putative hyperedges for the structure learning.

We assume that a hyperedge consisting of strongly interactive miRNAs and mRNAs is highly discriminative for classification in this study. Mutual information is used as a co-regulatory measuring criterion for efficiently selecting genes for hyperedge generation. Mutual information (MI) is an information-theoretic measure that specifies the degree of conditional independency between two random variables. When a genetic factor more strongly determines the cancer stage, the MI between the gene and the cancer stage is increased. A hyperedge is generated by probabilistically selecting miRNAs and mRNAs, and the MI between each gene and the class label determines the probability of selecting the genes. The probability _{
I
}(_{
i
}) of selecting the _{
i
} is defined such that miRNAs or mRNAs with high MI are selected more frequently:

where _{
i
};

s.t.

where _{
i
} and

In the parameter learning phase, the weights of the hyperedges are updated using the gradient descent method for all training data. The aim is to minimize the error in terms of the classification probability in (9) and the matching probability in (5):

where

where _{
max
} and _{
min
} denote the maximum and minimum number of replaced hyperedges, respectively. Therefore, the number of replaced hyperedges consecutively decreases as the structure learning proceeds, while high-discriminative modules are preserved. The algorithm for learning the hypergraph-based model is presented in Figure

Algorithm for learning the hypergraph-based model

**Algorithm for learning the hypergraph-based model.**

Representing interaction networks from hypergraphs

We construct a higher-order miRNA-mRNA interaction network at a specific cancer stage from the learned model. When analyzing complex biological networks based on graph mining, frequently occurring subgraphs in the networks are generally regarded as important building blocks which are merged to create the functional network _{|y '} =(

and _{
i
} is a clique corresponding to the _{
i
} (Figure

Procedure of converting a hypergraph to cancer stage-specific interaction networks

**Procedure of converting a hypergraph to cancer stage-specific interaction networks.** ‘P’ and ‘M’ denote metastatic and primary prostate cancer, respectively.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

SJK proposed the idea and wrote the manuscript and analyzed the data. JWH implemented the method and performed the computational experiments. BTZ supervised the study and revised the manuscript. All authors read and approved the final manuscript.

Acknowledgements

This work was supported by the National Research Foundation (NRF) Grant funded by the Korea government (MSIP) (NRF-2010-0017734, NRF-2013M3B5A2035921, and the Bio & Medical Technology Development Program, No.2012M3A9D1054622), supported by KEIT grant funded by the Korea government (MKE) (KEIT-10035348 and KEIT- 10044009), supported by AOARD R&D grant funded by AFORS (124087).