Post on 23-Mar-2020
Hindawi Publishing CorporationAdvances in BioinformaticsVolume 2013 Article ID 360678 11 pageshttpdxdoiorg1011552013360678
Research ArticleGene Regulation Modulation and Their Applications inGene Expression Data Analysis
Mario Flores1 Tzu-Hung Hsiao2 Yu-Chiao Chiu3 Eric Y Chuang3
Yufei Huang1 and Yidong Chen24
1 Department of Electrical and Computer Engineering University of Texas at San Antonio San Antonio TX 78249 USA2Greehey Childrenrsquos Cancer Research Institute University of Texas Health Science Center at San Antonio San Antonio TX 78229 USA3Graduate Institute of Biomedical Electronics and Bioinformatics National Taiwan University Taipei Taiwan4Department of Epidemiology and Biostatistics University of Texas Health Science Center at San Antonio San AntonioTX 78229 USA
Correspondence should be addressed to Yufei Huang yufeihuangutsaedu and Yidong Chen cheny8uthscsaedu
Received 2 December 2012 Accepted 24 January 2013
Academic Editor Mohamed Nounou
Copyright copy 2013 Mario Flores et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Commonmicroarray and next-generation sequencing data analysis concentrate on tumor subtype classification marker detectionand transcriptional regulation discovery during biological processes by exploring the correlated gene expression patterns and theirshared functions Genetic regulatory network (GRN) based approaches have been employed in many large studies in order toscrutinize for dysregulation and potential treatment controls In addition to gene regulation and network construction the conceptof the network modulator that has significant systemic impact has been proposed and detection algorithms have been developedin past years Here we provide a unified mathematic description of these methods followed with a brief survey of these modulatoridentification algorithms As an early attempt to extend the concept to new RNA regulation mechanism competitive endogenousRNA (ceRNA) into a modulator framework we provide two applications to illustrate the network construction modulation effectand the preliminary finding from these networks Those methods we surveyed and developed are used to dissect the regulatednetwork under different modulators Not limit to these the concept of ldquomodulationrdquo can adapt to various biological mechanismsto discover the novel gene regulation mechanisms
1 Introduction
With the development of microarray [1] and lately the nextgeneration sequencing techniques [2] transcriptional pro-filing of biological samples such as tumor samples [3ndash5]and samples from other model organisms have been carriedout in order to study sample subtypes at molecular level ortranscriptional regulation during the biological processes [6ndash8] While common data analysis methods employ hierarchi-cal clustering algorithms or pattern classification to explorecorrelated genes and their functions the genetic regulatorynetwork (GRN) approaches were employed to scrutinize fordysregulation between different tumor groups or biologicalprocesses (see reviews [9ndash12])
To construct the network most of research is focused onmethods based on gene expression data derived from high-throughput technologies by using metrics such as Pearsonor Spearman correlation [13] mutual information [14] co-determinationmethod [15 16] Bayesianmethods [17 18] andprobabilistic Boolean networks [19] Recently new transcrip-tional regulation via competitive endogenous RNA (ceRNAs)has been proposed [20 21] introducing additional dimensionin modeling gene regulation This type of regulation requiresthe knowledge of microRNA (miRNA) binding targets [2223] and the hypothesis of RNA regulations via competitionof miRNA binding Common GRN construction tries toconfine regulators to be transcription factor (TF) proteinsa primary transcription programming machine which relies
2 Advances in Bioinformatics
Interact
Regulator Target
119909 119910
(a)
Regulator
119909
119910
119911
119908
of 119910 and 119908
Regulator of 119911
Target of 119910
Target of 119909
Target of 119909
(b)
Figure 1 Regulator-target pair in genetic regulatory network model (a) basic regulator-target pair and (b) regulator-target complex
on sequence-specific binding sites at target genesrsquo promoterregions In contrast ceRNAs mediate gene regulation viacompeting miRNAs binding sites in target 31015840UTR regionwhich exist in gt50 of mRNAs [22 24] In this study wewill extend the current network construction methods byincorporating regulation via ceRNAs
In tumorigenesis gene mutation is the main cause ofthe cancer [25] The mutation may not directly reflect inthe change at the gene expression level however it willdisrupt gene regulation [26ndash28] In Hudson et al theyfound that mutated myostatin and MYL2 showed differentcoexpressionswhen comparing towild-typemyostatin Chunet al also showed that oncogenic KRAS modulates HIF-1120572 and HIF-2120572 target genes and in turn modulates cancermetabolism Stelniec-Klotz et al presented a complex hierar-chicalmodel ofKRASmodulated network followed by doubleperturbation experiments Shen et al [29] showed a temporalchange of GRNs modulated after the estradiol stimulationindicating important role of estrogen in modulating GRNsFunctionally modulation effect of high expression of ESR1was also reported byWilson andDering [30]where they stud-ied previously published microarray data with cells treatedwith hormone receptor agonists and antagonists [31ndash33] Inthis study a comprehensive review of existing algorithms touncover themodulators was provided Given either mutationor protein expression status was unknown under many ofreported studies the problem of how to partition the diversesamples with different conditions such as active or inactiveoncogene status (and perhaps a combination of multiplemutations) and the prediction of a putative modulator ofgene regulation remains a difficult task
By combining gene regulation obtained from coexpres-sion data and ceRNAs we report here an early attempt tounify two systems mathematically while assuming a knownmodulator estrogen receptor (ER) By employing the TCGA[3] breast tumor gene expressions data and their clinical testresult (ER status) we demonstrate the approach of obtainingGRN via ceRNAs and a new presentation of ER modulationeffects By integrating breast cancer data into our uniqueceRNAs discovery website we are uniquely positioned tofurther explore the ceRNA regulation network and further
develop the discovery algorithms in order to detect potentialmodulators of regulatory interactions
2 Models of Gene Regulation and Modulation
21 Regulation of Gene Expression Thecomplex relationshipsamong genes and their products in a cellular system canbe studied using genetic regulatory networks (GRNs) Thenetworksmodel the different states or phenotypes of a cellularsystem In this model the interactions are commonly mod-eled as regulator-target pairs with edges between regulatorand target pair representing their interaction direction asshown in Figure 1(a) In this model a target gene is a genewhose expression can be altered (activated or suppressed)by a regulator gene This definition of a target gene impliesthat any gene can be at some point a target gene or adirect or indirect regulator depending on its position in thegenetic regulatory network The regulator gene is a gene thatcontrols (activates or suppresses) its target genesrsquo expressionThe consequences of these activated (or suppressed) genessometimes are involved in specific biological functions suchas cell proliferation in cancer Examples of regulator-targetpair in biology are common For example a target geneCDCA7 (cell division cycle-associated protein 7) is a c-Myc(regulator) responsive gene and it is part of c-Myc-mediatedtransformation of lymphoblastoid cells Furthermore asshown in Figure 1(b) a regulator gene can also act as a targetgene if there exists an upstream regulator
If the interaction is modeled after Boolean network (BN)model [34] then
119910119894(119905 + 1) = 119891
119894(1199091198951(119905) 119909
119895119896(119905) 119910119894(119905)) (1)
where each regulator 119909119895isin 0 1 is a binary variable as well
as it is target 119910119894 As described by (1) the target 119910
119894at time
119905 + 1 is completely determined by the values of its regulatorsat time 119905 by means of a Boolean function 119891
119894isin 119865 where
119865 is a collection of Boolean functions Thus the Booleannetwork 119866(119881 119865) is defined as a set of nodes (genes) 119881 =1199091 1199092 119909
119899 and a list of functions (edges or interactions)
119865 = 1198911 1198912 119891
119899 Similarly such relationship can be
defined in the framework of Bayesian network where the
Advances in Bioinformatics 3
R
Target gene
TF
119910
(a)
Singaltransductionproteins
ReceptorHormone
(b)
microRNAs
3998400UTR
Target gene 119910
(c)
Figure 2 Three different cases of regulation of gene expression that share the network representation of a regulator target interaction
similar regulators-target relationship as defined in (1) can bemodeled by the distribution
119875 (119910119894(119905 + 1) 119909
1198951(119905) 119909
119895119896(119905) 119910119894(119905))
= 119875 (119910119894(119905 + 1) | Parents (119910
119894(119905 + 1)))
times 119875 (Parents (119910119894(119905 + 1)))
(2)
where Parents(119910119894(119905 + 1)) = 119909
1198951
(119905) 119909119895119896(119905) 119910119894(119905) is the set
of regulators or parents of 119910119894 119875(119910119894(119905 + 1) | Parents (119910
119894(119905 +
1))) is the conditional distribution defining the regulator-target relationship and 119875(Parents(119910
119894(119905 + 1))) models the
prior distribution of regulators Unlike in (1) the target andregulators in (2) are modeled as random variables Despiteof this difference in both (1) and (2) the target is always afunction (or conditional distribution) of the regulator (or par-ents)When the relationship is defined by a Boolean functionas in (1) the conditional distribution in (2) take the formof a binomial distribution (or a multinomial distributionwhen both regulators and target take more than two states)Other distributions such as the Gaussian and Poisson can beintroduced to model more complex relationships than theBoolean The network construction inference and controlhowever are beyond the scope of this paper and we leave thetopics to the literatures [9 35 36]
The interactions among genes and their products ina complex cellular process of gene expression are diversegoverned by the central dogma of molecular biology [37]There are different regulation mechanisms that can actuateduring different stages Figure 2 shows three different casesof regulation of gene expression Figure 2(a) shows the caseof regulation of expression in which a transcription factor(TF) regulates the expression of a protein-coding gene (indark grey) by binding to the promoter region of target gene 119910Figure 2(b) is the case of regulation at the protein level inwhich a ligand protein interacts with a receptor to activaterelay molecules to transduce outside signals directly into cellbehavior Figure 2(c) is the case of regulation at the RNAlevel in which one or more miRNAs regulate target mRNA119910 by translational repression or target transcript degradationvia binding to sequence-specific binding sites (called miRNAresponse elements or MREs) in 31015840UTR region As illustratedin Figure 2(c) the target genesproteins all contain a domainof binding or docking site enabling specific interactions
Modulator
Interact
Regulator Target
119909 119910
119898
Figure 3 Graphical representation of the triplet interaction ofregulator 119909 target 119910 and modulator119898
between regulator-target pairs a common element in net-work structure
22 Modulation of Gene Regulation Different from the con-cept of a coregulator commonly referred in the regulatorybiology a modulator denotes a gene or protein that is capableof altering the endogenous gene expression at one stage ortime In the context of this paper we specifically definea modulator to be a gene that can systemically influencethe interaction of regulator-target pair either to activateor suppress the interaction in the presenceabsence of themodulator One example of modulator is the widely studiedestrogen receptor (ER) in breast cancer studies [38ndash40] theER status determines not only the tumor progression but alsothe chemotherapy treatment outcomes It is well known thatbinding of estrogen to receptor facilitates the ER activitiesto activate or repress gene expression [41] thus effectivelymodulating the GRN Figure 3 illustrates the model of theinteraction between a modulator (119898) and a regulator (119909)target (119910) pair that it modulates
Following the convention used in (1) and (2) the modu-lation interaction in Figure 3 can be modeled by
119910 = F(119898)
(119909) (3)
where 119910 represents target expression 119909 represents the parents(regulators) of target 119910 andF(119898)(sdot) is the regulation function
4 Advances in Bioinformatics
modulated by119898 WhenF(119898)(sdot) is stochastic the relationshipis modeled by the conditional distribution as
119901 (119910 119909 | 119898) = 119901 (119910 | 119909119898) 119901 (119909 | 119898) (4)
where 119901(119910|119909119898) models the regulator-target relationshipmodulated by 119898 and 119901(119909|119898) defines the prior distributionof regulators (parents) expression modulated by119898 Differentdistribution models can be used to model different mech-anisms for modulation At the biological level there aredifferent mechanisms for modulation of the interaction 119909-119910and currently several algorithms for prediction of the mod-ulators has been developed This survey presents the latestformulations and algorithms for prediction of modulators
3 Survey of Algorithms of Gene Regulationand Modulation Discovery
During the past years many computational tools have beendeveloped for regulation network construction and thendepending on the hypothesis modulator concept can betested and extracted Here we will focus on modulator detec-tion algorithms (MINDy Mimosa GEM and Hermes) Tointroduce gene-gene interaction concept we will also brieflydiscuss algorithms for regulation network construction(ARACNE) and ceRNA identification algorithm (MuTaMe)
31 ARACNE (Algorithm for the Reconstruction of AccurateCellular Networks) ARACNE [14 42] is an algorithm thatextracts transcriptional networks from microarray data byusing an information-theoreticmethod to reduce the indirectinteractions ARACNE assumes that it is sufficient to estimate2-way marginal distributions when sample size119872 gt 100 ingenomics problems such that
119901 (119909119894) =
1
119911
119890minus[sum119873
119894=1120601119894(119909119894)+sum119873
119894119895120601119894119895(119909119894119910119895)] (5)
Or a candidate interaction can be identified using estima-tion of mutual information MI of genes 119909 and 119910 MI(119909 119910) =MI119909119910 where MI
119909119910= 1 if genes 119909 and 119910 are identical and
MI119909119910
is zero if 119901(119909 119910) = 119901(119909)119901(119910) or 119909 and 119910 are sta-tistically independent Specifically the estimation of mutualinformation of gene expressions 119909 and 119910 of regulator andtarget genes is done by using the Gaussian kernel estimatorTheARACNE takes additional two steps to clean the network(1) removing MI if its 119875 value is less than that derived fromtwo independent genes via random permutation and (2) dataprocessing inequality (DPI) The algorithm further assumesthat for a triplet gene (119892
119909 119892119910 119892119911) where 119892
119909regulates 119892
119911
through 119892119910 then
MI119909119911lt min (MI
119909119910MI119910119911) if 119909 997888rarr 119910 997888rarr 119911
with no alternative path(6)
where rarr represents regulation relationship In other wordsthe lowest mutual information MI
119909119911is from an indirect
interaction and thus shall be removed from the GRN by
ARACNE in the DPI step A similar algorithm was proposed[43] to utilize conditional mutual information to exploremore than 2 regulators
32 MINDy (Modulator Inference by Network Dynamics)Similar toARACNEMINDy is also an information-theoreticalgorithm [44] However MINDy aims to identify poten-tial transcription factor-(TF-target) gene pairs that can bemodulated by a candidate modulator MINDy assumes thatthe expressions of the modulated TF-target pairs are ofdifferent correlations under different expression state of themodulator For simplicity and computational considerationMINDy considers only two modulator expression states thatis up- (119898 = 1) or down-expression (119898 = 0) Then ittests if the expression correlations of potential TF-target pairsare significantly different for modulator up-expression versusdown-expression The modulator dependent correlation isassessed by the conditional mutual information (CMI) or119868(119909 119910 | 119898 = 0) and 119868(119909 119910 | 119898 = 1) Similar to ARACNEthe CMI is calculated using the Gaussian kernel estimator Totest if a pair of TF (119910) and target (119909) is modulated by 119898 theCMI difference can be calculated as
Δ119868 = 119868 (119909 119910 | 119898 = 1) minus 119868 (119909 119910 | 119898 = 0) (7)
The pair is determined to be modulated if Δ119868 = 0 The sig-nificance 119875 values for Δ119868 = 0 is computed using permutationtests
33 Mimosa Similarly to MINDy Mimosa [45] was pro-posed to identifymodulated TF-target pairs However it doesnot preselect a set of modulators of interest but rather aims toalso search for the modulators Mimosa also assumes that amodulator takes only two states that is absence and presenceor 0 and 1 The modulated regulator-target pair is furtherassumed to be correlated when a modulator is present butuncorrelated when it is absent Therefore the distributionof a modulated TF-target pair 119909 and 119910 naturally follows amixture distribution
119901 (119909 119910) = 120587119901 (119909 119910 | 119898 = 0) + (1 minus 120587) 119901 (119909 119910119898 = 1) (8)
where 120587 is the probability of themodulator being absent Par-ticularly an uncorrelated and correlated bivariant Gaussiandistributions were introduced to model different modulatedregulator-target relationship such that
119901 (119909 119910 | 119898 = 0) =
1
2120587
119890minus(12)(119909
2+1199102) (9a)
119901 (119909 119910 | 119898 = 1)
=
1
2120587radic1 minus 1205722
119890minus(12)(119909
2+1199102+2120572119909119910)(1minus120572
2)
(9b)
where 120572 models the correlation between 119909 and 119910 when themodulator is present With this model Mimosa sets out tofit the samples of every pair of potential regulator targetwith the mixture model (7) This is equivalent to finding
Advances in Bioinformatics 5
MicroRNAs
Modulator
Regulator
Target3998400UTR
3998400UTRmRNA119898
mRNA 119910
Figure 4 Modulation of gene regulation by competing mRNAs
a partition of the paired expression samples into the cor-related and uncorrelated samples The paired expressionsamples that possess such correlated-uncorrelated partition(03 lt 120587 lt 07 and |120572| gt 08) are determined to bemodulated To identify the modulator of a (or a group of)modulated pair(s) a weighted 119905-test was developed to searchfor the genes whose expressions are differentially expressedin the correlated partition versus the uncorrelated partition
34 GEM (Gene Expression Modulator) GEM [46] improvesover MINDy by predicting how a modulator-TF interactionaffects the expression of the target gene It can detect newtypes of interactions that result in stronger correlation butlow Δ119868 which therefore would be missed by MINDy GEMhypothesizes that the correlation between the expression of amodulator 119898 and a target 119909 must change as that of the TF119909 changes Unlike the previous surveyed algorithms GEMfirst transforms the continuous expression levels to binarystates (up- (1) or down-expression (0)) and then works onlywith discrete expression states To model the hypothesizedrelationship the following model is proposed
119875 (119909 = 1 | 119910119898) = 120572119888+ 120572119898119898 + 120572
119910119910 + 120574119898119910 (10)
where 120572119888is a constant 120572
119898and 120572
119910model the effect of
modulator and TF on the target genes and 120574 represents theeffect of modulator-TF interaction on the target gene If themodulator-TF interaction has an effect on 119909 then 120574 willbe nonzero For a given (119909 119910119898) triplets GEM devised analgorithm to estimate the model coefficients in (10) and a testto determine if 120574 is nonzero or119898 is a modulator of 119909 and 119910
35MuTaMe (Mutually TargetedMREEnrichment) Thegoalof MuTaMe [21] is to identify ceRNA networks of a gene ofinterest (GoI) ormRNA that sharemiRNA response elements(MREs) of samemiRNAs Figure 4 shows twomRNAs whereone is the GoIy and the other is a candidate ceRNA ormodulator 119898 In the figure the miRNA represented in colorred has MREs in both mRNA 119910 and mRNA 119898 in this casethe presence of mRNA 119898 will start the competition with 119910for miRNA represented in color red
Thehypothesis ofMuTaMe is thatmRNAs that havemanyof the same MREs can regulate each other by competingfor miRNAs binding The input of this algorithm is a GoIwhich is targeted by a group of miRNAs known to the userThen from a database of predicted MREs for the entiretranscriptome it is possible to obtain the binding sites andits predicted locations in the 31015840UTR for all mRNAsThis datais used to generate scores for each mRNA based on severalfeatures
(a) the number of miRNAs that an mRNA119898 shares withthe GoI 119910
(b) the density of the predicted MREs for the miRNA itfavors the cases in which more MREs are located inshorter distances
(c) the distribution of the MREs for every miRNA itfavors situations in which theMREs tend to be evenlydistributed
(d) the number of MREs predicted to target 119898 it favorssituationswhere eachmiRNA containsmoreMREs in119898
Then each candidate transcript119898will be assigned a scorethat results from multiplying the scores in (a) to (d) Thisscore indicates the likelihood of the candidates to be ceRNAsand will be used to predict ceRNAs
36 Hermes Hermes [20] is an extension of MINDy thatinfers candidate modulators of miRNA activity from expres-sion profiles of genes and miRNAs of the same samplesHermes makes inferences by estimating the MI and CMIHowever different from MINDy (7) Hermes extracts thedependences of this triplet by studying the difference betweenthe CMI of 119909 expression and 119910 expression conditional on theexpression of119898 and theMI of 119909 and 119910 expressions as follows
119868 = 119868 (119909 119910 | 119898) minus 119868 (119909 119910) (11)
These quantities and their associated statistical signif-icance can be computed from collections of expressionof genes with number of samples 250 or greater Hermesexpands MINDy by providing the capacity to identify can-didate modulator genes of miRNAs activity The presenceof these modulators (119898) will affect the relation betweenthe expression of the miRNAs targeting a gene (119909) and theexpression level of this gene (119909)
In summary we surveyed some of the most popularalgorithms for the inference of modulator Additional mod-ulator identification algorithms are summarized in Table 1It is worth noting that the concept of modulator appliesto cases beyond discussed in this paper Such exampleincludes themultilayer integrated regulatorymodel proposedin Yan et al [49] where the top layer of regulators could bealso considered as ldquomodulatorsrdquo
4 Applications to Breast Cancer GeneExpression Data
Algorithms of utilizing modulator concept have been imple-mented in various software packages Here we will discuss
6 Advances in Bioinformatics
Table 1 Gene regulation network and modulator identification methods
Algorithm Features ReferencesARACNE Interaction network constructed via mutual information (MI) [14 42]
Network profiler A varying-coefficient structural equation model (SEM) to represent themodulator-dependent conditional independence between genes [47]
MINDy Gene-pair interaction dependency on modulator candidates by using theconditional mutual information (CMI) [44]
Mimosa Search for modulator by partition samples with a Gaussian mixture model [45]
GEM A probabilistic method for detecting modulators of TFs that affect the expression oftarget gene by using a priori knowledge and gene expression profiles [46]
MuTaMe Based on the hypothesis that shared MREs can regulate mRNAs by competing formicroRNAs binding [21]
Hermes Extension of MINDy to include microRNAs as candidate modulators by using CMIand MI from expression profiles of genes and miRNAs of the same samples [20]
ER120572 modulator Analyzes the interaction between TF and target gene conditioned on a group ofspecific modulator genes via a multiple linear regression [48]
two new applications MEGRA and TraceRNA implementedin-house specifically to utilize the concept of differentialcorrelation coefficients and ceRNAs to construct a modu-lated GRN with a predetermined modulator In the case ofMGERA we chose estrogen receptor ESR1 as the initialstarting point since it is one of the dominant and systemicfactor in breast cancer in the case of TraceRNAwe also chosegene ESR1 and its modulated gene network Preliminaryresults of applications to TCGA breast cancer data arereported in the following 2 sections
41MGERA TheModulatedGeneRegulationAnalysis algo-rithm (MGERA) was designed to explore gene regulationpairs modulated by the modulator 119898 The regulation pairscan be identified by examining the coexpression of two genesbased on Pearson correlation (similar to (7) in the contextof correlation coefficient) Fisher transformation is adoptedto normalize the correlation coefficients biased by samplesizes to obtain equivalent statistical power among data withdifferent sample sizes Statistical significance of differencein the absolute correlation coefficients between two genes istested by the student 119905-test following Fisher transformationFor the gene pairs with significantly different coefficientsbetween two genes active and deactive statuses are iden-tified by examining the modulated gene expression pairs(MGEPs)TheMGEPs are further combined to construct the119898 modulated gene regulation network for a systematic andcomprehensive view of interaction under modulation
To demonstrate the ability of MGERA we set estrogenreceptor (ER) as the modulator and applied the algorithm toTCGA breast cancer expression data [3] which contains 588expression profiles (461 ER+ and 127 ERminus) By using 119875 valuelt001 and the difference in the absolute Pearson correlationcoefficients gt06 as criteria we identified 2324 putativeER+ MGEPs and a highly connected ER+ modulated generegulation network was constructed (Figure 5) The top tengenes with highest connectivity was show in Table 2 Thecysteinetyrosine-rich 1 gene (CYYR1) connected to 142genes was identified as the top hub gene in the network andthus may serve as a key regulator under ER+ modulation
Table 2 Hub genes derived from modulated gene regulationnetwork (Figure 5)
Gene Number of ER+ MGEPsCYYR1 142MRAS 109C9orf19 95LOC339524 93PLEKHG1 92FBLN5 91BOC 91ANKRD35 89FAM107A 83C16orf77 73
Gene Ontology analysis of CYYR1 and its connected neigh-bor genes revealed significant association with extracellularmatrix epithelial tube formation and angiogenesis
42 TraceRNA To identify the regulationnetwork of ceRNAsfor a GoI we developed a web-based application TraceRNApresented earlier in [50] with extension to regulation networkconstructionThe analysis flow chart of TraceRNAwas shownin Figure 6 For a selected GoI the GoI binding miRNAs(GBmiRs) were derived either validated miRNAs from miR-TarBase [51] or predicted miRNAs from SVMicrO [52]ThenmRNAs (other than the given GoI) also targeted by GBmiRswere identified as the candidates of ceRNAs The relevant (ortumor-specific) gene expression data were used to furtherstrengthen relationship between the ceRNA candidates andGoI The candidate ceRNAs which coexpressed with GoIwere reported as putative ceRNAs To construct the generegulation network via GBmiRs we set each ceRNA as thesecondary GoI and the ceRNAs of these secondary GoIswere identified by applying the algorithm recursively Uponidentifying all the ceRNAs the regulation network of ceRNAsof a given GoI was constructed
Advances in Bioinformatics 7
EDG1GIMAP8
ACVR2A
MEOX1P2RY5
ESAM
KLF6
SOX7GSN
PALMDLYVE1
EGFLAM
OLFML2AGIMAP6
CD93
HSPA12B
SH2D3C
ATP8B4
P2RY14
ELMO3
MED29
GUCY1B3
SPNS2
ELTD1
RGS5
GGTA1
FAAH
KIAA1244
HIVEP2
RBP7
CDR1
GIMAP5
CD36ABCA9
LIMS2FOXO1
CYYR1 EBF3
MALLRSU1
FILIP1
KCNJ2 ROBO4
GIMAP1
CFH
LPPR4
C13orf33
C1QTNF7 EMX2OS
C10orf54
CIDEA
VWF
PODNPALLD
CILP
KGFLP1 LRRC32
OMD
ADH1BPLA2G5
C14orf37
WFWBCL6B
CCIN
ADAMTS16
AANAT
KIR3DL1
TPK1
EMILIN2
PLEKHQ1
ACVRL1F13A1
CCL4 PTGDS
BOC
LAYN
TSHZ2
RARRES1
ENPP6
MMP7
KRT6B
DSC1
ARID5BOLFM4
CX3CL1FLJ45803
GABRPDNM3
KCNMB111FAM3D
CPAMD8
SCPEP1PIK3C2G
BCL6
C12orf54
UTRN
IFNGR1IL33
USHBP1
GPBAR1PTH2R
NDRG2
CASQ2MAB21L1
CCL23CCL15CCCCEMP1
CRTAC1
SKIP
TP63
RRAGC
MMRN1
PLEKHF1
GUCY1A3ADCY4
HSD11B1
EBF2
RASD1 UACA
LGI4
ANKRD6
EGR2
RP5-1054A223ELN
GIMAP7
ADIPOQ
TLR4
HAND2
FAM20A
DPT
APDE1B
HHSOCS3
CD248
LCN6
TLL1
RSPO1
CXorf36
RRAD
CRISPLD2
PLB1
DYSF
RASSF222FAM101A
ST5
HSPG2
FGF1
ITGBL1LGI2
DCHS1
PLXDC2
RCN3DACT3SFRP2SPSB1
KCNJ15SRPX2IGFBP7 COL1A2
OR10A5
ACTA2PRICKLE1
XG
FMO1
C16orf30CLEC14A
TCF4C1S
LMAN1L
FBLN1SH3PXD2B
FMO3
MXRA8D22
HTRA1COL8A2
1DACT1
COL16A1
GLT8D2
LMOD1
SPON2
RFTN2
FILIP1L
MFAP5
MMP23BPRRX1 TNFAIP6
LUMCTSKFIBIN
COL6A3MAFB
SFRP4
KERACLDN5RSPO3FGF7
BHLHB3
C16orf77
BST1
SSPN
COL15A1TIE1
ABCC9
CRYABSV2B
LAMA2
SAA1NKAPL
FAM49AANKRD35CHL1
SAA2
SAV1
FBLN5TCEAL7KIAA0355CCDC80
C2orf40
THSD7APRELP
FRZBMMRN2
ALDH2
RASL12SERPING1
TAGLNLTBP2
CRB3
EPN3
TPD52
AP1M2
KIAA1543
TMC4
TJP3
MB
CDH23ZFP36
CLIC5
KL
FHL5
CCDC24
FAM86A
AVPR1ASTX12
FABP4
PDZRN3
ADH1CPPP1R12B
LASS6
SCARF1
LOC338328
ANKRD47SPARCL1SNOTCH4
JPH2TCF7L2
TM4SF18HECA
BSPRY
RASEF
SHROOM3
PVRL4
SPINT1
PROM2
SH2D3A
CLDN3
FAAH2
RAB25
LOC124220
RBM35A
C2orf15
PKP3
MARVELD3
FAM84B
VIL2
ELF3TACSTD1
PRSS8 C1orf172
C1orf34
OVOL1
ARL6IP1
MAGIX
CDS1
BIK
FXYD3
GSTO2
IGSF9
CAPN13
TRPS1
OCLN
DNASE2
C1orf210
P2RX4
LSR
C9orf7
EPHA10
AVAVAVAVFAM83H
KIAA1688
LRRC8E
FMOD
MMP11 RUNDC3B
CCDC107CLDN7
KRT19
CBLCCGN
C19orf46SLC44A4
KIAA1324
PRRG2
DDR1
NEBL
TMEM125
ATAD4
LRFN2
MARVELD2
IRX5
CLDN4
LRBA
SPDEF
LOC400451
SPINT2
DLG3
CREB3L4GRHL2
TFAP2A
PSD4
IRF6
RGS13
C1orf186
HDC
CPA3
AQP2
LCE1F
CYP11B1
GRAMD2
CLDN8
GPRIN2
TNN
PDGFD
GIMAP4
GJA12
SHANK3
MAGED2
PTCH2
KLK5PIGR
C4orf7
LGR6
SLC34A2
SELP
APOD
KIF19
PACRG
ZNF446
ZNF574
CREB3HHAT
PEX11B
C10orf81
KRT7
PSEN2
SLC25A44
SEC16A
CRHR1CCDC114
SPINK7
LCE3ETMEM95
OTOF
OR4D1
SSTR3
KLK9
OR10H2
ATN1
CDK5R2FLJ32214
KRT3
C1orf175
PROP1GHSR
PLA2G4D
TRGV5
TIGD5
VPS28
KIAA0143
DPY19L4
POLR2K
COMMD5
DYRK1B
PSORS1C2OR5H1
C10orf27
KRTAP11-1
LOC652968
OR2B11
OR5BF1
OR10P1
MGC40574
P2RX2 RASGRP4
Gene
Positive correlation under ER+ modulationNegative correlation under ER+ modulation
Figure 5 ER+ modulated gene regulation network
To identify ceRNA candidates three miRNAs bindingprediction algorithms SiteTest SVMicrO and BCMicrOwere used in TracRNA SiteTest is an algorithm similar toMuTaMe and uses UTR features for target prediction SVMi-crO [52] is an algorithm that uses a large number of sequence-level site as well as UTR features including binding secondarystructure energy and conservation whereas BCMicrO [53]employs a Bayesian approach that integrates predictions from6 popular algorithms including TargetScan miRanda PicTarmirTarget PITA and DIANA-microT Pearson correlationcoefficient was used to test the coexpression between the GoIand the candidate ceRNAs We utilized TCGA breast cancercohort [3] as the expression data by using 60 of GBmiRs
as commonmiRNAs and Pearson correlation coefficient gt09as criteria The final scores of putative ceRNAs (see Table 3last column) were generated by using Bordamergingmethodwhich rerank the sum of ranks from both GBmiR bindingand coexpression 119875 values [54] To illustrate the utility of theTraceRNA algorithm for breast cancer study we also focus onthe genes interacted with the estrogen receptor alpha ESR1with GBmiRs including miR-18a miR-18b miR-193b miR-19a miR-19b miR-206 miR-20b miR-22 miR-221 miR-222miR-29b and miR-302c The regulation network generatedby ESR1 as the initial GoI is shown in Figure 7 and the top18 ceRNAs are provided in Table 3 The TraceRNA algorithmcan be accessed httpcompgenomicsutsaeducerna
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
2 Advances in Bioinformatics
Interact
Regulator Target
119909 119910
(a)
Regulator
119909
119910
119911
119908
of 119910 and 119908
Regulator of 119911
Target of 119910
Target of 119909
Target of 119909
(b)
Figure 1 Regulator-target pair in genetic regulatory network model (a) basic regulator-target pair and (b) regulator-target complex
on sequence-specific binding sites at target genesrsquo promoterregions In contrast ceRNAs mediate gene regulation viacompeting miRNAs binding sites in target 31015840UTR regionwhich exist in gt50 of mRNAs [22 24] In this study wewill extend the current network construction methods byincorporating regulation via ceRNAs
In tumorigenesis gene mutation is the main cause ofthe cancer [25] The mutation may not directly reflect inthe change at the gene expression level however it willdisrupt gene regulation [26ndash28] In Hudson et al theyfound that mutated myostatin and MYL2 showed differentcoexpressionswhen comparing towild-typemyostatin Chunet al also showed that oncogenic KRAS modulates HIF-1120572 and HIF-2120572 target genes and in turn modulates cancermetabolism Stelniec-Klotz et al presented a complex hierar-chicalmodel ofKRASmodulated network followed by doubleperturbation experiments Shen et al [29] showed a temporalchange of GRNs modulated after the estradiol stimulationindicating important role of estrogen in modulating GRNsFunctionally modulation effect of high expression of ESR1was also reported byWilson andDering [30]where they stud-ied previously published microarray data with cells treatedwith hormone receptor agonists and antagonists [31ndash33] Inthis study a comprehensive review of existing algorithms touncover themodulators was provided Given either mutationor protein expression status was unknown under many ofreported studies the problem of how to partition the diversesamples with different conditions such as active or inactiveoncogene status (and perhaps a combination of multiplemutations) and the prediction of a putative modulator ofgene regulation remains a difficult task
By combining gene regulation obtained from coexpres-sion data and ceRNAs we report here an early attempt tounify two systems mathematically while assuming a knownmodulator estrogen receptor (ER) By employing the TCGA[3] breast tumor gene expressions data and their clinical testresult (ER status) we demonstrate the approach of obtainingGRN via ceRNAs and a new presentation of ER modulationeffects By integrating breast cancer data into our uniqueceRNAs discovery website we are uniquely positioned tofurther explore the ceRNA regulation network and further
develop the discovery algorithms in order to detect potentialmodulators of regulatory interactions
2 Models of Gene Regulation and Modulation
21 Regulation of Gene Expression Thecomplex relationshipsamong genes and their products in a cellular system canbe studied using genetic regulatory networks (GRNs) Thenetworksmodel the different states or phenotypes of a cellularsystem In this model the interactions are commonly mod-eled as regulator-target pairs with edges between regulatorand target pair representing their interaction direction asshown in Figure 1(a) In this model a target gene is a genewhose expression can be altered (activated or suppressed)by a regulator gene This definition of a target gene impliesthat any gene can be at some point a target gene or adirect or indirect regulator depending on its position in thegenetic regulatory network The regulator gene is a gene thatcontrols (activates or suppresses) its target genesrsquo expressionThe consequences of these activated (or suppressed) genessometimes are involved in specific biological functions suchas cell proliferation in cancer Examples of regulator-targetpair in biology are common For example a target geneCDCA7 (cell division cycle-associated protein 7) is a c-Myc(regulator) responsive gene and it is part of c-Myc-mediatedtransformation of lymphoblastoid cells Furthermore asshown in Figure 1(b) a regulator gene can also act as a targetgene if there exists an upstream regulator
If the interaction is modeled after Boolean network (BN)model [34] then
119910119894(119905 + 1) = 119891
119894(1199091198951(119905) 119909
119895119896(119905) 119910119894(119905)) (1)
where each regulator 119909119895isin 0 1 is a binary variable as well
as it is target 119910119894 As described by (1) the target 119910
119894at time
119905 + 1 is completely determined by the values of its regulatorsat time 119905 by means of a Boolean function 119891
119894isin 119865 where
119865 is a collection of Boolean functions Thus the Booleannetwork 119866(119881 119865) is defined as a set of nodes (genes) 119881 =1199091 1199092 119909
119899 and a list of functions (edges or interactions)
119865 = 1198911 1198912 119891
119899 Similarly such relationship can be
defined in the framework of Bayesian network where the
Advances in Bioinformatics 3
R
Target gene
TF
119910
(a)
Singaltransductionproteins
ReceptorHormone
(b)
microRNAs
3998400UTR
Target gene 119910
(c)
Figure 2 Three different cases of regulation of gene expression that share the network representation of a regulator target interaction
similar regulators-target relationship as defined in (1) can bemodeled by the distribution
119875 (119910119894(119905 + 1) 119909
1198951(119905) 119909
119895119896(119905) 119910119894(119905))
= 119875 (119910119894(119905 + 1) | Parents (119910
119894(119905 + 1)))
times 119875 (Parents (119910119894(119905 + 1)))
(2)
where Parents(119910119894(119905 + 1)) = 119909
1198951
(119905) 119909119895119896(119905) 119910119894(119905) is the set
of regulators or parents of 119910119894 119875(119910119894(119905 + 1) | Parents (119910
119894(119905 +
1))) is the conditional distribution defining the regulator-target relationship and 119875(Parents(119910
119894(119905 + 1))) models the
prior distribution of regulators Unlike in (1) the target andregulators in (2) are modeled as random variables Despiteof this difference in both (1) and (2) the target is always afunction (or conditional distribution) of the regulator (or par-ents)When the relationship is defined by a Boolean functionas in (1) the conditional distribution in (2) take the formof a binomial distribution (or a multinomial distributionwhen both regulators and target take more than two states)Other distributions such as the Gaussian and Poisson can beintroduced to model more complex relationships than theBoolean The network construction inference and controlhowever are beyond the scope of this paper and we leave thetopics to the literatures [9 35 36]
The interactions among genes and their products ina complex cellular process of gene expression are diversegoverned by the central dogma of molecular biology [37]There are different regulation mechanisms that can actuateduring different stages Figure 2 shows three different casesof regulation of gene expression Figure 2(a) shows the caseof regulation of expression in which a transcription factor(TF) regulates the expression of a protein-coding gene (indark grey) by binding to the promoter region of target gene 119910Figure 2(b) is the case of regulation at the protein level inwhich a ligand protein interacts with a receptor to activaterelay molecules to transduce outside signals directly into cellbehavior Figure 2(c) is the case of regulation at the RNAlevel in which one or more miRNAs regulate target mRNA119910 by translational repression or target transcript degradationvia binding to sequence-specific binding sites (called miRNAresponse elements or MREs) in 31015840UTR region As illustratedin Figure 2(c) the target genesproteins all contain a domainof binding or docking site enabling specific interactions
Modulator
Interact
Regulator Target
119909 119910
119898
Figure 3 Graphical representation of the triplet interaction ofregulator 119909 target 119910 and modulator119898
between regulator-target pairs a common element in net-work structure
22 Modulation of Gene Regulation Different from the con-cept of a coregulator commonly referred in the regulatorybiology a modulator denotes a gene or protein that is capableof altering the endogenous gene expression at one stage ortime In the context of this paper we specifically definea modulator to be a gene that can systemically influencethe interaction of regulator-target pair either to activateor suppress the interaction in the presenceabsence of themodulator One example of modulator is the widely studiedestrogen receptor (ER) in breast cancer studies [38ndash40] theER status determines not only the tumor progression but alsothe chemotherapy treatment outcomes It is well known thatbinding of estrogen to receptor facilitates the ER activitiesto activate or repress gene expression [41] thus effectivelymodulating the GRN Figure 3 illustrates the model of theinteraction between a modulator (119898) and a regulator (119909)target (119910) pair that it modulates
Following the convention used in (1) and (2) the modu-lation interaction in Figure 3 can be modeled by
119910 = F(119898)
(119909) (3)
where 119910 represents target expression 119909 represents the parents(regulators) of target 119910 andF(119898)(sdot) is the regulation function
4 Advances in Bioinformatics
modulated by119898 WhenF(119898)(sdot) is stochastic the relationshipis modeled by the conditional distribution as
119901 (119910 119909 | 119898) = 119901 (119910 | 119909119898) 119901 (119909 | 119898) (4)
where 119901(119910|119909119898) models the regulator-target relationshipmodulated by 119898 and 119901(119909|119898) defines the prior distributionof regulators (parents) expression modulated by119898 Differentdistribution models can be used to model different mech-anisms for modulation At the biological level there aredifferent mechanisms for modulation of the interaction 119909-119910and currently several algorithms for prediction of the mod-ulators has been developed This survey presents the latestformulations and algorithms for prediction of modulators
3 Survey of Algorithms of Gene Regulationand Modulation Discovery
During the past years many computational tools have beendeveloped for regulation network construction and thendepending on the hypothesis modulator concept can betested and extracted Here we will focus on modulator detec-tion algorithms (MINDy Mimosa GEM and Hermes) Tointroduce gene-gene interaction concept we will also brieflydiscuss algorithms for regulation network construction(ARACNE) and ceRNA identification algorithm (MuTaMe)
31 ARACNE (Algorithm for the Reconstruction of AccurateCellular Networks) ARACNE [14 42] is an algorithm thatextracts transcriptional networks from microarray data byusing an information-theoreticmethod to reduce the indirectinteractions ARACNE assumes that it is sufficient to estimate2-way marginal distributions when sample size119872 gt 100 ingenomics problems such that
119901 (119909119894) =
1
119911
119890minus[sum119873
119894=1120601119894(119909119894)+sum119873
119894119895120601119894119895(119909119894119910119895)] (5)
Or a candidate interaction can be identified using estima-tion of mutual information MI of genes 119909 and 119910 MI(119909 119910) =MI119909119910 where MI
119909119910= 1 if genes 119909 and 119910 are identical and
MI119909119910
is zero if 119901(119909 119910) = 119901(119909)119901(119910) or 119909 and 119910 are sta-tistically independent Specifically the estimation of mutualinformation of gene expressions 119909 and 119910 of regulator andtarget genes is done by using the Gaussian kernel estimatorTheARACNE takes additional two steps to clean the network(1) removing MI if its 119875 value is less than that derived fromtwo independent genes via random permutation and (2) dataprocessing inequality (DPI) The algorithm further assumesthat for a triplet gene (119892
119909 119892119910 119892119911) where 119892
119909regulates 119892
119911
through 119892119910 then
MI119909119911lt min (MI
119909119910MI119910119911) if 119909 997888rarr 119910 997888rarr 119911
with no alternative path(6)
where rarr represents regulation relationship In other wordsthe lowest mutual information MI
119909119911is from an indirect
interaction and thus shall be removed from the GRN by
ARACNE in the DPI step A similar algorithm was proposed[43] to utilize conditional mutual information to exploremore than 2 regulators
32 MINDy (Modulator Inference by Network Dynamics)Similar toARACNEMINDy is also an information-theoreticalgorithm [44] However MINDy aims to identify poten-tial transcription factor-(TF-target) gene pairs that can bemodulated by a candidate modulator MINDy assumes thatthe expressions of the modulated TF-target pairs are ofdifferent correlations under different expression state of themodulator For simplicity and computational considerationMINDy considers only two modulator expression states thatis up- (119898 = 1) or down-expression (119898 = 0) Then ittests if the expression correlations of potential TF-target pairsare significantly different for modulator up-expression versusdown-expression The modulator dependent correlation isassessed by the conditional mutual information (CMI) or119868(119909 119910 | 119898 = 0) and 119868(119909 119910 | 119898 = 1) Similar to ARACNEthe CMI is calculated using the Gaussian kernel estimator Totest if a pair of TF (119910) and target (119909) is modulated by 119898 theCMI difference can be calculated as
Δ119868 = 119868 (119909 119910 | 119898 = 1) minus 119868 (119909 119910 | 119898 = 0) (7)
The pair is determined to be modulated if Δ119868 = 0 The sig-nificance 119875 values for Δ119868 = 0 is computed using permutationtests
33 Mimosa Similarly to MINDy Mimosa [45] was pro-posed to identifymodulated TF-target pairs However it doesnot preselect a set of modulators of interest but rather aims toalso search for the modulators Mimosa also assumes that amodulator takes only two states that is absence and presenceor 0 and 1 The modulated regulator-target pair is furtherassumed to be correlated when a modulator is present butuncorrelated when it is absent Therefore the distributionof a modulated TF-target pair 119909 and 119910 naturally follows amixture distribution
119901 (119909 119910) = 120587119901 (119909 119910 | 119898 = 0) + (1 minus 120587) 119901 (119909 119910119898 = 1) (8)
where 120587 is the probability of themodulator being absent Par-ticularly an uncorrelated and correlated bivariant Gaussiandistributions were introduced to model different modulatedregulator-target relationship such that
119901 (119909 119910 | 119898 = 0) =
1
2120587
119890minus(12)(119909
2+1199102) (9a)
119901 (119909 119910 | 119898 = 1)
=
1
2120587radic1 minus 1205722
119890minus(12)(119909
2+1199102+2120572119909119910)(1minus120572
2)
(9b)
where 120572 models the correlation between 119909 and 119910 when themodulator is present With this model Mimosa sets out tofit the samples of every pair of potential regulator targetwith the mixture model (7) This is equivalent to finding
Advances in Bioinformatics 5
MicroRNAs
Modulator
Regulator
Target3998400UTR
3998400UTRmRNA119898
mRNA 119910
Figure 4 Modulation of gene regulation by competing mRNAs
a partition of the paired expression samples into the cor-related and uncorrelated samples The paired expressionsamples that possess such correlated-uncorrelated partition(03 lt 120587 lt 07 and |120572| gt 08) are determined to bemodulated To identify the modulator of a (or a group of)modulated pair(s) a weighted 119905-test was developed to searchfor the genes whose expressions are differentially expressedin the correlated partition versus the uncorrelated partition
34 GEM (Gene Expression Modulator) GEM [46] improvesover MINDy by predicting how a modulator-TF interactionaffects the expression of the target gene It can detect newtypes of interactions that result in stronger correlation butlow Δ119868 which therefore would be missed by MINDy GEMhypothesizes that the correlation between the expression of amodulator 119898 and a target 119909 must change as that of the TF119909 changes Unlike the previous surveyed algorithms GEMfirst transforms the continuous expression levels to binarystates (up- (1) or down-expression (0)) and then works onlywith discrete expression states To model the hypothesizedrelationship the following model is proposed
119875 (119909 = 1 | 119910119898) = 120572119888+ 120572119898119898 + 120572
119910119910 + 120574119898119910 (10)
where 120572119888is a constant 120572
119898and 120572
119910model the effect of
modulator and TF on the target genes and 120574 represents theeffect of modulator-TF interaction on the target gene If themodulator-TF interaction has an effect on 119909 then 120574 willbe nonzero For a given (119909 119910119898) triplets GEM devised analgorithm to estimate the model coefficients in (10) and a testto determine if 120574 is nonzero or119898 is a modulator of 119909 and 119910
35MuTaMe (Mutually TargetedMREEnrichment) Thegoalof MuTaMe [21] is to identify ceRNA networks of a gene ofinterest (GoI) ormRNA that sharemiRNA response elements(MREs) of samemiRNAs Figure 4 shows twomRNAs whereone is the GoIy and the other is a candidate ceRNA ormodulator 119898 In the figure the miRNA represented in colorred has MREs in both mRNA 119910 and mRNA 119898 in this casethe presence of mRNA 119898 will start the competition with 119910for miRNA represented in color red
Thehypothesis ofMuTaMe is thatmRNAs that havemanyof the same MREs can regulate each other by competingfor miRNAs binding The input of this algorithm is a GoIwhich is targeted by a group of miRNAs known to the userThen from a database of predicted MREs for the entiretranscriptome it is possible to obtain the binding sites andits predicted locations in the 31015840UTR for all mRNAsThis datais used to generate scores for each mRNA based on severalfeatures
(a) the number of miRNAs that an mRNA119898 shares withthe GoI 119910
(b) the density of the predicted MREs for the miRNA itfavors the cases in which more MREs are located inshorter distances
(c) the distribution of the MREs for every miRNA itfavors situations in which theMREs tend to be evenlydistributed
(d) the number of MREs predicted to target 119898 it favorssituationswhere eachmiRNA containsmoreMREs in119898
Then each candidate transcript119898will be assigned a scorethat results from multiplying the scores in (a) to (d) Thisscore indicates the likelihood of the candidates to be ceRNAsand will be used to predict ceRNAs
36 Hermes Hermes [20] is an extension of MINDy thatinfers candidate modulators of miRNA activity from expres-sion profiles of genes and miRNAs of the same samplesHermes makes inferences by estimating the MI and CMIHowever different from MINDy (7) Hermes extracts thedependences of this triplet by studying the difference betweenthe CMI of 119909 expression and 119910 expression conditional on theexpression of119898 and theMI of 119909 and 119910 expressions as follows
119868 = 119868 (119909 119910 | 119898) minus 119868 (119909 119910) (11)
These quantities and their associated statistical signif-icance can be computed from collections of expressionof genes with number of samples 250 or greater Hermesexpands MINDy by providing the capacity to identify can-didate modulator genes of miRNAs activity The presenceof these modulators (119898) will affect the relation betweenthe expression of the miRNAs targeting a gene (119909) and theexpression level of this gene (119909)
In summary we surveyed some of the most popularalgorithms for the inference of modulator Additional mod-ulator identification algorithms are summarized in Table 1It is worth noting that the concept of modulator appliesto cases beyond discussed in this paper Such exampleincludes themultilayer integrated regulatorymodel proposedin Yan et al [49] where the top layer of regulators could bealso considered as ldquomodulatorsrdquo
4 Applications to Breast Cancer GeneExpression Data
Algorithms of utilizing modulator concept have been imple-mented in various software packages Here we will discuss
6 Advances in Bioinformatics
Table 1 Gene regulation network and modulator identification methods
Algorithm Features ReferencesARACNE Interaction network constructed via mutual information (MI) [14 42]
Network profiler A varying-coefficient structural equation model (SEM) to represent themodulator-dependent conditional independence between genes [47]
MINDy Gene-pair interaction dependency on modulator candidates by using theconditional mutual information (CMI) [44]
Mimosa Search for modulator by partition samples with a Gaussian mixture model [45]
GEM A probabilistic method for detecting modulators of TFs that affect the expression oftarget gene by using a priori knowledge and gene expression profiles [46]
MuTaMe Based on the hypothesis that shared MREs can regulate mRNAs by competing formicroRNAs binding [21]
Hermes Extension of MINDy to include microRNAs as candidate modulators by using CMIand MI from expression profiles of genes and miRNAs of the same samples [20]
ER120572 modulator Analyzes the interaction between TF and target gene conditioned on a group ofspecific modulator genes via a multiple linear regression [48]
two new applications MEGRA and TraceRNA implementedin-house specifically to utilize the concept of differentialcorrelation coefficients and ceRNAs to construct a modu-lated GRN with a predetermined modulator In the case ofMGERA we chose estrogen receptor ESR1 as the initialstarting point since it is one of the dominant and systemicfactor in breast cancer in the case of TraceRNAwe also chosegene ESR1 and its modulated gene network Preliminaryresults of applications to TCGA breast cancer data arereported in the following 2 sections
41MGERA TheModulatedGeneRegulationAnalysis algo-rithm (MGERA) was designed to explore gene regulationpairs modulated by the modulator 119898 The regulation pairscan be identified by examining the coexpression of two genesbased on Pearson correlation (similar to (7) in the contextof correlation coefficient) Fisher transformation is adoptedto normalize the correlation coefficients biased by samplesizes to obtain equivalent statistical power among data withdifferent sample sizes Statistical significance of differencein the absolute correlation coefficients between two genes istested by the student 119905-test following Fisher transformationFor the gene pairs with significantly different coefficientsbetween two genes active and deactive statuses are iden-tified by examining the modulated gene expression pairs(MGEPs)TheMGEPs are further combined to construct the119898 modulated gene regulation network for a systematic andcomprehensive view of interaction under modulation
To demonstrate the ability of MGERA we set estrogenreceptor (ER) as the modulator and applied the algorithm toTCGA breast cancer expression data [3] which contains 588expression profiles (461 ER+ and 127 ERminus) By using 119875 valuelt001 and the difference in the absolute Pearson correlationcoefficients gt06 as criteria we identified 2324 putativeER+ MGEPs and a highly connected ER+ modulated generegulation network was constructed (Figure 5) The top tengenes with highest connectivity was show in Table 2 Thecysteinetyrosine-rich 1 gene (CYYR1) connected to 142genes was identified as the top hub gene in the network andthus may serve as a key regulator under ER+ modulation
Table 2 Hub genes derived from modulated gene regulationnetwork (Figure 5)
Gene Number of ER+ MGEPsCYYR1 142MRAS 109C9orf19 95LOC339524 93PLEKHG1 92FBLN5 91BOC 91ANKRD35 89FAM107A 83C16orf77 73
Gene Ontology analysis of CYYR1 and its connected neigh-bor genes revealed significant association with extracellularmatrix epithelial tube formation and angiogenesis
42 TraceRNA To identify the regulationnetwork of ceRNAsfor a GoI we developed a web-based application TraceRNApresented earlier in [50] with extension to regulation networkconstructionThe analysis flow chart of TraceRNAwas shownin Figure 6 For a selected GoI the GoI binding miRNAs(GBmiRs) were derived either validated miRNAs from miR-TarBase [51] or predicted miRNAs from SVMicrO [52]ThenmRNAs (other than the given GoI) also targeted by GBmiRswere identified as the candidates of ceRNAs The relevant (ortumor-specific) gene expression data were used to furtherstrengthen relationship between the ceRNA candidates andGoI The candidate ceRNAs which coexpressed with GoIwere reported as putative ceRNAs To construct the generegulation network via GBmiRs we set each ceRNA as thesecondary GoI and the ceRNAs of these secondary GoIswere identified by applying the algorithm recursively Uponidentifying all the ceRNAs the regulation network of ceRNAsof a given GoI was constructed
Advances in Bioinformatics 7
EDG1GIMAP8
ACVR2A
MEOX1P2RY5
ESAM
KLF6
SOX7GSN
PALMDLYVE1
EGFLAM
OLFML2AGIMAP6
CD93
HSPA12B
SH2D3C
ATP8B4
P2RY14
ELMO3
MED29
GUCY1B3
SPNS2
ELTD1
RGS5
GGTA1
FAAH
KIAA1244
HIVEP2
RBP7
CDR1
GIMAP5
CD36ABCA9
LIMS2FOXO1
CYYR1 EBF3
MALLRSU1
FILIP1
KCNJ2 ROBO4
GIMAP1
CFH
LPPR4
C13orf33
C1QTNF7 EMX2OS
C10orf54
CIDEA
VWF
PODNPALLD
CILP
KGFLP1 LRRC32
OMD
ADH1BPLA2G5
C14orf37
WFWBCL6B
CCIN
ADAMTS16
AANAT
KIR3DL1
TPK1
EMILIN2
PLEKHQ1
ACVRL1F13A1
CCL4 PTGDS
BOC
LAYN
TSHZ2
RARRES1
ENPP6
MMP7
KRT6B
DSC1
ARID5BOLFM4
CX3CL1FLJ45803
GABRPDNM3
KCNMB111FAM3D
CPAMD8
SCPEP1PIK3C2G
BCL6
C12orf54
UTRN
IFNGR1IL33
USHBP1
GPBAR1PTH2R
NDRG2
CASQ2MAB21L1
CCL23CCL15CCCCEMP1
CRTAC1
SKIP
TP63
RRAGC
MMRN1
PLEKHF1
GUCY1A3ADCY4
HSD11B1
EBF2
RASD1 UACA
LGI4
ANKRD6
EGR2
RP5-1054A223ELN
GIMAP7
ADIPOQ
TLR4
HAND2
FAM20A
DPT
APDE1B
HHSOCS3
CD248
LCN6
TLL1
RSPO1
CXorf36
RRAD
CRISPLD2
PLB1
DYSF
RASSF222FAM101A
ST5
HSPG2
FGF1
ITGBL1LGI2
DCHS1
PLXDC2
RCN3DACT3SFRP2SPSB1
KCNJ15SRPX2IGFBP7 COL1A2
OR10A5
ACTA2PRICKLE1
XG
FMO1
C16orf30CLEC14A
TCF4C1S
LMAN1L
FBLN1SH3PXD2B
FMO3
MXRA8D22
HTRA1COL8A2
1DACT1
COL16A1
GLT8D2
LMOD1
SPON2
RFTN2
FILIP1L
MFAP5
MMP23BPRRX1 TNFAIP6
LUMCTSKFIBIN
COL6A3MAFB
SFRP4
KERACLDN5RSPO3FGF7
BHLHB3
C16orf77
BST1
SSPN
COL15A1TIE1
ABCC9
CRYABSV2B
LAMA2
SAA1NKAPL
FAM49AANKRD35CHL1
SAA2
SAV1
FBLN5TCEAL7KIAA0355CCDC80
C2orf40
THSD7APRELP
FRZBMMRN2
ALDH2
RASL12SERPING1
TAGLNLTBP2
CRB3
EPN3
TPD52
AP1M2
KIAA1543
TMC4
TJP3
MB
CDH23ZFP36
CLIC5
KL
FHL5
CCDC24
FAM86A
AVPR1ASTX12
FABP4
PDZRN3
ADH1CPPP1R12B
LASS6
SCARF1
LOC338328
ANKRD47SPARCL1SNOTCH4
JPH2TCF7L2
TM4SF18HECA
BSPRY
RASEF
SHROOM3
PVRL4
SPINT1
PROM2
SH2D3A
CLDN3
FAAH2
RAB25
LOC124220
RBM35A
C2orf15
PKP3
MARVELD3
FAM84B
VIL2
ELF3TACSTD1
PRSS8 C1orf172
C1orf34
OVOL1
ARL6IP1
MAGIX
CDS1
BIK
FXYD3
GSTO2
IGSF9
CAPN13
TRPS1
OCLN
DNASE2
C1orf210
P2RX4
LSR
C9orf7
EPHA10
AVAVAVAVFAM83H
KIAA1688
LRRC8E
FMOD
MMP11 RUNDC3B
CCDC107CLDN7
KRT19
CBLCCGN
C19orf46SLC44A4
KIAA1324
PRRG2
DDR1
NEBL
TMEM125
ATAD4
LRFN2
MARVELD2
IRX5
CLDN4
LRBA
SPDEF
LOC400451
SPINT2
DLG3
CREB3L4GRHL2
TFAP2A
PSD4
IRF6
RGS13
C1orf186
HDC
CPA3
AQP2
LCE1F
CYP11B1
GRAMD2
CLDN8
GPRIN2
TNN
PDGFD
GIMAP4
GJA12
SHANK3
MAGED2
PTCH2
KLK5PIGR
C4orf7
LGR6
SLC34A2
SELP
APOD
KIF19
PACRG
ZNF446
ZNF574
CREB3HHAT
PEX11B
C10orf81
KRT7
PSEN2
SLC25A44
SEC16A
CRHR1CCDC114
SPINK7
LCE3ETMEM95
OTOF
OR4D1
SSTR3
KLK9
OR10H2
ATN1
CDK5R2FLJ32214
KRT3
C1orf175
PROP1GHSR
PLA2G4D
TRGV5
TIGD5
VPS28
KIAA0143
DPY19L4
POLR2K
COMMD5
DYRK1B
PSORS1C2OR5H1
C10orf27
KRTAP11-1
LOC652968
OR2B11
OR5BF1
OR10P1
MGC40574
P2RX2 RASGRP4
Gene
Positive correlation under ER+ modulationNegative correlation under ER+ modulation
Figure 5 ER+ modulated gene regulation network
To identify ceRNA candidates three miRNAs bindingprediction algorithms SiteTest SVMicrO and BCMicrOwere used in TracRNA SiteTest is an algorithm similar toMuTaMe and uses UTR features for target prediction SVMi-crO [52] is an algorithm that uses a large number of sequence-level site as well as UTR features including binding secondarystructure energy and conservation whereas BCMicrO [53]employs a Bayesian approach that integrates predictions from6 popular algorithms including TargetScan miRanda PicTarmirTarget PITA and DIANA-microT Pearson correlationcoefficient was used to test the coexpression between the GoIand the candidate ceRNAs We utilized TCGA breast cancercohort [3] as the expression data by using 60 of GBmiRs
as commonmiRNAs and Pearson correlation coefficient gt09as criteria The final scores of putative ceRNAs (see Table 3last column) were generated by using Bordamergingmethodwhich rerank the sum of ranks from both GBmiR bindingand coexpression 119875 values [54] To illustrate the utility of theTraceRNA algorithm for breast cancer study we also focus onthe genes interacted with the estrogen receptor alpha ESR1with GBmiRs including miR-18a miR-18b miR-193b miR-19a miR-19b miR-206 miR-20b miR-22 miR-221 miR-222miR-29b and miR-302c The regulation network generatedby ESR1 as the initial GoI is shown in Figure 7 and the top18 ceRNAs are provided in Table 3 The TraceRNA algorithmcan be accessed httpcompgenomicsutsaeducerna
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
Advances in Bioinformatics 3
R
Target gene
TF
119910
(a)
Singaltransductionproteins
ReceptorHormone
(b)
microRNAs
3998400UTR
Target gene 119910
(c)
Figure 2 Three different cases of regulation of gene expression that share the network representation of a regulator target interaction
similar regulators-target relationship as defined in (1) can bemodeled by the distribution
119875 (119910119894(119905 + 1) 119909
1198951(119905) 119909
119895119896(119905) 119910119894(119905))
= 119875 (119910119894(119905 + 1) | Parents (119910
119894(119905 + 1)))
times 119875 (Parents (119910119894(119905 + 1)))
(2)
where Parents(119910119894(119905 + 1)) = 119909
1198951
(119905) 119909119895119896(119905) 119910119894(119905) is the set
of regulators or parents of 119910119894 119875(119910119894(119905 + 1) | Parents (119910
119894(119905 +
1))) is the conditional distribution defining the regulator-target relationship and 119875(Parents(119910
119894(119905 + 1))) models the
prior distribution of regulators Unlike in (1) the target andregulators in (2) are modeled as random variables Despiteof this difference in both (1) and (2) the target is always afunction (or conditional distribution) of the regulator (or par-ents)When the relationship is defined by a Boolean functionas in (1) the conditional distribution in (2) take the formof a binomial distribution (or a multinomial distributionwhen both regulators and target take more than two states)Other distributions such as the Gaussian and Poisson can beintroduced to model more complex relationships than theBoolean The network construction inference and controlhowever are beyond the scope of this paper and we leave thetopics to the literatures [9 35 36]
The interactions among genes and their products ina complex cellular process of gene expression are diversegoverned by the central dogma of molecular biology [37]There are different regulation mechanisms that can actuateduring different stages Figure 2 shows three different casesof regulation of gene expression Figure 2(a) shows the caseof regulation of expression in which a transcription factor(TF) regulates the expression of a protein-coding gene (indark grey) by binding to the promoter region of target gene 119910Figure 2(b) is the case of regulation at the protein level inwhich a ligand protein interacts with a receptor to activaterelay molecules to transduce outside signals directly into cellbehavior Figure 2(c) is the case of regulation at the RNAlevel in which one or more miRNAs regulate target mRNA119910 by translational repression or target transcript degradationvia binding to sequence-specific binding sites (called miRNAresponse elements or MREs) in 31015840UTR region As illustratedin Figure 2(c) the target genesproteins all contain a domainof binding or docking site enabling specific interactions
Modulator
Interact
Regulator Target
119909 119910
119898
Figure 3 Graphical representation of the triplet interaction ofregulator 119909 target 119910 and modulator119898
between regulator-target pairs a common element in net-work structure
22 Modulation of Gene Regulation Different from the con-cept of a coregulator commonly referred in the regulatorybiology a modulator denotes a gene or protein that is capableof altering the endogenous gene expression at one stage ortime In the context of this paper we specifically definea modulator to be a gene that can systemically influencethe interaction of regulator-target pair either to activateor suppress the interaction in the presenceabsence of themodulator One example of modulator is the widely studiedestrogen receptor (ER) in breast cancer studies [38ndash40] theER status determines not only the tumor progression but alsothe chemotherapy treatment outcomes It is well known thatbinding of estrogen to receptor facilitates the ER activitiesto activate or repress gene expression [41] thus effectivelymodulating the GRN Figure 3 illustrates the model of theinteraction between a modulator (119898) and a regulator (119909)target (119910) pair that it modulates
Following the convention used in (1) and (2) the modu-lation interaction in Figure 3 can be modeled by
119910 = F(119898)
(119909) (3)
where 119910 represents target expression 119909 represents the parents(regulators) of target 119910 andF(119898)(sdot) is the regulation function
4 Advances in Bioinformatics
modulated by119898 WhenF(119898)(sdot) is stochastic the relationshipis modeled by the conditional distribution as
119901 (119910 119909 | 119898) = 119901 (119910 | 119909119898) 119901 (119909 | 119898) (4)
where 119901(119910|119909119898) models the regulator-target relationshipmodulated by 119898 and 119901(119909|119898) defines the prior distributionof regulators (parents) expression modulated by119898 Differentdistribution models can be used to model different mech-anisms for modulation At the biological level there aredifferent mechanisms for modulation of the interaction 119909-119910and currently several algorithms for prediction of the mod-ulators has been developed This survey presents the latestformulations and algorithms for prediction of modulators
3 Survey of Algorithms of Gene Regulationand Modulation Discovery
During the past years many computational tools have beendeveloped for regulation network construction and thendepending on the hypothesis modulator concept can betested and extracted Here we will focus on modulator detec-tion algorithms (MINDy Mimosa GEM and Hermes) Tointroduce gene-gene interaction concept we will also brieflydiscuss algorithms for regulation network construction(ARACNE) and ceRNA identification algorithm (MuTaMe)
31 ARACNE (Algorithm for the Reconstruction of AccurateCellular Networks) ARACNE [14 42] is an algorithm thatextracts transcriptional networks from microarray data byusing an information-theoreticmethod to reduce the indirectinteractions ARACNE assumes that it is sufficient to estimate2-way marginal distributions when sample size119872 gt 100 ingenomics problems such that
119901 (119909119894) =
1
119911
119890minus[sum119873
119894=1120601119894(119909119894)+sum119873
119894119895120601119894119895(119909119894119910119895)] (5)
Or a candidate interaction can be identified using estima-tion of mutual information MI of genes 119909 and 119910 MI(119909 119910) =MI119909119910 where MI
119909119910= 1 if genes 119909 and 119910 are identical and
MI119909119910
is zero if 119901(119909 119910) = 119901(119909)119901(119910) or 119909 and 119910 are sta-tistically independent Specifically the estimation of mutualinformation of gene expressions 119909 and 119910 of regulator andtarget genes is done by using the Gaussian kernel estimatorTheARACNE takes additional two steps to clean the network(1) removing MI if its 119875 value is less than that derived fromtwo independent genes via random permutation and (2) dataprocessing inequality (DPI) The algorithm further assumesthat for a triplet gene (119892
119909 119892119910 119892119911) where 119892
119909regulates 119892
119911
through 119892119910 then
MI119909119911lt min (MI
119909119910MI119910119911) if 119909 997888rarr 119910 997888rarr 119911
with no alternative path(6)
where rarr represents regulation relationship In other wordsthe lowest mutual information MI
119909119911is from an indirect
interaction and thus shall be removed from the GRN by
ARACNE in the DPI step A similar algorithm was proposed[43] to utilize conditional mutual information to exploremore than 2 regulators
32 MINDy (Modulator Inference by Network Dynamics)Similar toARACNEMINDy is also an information-theoreticalgorithm [44] However MINDy aims to identify poten-tial transcription factor-(TF-target) gene pairs that can bemodulated by a candidate modulator MINDy assumes thatthe expressions of the modulated TF-target pairs are ofdifferent correlations under different expression state of themodulator For simplicity and computational considerationMINDy considers only two modulator expression states thatis up- (119898 = 1) or down-expression (119898 = 0) Then ittests if the expression correlations of potential TF-target pairsare significantly different for modulator up-expression versusdown-expression The modulator dependent correlation isassessed by the conditional mutual information (CMI) or119868(119909 119910 | 119898 = 0) and 119868(119909 119910 | 119898 = 1) Similar to ARACNEthe CMI is calculated using the Gaussian kernel estimator Totest if a pair of TF (119910) and target (119909) is modulated by 119898 theCMI difference can be calculated as
Δ119868 = 119868 (119909 119910 | 119898 = 1) minus 119868 (119909 119910 | 119898 = 0) (7)
The pair is determined to be modulated if Δ119868 = 0 The sig-nificance 119875 values for Δ119868 = 0 is computed using permutationtests
33 Mimosa Similarly to MINDy Mimosa [45] was pro-posed to identifymodulated TF-target pairs However it doesnot preselect a set of modulators of interest but rather aims toalso search for the modulators Mimosa also assumes that amodulator takes only two states that is absence and presenceor 0 and 1 The modulated regulator-target pair is furtherassumed to be correlated when a modulator is present butuncorrelated when it is absent Therefore the distributionof a modulated TF-target pair 119909 and 119910 naturally follows amixture distribution
119901 (119909 119910) = 120587119901 (119909 119910 | 119898 = 0) + (1 minus 120587) 119901 (119909 119910119898 = 1) (8)
where 120587 is the probability of themodulator being absent Par-ticularly an uncorrelated and correlated bivariant Gaussiandistributions were introduced to model different modulatedregulator-target relationship such that
119901 (119909 119910 | 119898 = 0) =
1
2120587
119890minus(12)(119909
2+1199102) (9a)
119901 (119909 119910 | 119898 = 1)
=
1
2120587radic1 minus 1205722
119890minus(12)(119909
2+1199102+2120572119909119910)(1minus120572
2)
(9b)
where 120572 models the correlation between 119909 and 119910 when themodulator is present With this model Mimosa sets out tofit the samples of every pair of potential regulator targetwith the mixture model (7) This is equivalent to finding
Advances in Bioinformatics 5
MicroRNAs
Modulator
Regulator
Target3998400UTR
3998400UTRmRNA119898
mRNA 119910
Figure 4 Modulation of gene regulation by competing mRNAs
a partition of the paired expression samples into the cor-related and uncorrelated samples The paired expressionsamples that possess such correlated-uncorrelated partition(03 lt 120587 lt 07 and |120572| gt 08) are determined to bemodulated To identify the modulator of a (or a group of)modulated pair(s) a weighted 119905-test was developed to searchfor the genes whose expressions are differentially expressedin the correlated partition versus the uncorrelated partition
34 GEM (Gene Expression Modulator) GEM [46] improvesover MINDy by predicting how a modulator-TF interactionaffects the expression of the target gene It can detect newtypes of interactions that result in stronger correlation butlow Δ119868 which therefore would be missed by MINDy GEMhypothesizes that the correlation between the expression of amodulator 119898 and a target 119909 must change as that of the TF119909 changes Unlike the previous surveyed algorithms GEMfirst transforms the continuous expression levels to binarystates (up- (1) or down-expression (0)) and then works onlywith discrete expression states To model the hypothesizedrelationship the following model is proposed
119875 (119909 = 1 | 119910119898) = 120572119888+ 120572119898119898 + 120572
119910119910 + 120574119898119910 (10)
where 120572119888is a constant 120572
119898and 120572
119910model the effect of
modulator and TF on the target genes and 120574 represents theeffect of modulator-TF interaction on the target gene If themodulator-TF interaction has an effect on 119909 then 120574 willbe nonzero For a given (119909 119910119898) triplets GEM devised analgorithm to estimate the model coefficients in (10) and a testto determine if 120574 is nonzero or119898 is a modulator of 119909 and 119910
35MuTaMe (Mutually TargetedMREEnrichment) Thegoalof MuTaMe [21] is to identify ceRNA networks of a gene ofinterest (GoI) ormRNA that sharemiRNA response elements(MREs) of samemiRNAs Figure 4 shows twomRNAs whereone is the GoIy and the other is a candidate ceRNA ormodulator 119898 In the figure the miRNA represented in colorred has MREs in both mRNA 119910 and mRNA 119898 in this casethe presence of mRNA 119898 will start the competition with 119910for miRNA represented in color red
Thehypothesis ofMuTaMe is thatmRNAs that havemanyof the same MREs can regulate each other by competingfor miRNAs binding The input of this algorithm is a GoIwhich is targeted by a group of miRNAs known to the userThen from a database of predicted MREs for the entiretranscriptome it is possible to obtain the binding sites andits predicted locations in the 31015840UTR for all mRNAsThis datais used to generate scores for each mRNA based on severalfeatures
(a) the number of miRNAs that an mRNA119898 shares withthe GoI 119910
(b) the density of the predicted MREs for the miRNA itfavors the cases in which more MREs are located inshorter distances
(c) the distribution of the MREs for every miRNA itfavors situations in which theMREs tend to be evenlydistributed
(d) the number of MREs predicted to target 119898 it favorssituationswhere eachmiRNA containsmoreMREs in119898
Then each candidate transcript119898will be assigned a scorethat results from multiplying the scores in (a) to (d) Thisscore indicates the likelihood of the candidates to be ceRNAsand will be used to predict ceRNAs
36 Hermes Hermes [20] is an extension of MINDy thatinfers candidate modulators of miRNA activity from expres-sion profiles of genes and miRNAs of the same samplesHermes makes inferences by estimating the MI and CMIHowever different from MINDy (7) Hermes extracts thedependences of this triplet by studying the difference betweenthe CMI of 119909 expression and 119910 expression conditional on theexpression of119898 and theMI of 119909 and 119910 expressions as follows
119868 = 119868 (119909 119910 | 119898) minus 119868 (119909 119910) (11)
These quantities and their associated statistical signif-icance can be computed from collections of expressionof genes with number of samples 250 or greater Hermesexpands MINDy by providing the capacity to identify can-didate modulator genes of miRNAs activity The presenceof these modulators (119898) will affect the relation betweenthe expression of the miRNAs targeting a gene (119909) and theexpression level of this gene (119909)
In summary we surveyed some of the most popularalgorithms for the inference of modulator Additional mod-ulator identification algorithms are summarized in Table 1It is worth noting that the concept of modulator appliesto cases beyond discussed in this paper Such exampleincludes themultilayer integrated regulatorymodel proposedin Yan et al [49] where the top layer of regulators could bealso considered as ldquomodulatorsrdquo
4 Applications to Breast Cancer GeneExpression Data
Algorithms of utilizing modulator concept have been imple-mented in various software packages Here we will discuss
6 Advances in Bioinformatics
Table 1 Gene regulation network and modulator identification methods
Algorithm Features ReferencesARACNE Interaction network constructed via mutual information (MI) [14 42]
Network profiler A varying-coefficient structural equation model (SEM) to represent themodulator-dependent conditional independence between genes [47]
MINDy Gene-pair interaction dependency on modulator candidates by using theconditional mutual information (CMI) [44]
Mimosa Search for modulator by partition samples with a Gaussian mixture model [45]
GEM A probabilistic method for detecting modulators of TFs that affect the expression oftarget gene by using a priori knowledge and gene expression profiles [46]
MuTaMe Based on the hypothesis that shared MREs can regulate mRNAs by competing formicroRNAs binding [21]
Hermes Extension of MINDy to include microRNAs as candidate modulators by using CMIand MI from expression profiles of genes and miRNAs of the same samples [20]
ER120572 modulator Analyzes the interaction between TF and target gene conditioned on a group ofspecific modulator genes via a multiple linear regression [48]
two new applications MEGRA and TraceRNA implementedin-house specifically to utilize the concept of differentialcorrelation coefficients and ceRNAs to construct a modu-lated GRN with a predetermined modulator In the case ofMGERA we chose estrogen receptor ESR1 as the initialstarting point since it is one of the dominant and systemicfactor in breast cancer in the case of TraceRNAwe also chosegene ESR1 and its modulated gene network Preliminaryresults of applications to TCGA breast cancer data arereported in the following 2 sections
41MGERA TheModulatedGeneRegulationAnalysis algo-rithm (MGERA) was designed to explore gene regulationpairs modulated by the modulator 119898 The regulation pairscan be identified by examining the coexpression of two genesbased on Pearson correlation (similar to (7) in the contextof correlation coefficient) Fisher transformation is adoptedto normalize the correlation coefficients biased by samplesizes to obtain equivalent statistical power among data withdifferent sample sizes Statistical significance of differencein the absolute correlation coefficients between two genes istested by the student 119905-test following Fisher transformationFor the gene pairs with significantly different coefficientsbetween two genes active and deactive statuses are iden-tified by examining the modulated gene expression pairs(MGEPs)TheMGEPs are further combined to construct the119898 modulated gene regulation network for a systematic andcomprehensive view of interaction under modulation
To demonstrate the ability of MGERA we set estrogenreceptor (ER) as the modulator and applied the algorithm toTCGA breast cancer expression data [3] which contains 588expression profiles (461 ER+ and 127 ERminus) By using 119875 valuelt001 and the difference in the absolute Pearson correlationcoefficients gt06 as criteria we identified 2324 putativeER+ MGEPs and a highly connected ER+ modulated generegulation network was constructed (Figure 5) The top tengenes with highest connectivity was show in Table 2 Thecysteinetyrosine-rich 1 gene (CYYR1) connected to 142genes was identified as the top hub gene in the network andthus may serve as a key regulator under ER+ modulation
Table 2 Hub genes derived from modulated gene regulationnetwork (Figure 5)
Gene Number of ER+ MGEPsCYYR1 142MRAS 109C9orf19 95LOC339524 93PLEKHG1 92FBLN5 91BOC 91ANKRD35 89FAM107A 83C16orf77 73
Gene Ontology analysis of CYYR1 and its connected neigh-bor genes revealed significant association with extracellularmatrix epithelial tube formation and angiogenesis
42 TraceRNA To identify the regulationnetwork of ceRNAsfor a GoI we developed a web-based application TraceRNApresented earlier in [50] with extension to regulation networkconstructionThe analysis flow chart of TraceRNAwas shownin Figure 6 For a selected GoI the GoI binding miRNAs(GBmiRs) were derived either validated miRNAs from miR-TarBase [51] or predicted miRNAs from SVMicrO [52]ThenmRNAs (other than the given GoI) also targeted by GBmiRswere identified as the candidates of ceRNAs The relevant (ortumor-specific) gene expression data were used to furtherstrengthen relationship between the ceRNA candidates andGoI The candidate ceRNAs which coexpressed with GoIwere reported as putative ceRNAs To construct the generegulation network via GBmiRs we set each ceRNA as thesecondary GoI and the ceRNAs of these secondary GoIswere identified by applying the algorithm recursively Uponidentifying all the ceRNAs the regulation network of ceRNAsof a given GoI was constructed
Advances in Bioinformatics 7
EDG1GIMAP8
ACVR2A
MEOX1P2RY5
ESAM
KLF6
SOX7GSN
PALMDLYVE1
EGFLAM
OLFML2AGIMAP6
CD93
HSPA12B
SH2D3C
ATP8B4
P2RY14
ELMO3
MED29
GUCY1B3
SPNS2
ELTD1
RGS5
GGTA1
FAAH
KIAA1244
HIVEP2
RBP7
CDR1
GIMAP5
CD36ABCA9
LIMS2FOXO1
CYYR1 EBF3
MALLRSU1
FILIP1
KCNJ2 ROBO4
GIMAP1
CFH
LPPR4
C13orf33
C1QTNF7 EMX2OS
C10orf54
CIDEA
VWF
PODNPALLD
CILP
KGFLP1 LRRC32
OMD
ADH1BPLA2G5
C14orf37
WFWBCL6B
CCIN
ADAMTS16
AANAT
KIR3DL1
TPK1
EMILIN2
PLEKHQ1
ACVRL1F13A1
CCL4 PTGDS
BOC
LAYN
TSHZ2
RARRES1
ENPP6
MMP7
KRT6B
DSC1
ARID5BOLFM4
CX3CL1FLJ45803
GABRPDNM3
KCNMB111FAM3D
CPAMD8
SCPEP1PIK3C2G
BCL6
C12orf54
UTRN
IFNGR1IL33
USHBP1
GPBAR1PTH2R
NDRG2
CASQ2MAB21L1
CCL23CCL15CCCCEMP1
CRTAC1
SKIP
TP63
RRAGC
MMRN1
PLEKHF1
GUCY1A3ADCY4
HSD11B1
EBF2
RASD1 UACA
LGI4
ANKRD6
EGR2
RP5-1054A223ELN
GIMAP7
ADIPOQ
TLR4
HAND2
FAM20A
DPT
APDE1B
HHSOCS3
CD248
LCN6
TLL1
RSPO1
CXorf36
RRAD
CRISPLD2
PLB1
DYSF
RASSF222FAM101A
ST5
HSPG2
FGF1
ITGBL1LGI2
DCHS1
PLXDC2
RCN3DACT3SFRP2SPSB1
KCNJ15SRPX2IGFBP7 COL1A2
OR10A5
ACTA2PRICKLE1
XG
FMO1
C16orf30CLEC14A
TCF4C1S
LMAN1L
FBLN1SH3PXD2B
FMO3
MXRA8D22
HTRA1COL8A2
1DACT1
COL16A1
GLT8D2
LMOD1
SPON2
RFTN2
FILIP1L
MFAP5
MMP23BPRRX1 TNFAIP6
LUMCTSKFIBIN
COL6A3MAFB
SFRP4
KERACLDN5RSPO3FGF7
BHLHB3
C16orf77
BST1
SSPN
COL15A1TIE1
ABCC9
CRYABSV2B
LAMA2
SAA1NKAPL
FAM49AANKRD35CHL1
SAA2
SAV1
FBLN5TCEAL7KIAA0355CCDC80
C2orf40
THSD7APRELP
FRZBMMRN2
ALDH2
RASL12SERPING1
TAGLNLTBP2
CRB3
EPN3
TPD52
AP1M2
KIAA1543
TMC4
TJP3
MB
CDH23ZFP36
CLIC5
KL
FHL5
CCDC24
FAM86A
AVPR1ASTX12
FABP4
PDZRN3
ADH1CPPP1R12B
LASS6
SCARF1
LOC338328
ANKRD47SPARCL1SNOTCH4
JPH2TCF7L2
TM4SF18HECA
BSPRY
RASEF
SHROOM3
PVRL4
SPINT1
PROM2
SH2D3A
CLDN3
FAAH2
RAB25
LOC124220
RBM35A
C2orf15
PKP3
MARVELD3
FAM84B
VIL2
ELF3TACSTD1
PRSS8 C1orf172
C1orf34
OVOL1
ARL6IP1
MAGIX
CDS1
BIK
FXYD3
GSTO2
IGSF9
CAPN13
TRPS1
OCLN
DNASE2
C1orf210
P2RX4
LSR
C9orf7
EPHA10
AVAVAVAVFAM83H
KIAA1688
LRRC8E
FMOD
MMP11 RUNDC3B
CCDC107CLDN7
KRT19
CBLCCGN
C19orf46SLC44A4
KIAA1324
PRRG2
DDR1
NEBL
TMEM125
ATAD4
LRFN2
MARVELD2
IRX5
CLDN4
LRBA
SPDEF
LOC400451
SPINT2
DLG3
CREB3L4GRHL2
TFAP2A
PSD4
IRF6
RGS13
C1orf186
HDC
CPA3
AQP2
LCE1F
CYP11B1
GRAMD2
CLDN8
GPRIN2
TNN
PDGFD
GIMAP4
GJA12
SHANK3
MAGED2
PTCH2
KLK5PIGR
C4orf7
LGR6
SLC34A2
SELP
APOD
KIF19
PACRG
ZNF446
ZNF574
CREB3HHAT
PEX11B
C10orf81
KRT7
PSEN2
SLC25A44
SEC16A
CRHR1CCDC114
SPINK7
LCE3ETMEM95
OTOF
OR4D1
SSTR3
KLK9
OR10H2
ATN1
CDK5R2FLJ32214
KRT3
C1orf175
PROP1GHSR
PLA2G4D
TRGV5
TIGD5
VPS28
KIAA0143
DPY19L4
POLR2K
COMMD5
DYRK1B
PSORS1C2OR5H1
C10orf27
KRTAP11-1
LOC652968
OR2B11
OR5BF1
OR10P1
MGC40574
P2RX2 RASGRP4
Gene
Positive correlation under ER+ modulationNegative correlation under ER+ modulation
Figure 5 ER+ modulated gene regulation network
To identify ceRNA candidates three miRNAs bindingprediction algorithms SiteTest SVMicrO and BCMicrOwere used in TracRNA SiteTest is an algorithm similar toMuTaMe and uses UTR features for target prediction SVMi-crO [52] is an algorithm that uses a large number of sequence-level site as well as UTR features including binding secondarystructure energy and conservation whereas BCMicrO [53]employs a Bayesian approach that integrates predictions from6 popular algorithms including TargetScan miRanda PicTarmirTarget PITA and DIANA-microT Pearson correlationcoefficient was used to test the coexpression between the GoIand the candidate ceRNAs We utilized TCGA breast cancercohort [3] as the expression data by using 60 of GBmiRs
as commonmiRNAs and Pearson correlation coefficient gt09as criteria The final scores of putative ceRNAs (see Table 3last column) were generated by using Bordamergingmethodwhich rerank the sum of ranks from both GBmiR bindingand coexpression 119875 values [54] To illustrate the utility of theTraceRNA algorithm for breast cancer study we also focus onthe genes interacted with the estrogen receptor alpha ESR1with GBmiRs including miR-18a miR-18b miR-193b miR-19a miR-19b miR-206 miR-20b miR-22 miR-221 miR-222miR-29b and miR-302c The regulation network generatedby ESR1 as the initial GoI is shown in Figure 7 and the top18 ceRNAs are provided in Table 3 The TraceRNA algorithmcan be accessed httpcompgenomicsutsaeducerna
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
4 Advances in Bioinformatics
modulated by119898 WhenF(119898)(sdot) is stochastic the relationshipis modeled by the conditional distribution as
119901 (119910 119909 | 119898) = 119901 (119910 | 119909119898) 119901 (119909 | 119898) (4)
where 119901(119910|119909119898) models the regulator-target relationshipmodulated by 119898 and 119901(119909|119898) defines the prior distributionof regulators (parents) expression modulated by119898 Differentdistribution models can be used to model different mech-anisms for modulation At the biological level there aredifferent mechanisms for modulation of the interaction 119909-119910and currently several algorithms for prediction of the mod-ulators has been developed This survey presents the latestformulations and algorithms for prediction of modulators
3 Survey of Algorithms of Gene Regulationand Modulation Discovery
During the past years many computational tools have beendeveloped for regulation network construction and thendepending on the hypothesis modulator concept can betested and extracted Here we will focus on modulator detec-tion algorithms (MINDy Mimosa GEM and Hermes) Tointroduce gene-gene interaction concept we will also brieflydiscuss algorithms for regulation network construction(ARACNE) and ceRNA identification algorithm (MuTaMe)
31 ARACNE (Algorithm for the Reconstruction of AccurateCellular Networks) ARACNE [14 42] is an algorithm thatextracts transcriptional networks from microarray data byusing an information-theoreticmethod to reduce the indirectinteractions ARACNE assumes that it is sufficient to estimate2-way marginal distributions when sample size119872 gt 100 ingenomics problems such that
119901 (119909119894) =
1
119911
119890minus[sum119873
119894=1120601119894(119909119894)+sum119873
119894119895120601119894119895(119909119894119910119895)] (5)
Or a candidate interaction can be identified using estima-tion of mutual information MI of genes 119909 and 119910 MI(119909 119910) =MI119909119910 where MI
119909119910= 1 if genes 119909 and 119910 are identical and
MI119909119910
is zero if 119901(119909 119910) = 119901(119909)119901(119910) or 119909 and 119910 are sta-tistically independent Specifically the estimation of mutualinformation of gene expressions 119909 and 119910 of regulator andtarget genes is done by using the Gaussian kernel estimatorTheARACNE takes additional two steps to clean the network(1) removing MI if its 119875 value is less than that derived fromtwo independent genes via random permutation and (2) dataprocessing inequality (DPI) The algorithm further assumesthat for a triplet gene (119892
119909 119892119910 119892119911) where 119892
119909regulates 119892
119911
through 119892119910 then
MI119909119911lt min (MI
119909119910MI119910119911) if 119909 997888rarr 119910 997888rarr 119911
with no alternative path(6)
where rarr represents regulation relationship In other wordsthe lowest mutual information MI
119909119911is from an indirect
interaction and thus shall be removed from the GRN by
ARACNE in the DPI step A similar algorithm was proposed[43] to utilize conditional mutual information to exploremore than 2 regulators
32 MINDy (Modulator Inference by Network Dynamics)Similar toARACNEMINDy is also an information-theoreticalgorithm [44] However MINDy aims to identify poten-tial transcription factor-(TF-target) gene pairs that can bemodulated by a candidate modulator MINDy assumes thatthe expressions of the modulated TF-target pairs are ofdifferent correlations under different expression state of themodulator For simplicity and computational considerationMINDy considers only two modulator expression states thatis up- (119898 = 1) or down-expression (119898 = 0) Then ittests if the expression correlations of potential TF-target pairsare significantly different for modulator up-expression versusdown-expression The modulator dependent correlation isassessed by the conditional mutual information (CMI) or119868(119909 119910 | 119898 = 0) and 119868(119909 119910 | 119898 = 1) Similar to ARACNEthe CMI is calculated using the Gaussian kernel estimator Totest if a pair of TF (119910) and target (119909) is modulated by 119898 theCMI difference can be calculated as
Δ119868 = 119868 (119909 119910 | 119898 = 1) minus 119868 (119909 119910 | 119898 = 0) (7)
The pair is determined to be modulated if Δ119868 = 0 The sig-nificance 119875 values for Δ119868 = 0 is computed using permutationtests
33 Mimosa Similarly to MINDy Mimosa [45] was pro-posed to identifymodulated TF-target pairs However it doesnot preselect a set of modulators of interest but rather aims toalso search for the modulators Mimosa also assumes that amodulator takes only two states that is absence and presenceor 0 and 1 The modulated regulator-target pair is furtherassumed to be correlated when a modulator is present butuncorrelated when it is absent Therefore the distributionof a modulated TF-target pair 119909 and 119910 naturally follows amixture distribution
119901 (119909 119910) = 120587119901 (119909 119910 | 119898 = 0) + (1 minus 120587) 119901 (119909 119910119898 = 1) (8)
where 120587 is the probability of themodulator being absent Par-ticularly an uncorrelated and correlated bivariant Gaussiandistributions were introduced to model different modulatedregulator-target relationship such that
119901 (119909 119910 | 119898 = 0) =
1
2120587
119890minus(12)(119909
2+1199102) (9a)
119901 (119909 119910 | 119898 = 1)
=
1
2120587radic1 minus 1205722
119890minus(12)(119909
2+1199102+2120572119909119910)(1minus120572
2)
(9b)
where 120572 models the correlation between 119909 and 119910 when themodulator is present With this model Mimosa sets out tofit the samples of every pair of potential regulator targetwith the mixture model (7) This is equivalent to finding
Advances in Bioinformatics 5
MicroRNAs
Modulator
Regulator
Target3998400UTR
3998400UTRmRNA119898
mRNA 119910
Figure 4 Modulation of gene regulation by competing mRNAs
a partition of the paired expression samples into the cor-related and uncorrelated samples The paired expressionsamples that possess such correlated-uncorrelated partition(03 lt 120587 lt 07 and |120572| gt 08) are determined to bemodulated To identify the modulator of a (or a group of)modulated pair(s) a weighted 119905-test was developed to searchfor the genes whose expressions are differentially expressedin the correlated partition versus the uncorrelated partition
34 GEM (Gene Expression Modulator) GEM [46] improvesover MINDy by predicting how a modulator-TF interactionaffects the expression of the target gene It can detect newtypes of interactions that result in stronger correlation butlow Δ119868 which therefore would be missed by MINDy GEMhypothesizes that the correlation between the expression of amodulator 119898 and a target 119909 must change as that of the TF119909 changes Unlike the previous surveyed algorithms GEMfirst transforms the continuous expression levels to binarystates (up- (1) or down-expression (0)) and then works onlywith discrete expression states To model the hypothesizedrelationship the following model is proposed
119875 (119909 = 1 | 119910119898) = 120572119888+ 120572119898119898 + 120572
119910119910 + 120574119898119910 (10)
where 120572119888is a constant 120572
119898and 120572
119910model the effect of
modulator and TF on the target genes and 120574 represents theeffect of modulator-TF interaction on the target gene If themodulator-TF interaction has an effect on 119909 then 120574 willbe nonzero For a given (119909 119910119898) triplets GEM devised analgorithm to estimate the model coefficients in (10) and a testto determine if 120574 is nonzero or119898 is a modulator of 119909 and 119910
35MuTaMe (Mutually TargetedMREEnrichment) Thegoalof MuTaMe [21] is to identify ceRNA networks of a gene ofinterest (GoI) ormRNA that sharemiRNA response elements(MREs) of samemiRNAs Figure 4 shows twomRNAs whereone is the GoIy and the other is a candidate ceRNA ormodulator 119898 In the figure the miRNA represented in colorred has MREs in both mRNA 119910 and mRNA 119898 in this casethe presence of mRNA 119898 will start the competition with 119910for miRNA represented in color red
Thehypothesis ofMuTaMe is thatmRNAs that havemanyof the same MREs can regulate each other by competingfor miRNAs binding The input of this algorithm is a GoIwhich is targeted by a group of miRNAs known to the userThen from a database of predicted MREs for the entiretranscriptome it is possible to obtain the binding sites andits predicted locations in the 31015840UTR for all mRNAsThis datais used to generate scores for each mRNA based on severalfeatures
(a) the number of miRNAs that an mRNA119898 shares withthe GoI 119910
(b) the density of the predicted MREs for the miRNA itfavors the cases in which more MREs are located inshorter distances
(c) the distribution of the MREs for every miRNA itfavors situations in which theMREs tend to be evenlydistributed
(d) the number of MREs predicted to target 119898 it favorssituationswhere eachmiRNA containsmoreMREs in119898
Then each candidate transcript119898will be assigned a scorethat results from multiplying the scores in (a) to (d) Thisscore indicates the likelihood of the candidates to be ceRNAsand will be used to predict ceRNAs
36 Hermes Hermes [20] is an extension of MINDy thatinfers candidate modulators of miRNA activity from expres-sion profiles of genes and miRNAs of the same samplesHermes makes inferences by estimating the MI and CMIHowever different from MINDy (7) Hermes extracts thedependences of this triplet by studying the difference betweenthe CMI of 119909 expression and 119910 expression conditional on theexpression of119898 and theMI of 119909 and 119910 expressions as follows
119868 = 119868 (119909 119910 | 119898) minus 119868 (119909 119910) (11)
These quantities and their associated statistical signif-icance can be computed from collections of expressionof genes with number of samples 250 or greater Hermesexpands MINDy by providing the capacity to identify can-didate modulator genes of miRNAs activity The presenceof these modulators (119898) will affect the relation betweenthe expression of the miRNAs targeting a gene (119909) and theexpression level of this gene (119909)
In summary we surveyed some of the most popularalgorithms for the inference of modulator Additional mod-ulator identification algorithms are summarized in Table 1It is worth noting that the concept of modulator appliesto cases beyond discussed in this paper Such exampleincludes themultilayer integrated regulatorymodel proposedin Yan et al [49] where the top layer of regulators could bealso considered as ldquomodulatorsrdquo
4 Applications to Breast Cancer GeneExpression Data
Algorithms of utilizing modulator concept have been imple-mented in various software packages Here we will discuss
6 Advances in Bioinformatics
Table 1 Gene regulation network and modulator identification methods
Algorithm Features ReferencesARACNE Interaction network constructed via mutual information (MI) [14 42]
Network profiler A varying-coefficient structural equation model (SEM) to represent themodulator-dependent conditional independence between genes [47]
MINDy Gene-pair interaction dependency on modulator candidates by using theconditional mutual information (CMI) [44]
Mimosa Search for modulator by partition samples with a Gaussian mixture model [45]
GEM A probabilistic method for detecting modulators of TFs that affect the expression oftarget gene by using a priori knowledge and gene expression profiles [46]
MuTaMe Based on the hypothesis that shared MREs can regulate mRNAs by competing formicroRNAs binding [21]
Hermes Extension of MINDy to include microRNAs as candidate modulators by using CMIand MI from expression profiles of genes and miRNAs of the same samples [20]
ER120572 modulator Analyzes the interaction between TF and target gene conditioned on a group ofspecific modulator genes via a multiple linear regression [48]
two new applications MEGRA and TraceRNA implementedin-house specifically to utilize the concept of differentialcorrelation coefficients and ceRNAs to construct a modu-lated GRN with a predetermined modulator In the case ofMGERA we chose estrogen receptor ESR1 as the initialstarting point since it is one of the dominant and systemicfactor in breast cancer in the case of TraceRNAwe also chosegene ESR1 and its modulated gene network Preliminaryresults of applications to TCGA breast cancer data arereported in the following 2 sections
41MGERA TheModulatedGeneRegulationAnalysis algo-rithm (MGERA) was designed to explore gene regulationpairs modulated by the modulator 119898 The regulation pairscan be identified by examining the coexpression of two genesbased on Pearson correlation (similar to (7) in the contextof correlation coefficient) Fisher transformation is adoptedto normalize the correlation coefficients biased by samplesizes to obtain equivalent statistical power among data withdifferent sample sizes Statistical significance of differencein the absolute correlation coefficients between two genes istested by the student 119905-test following Fisher transformationFor the gene pairs with significantly different coefficientsbetween two genes active and deactive statuses are iden-tified by examining the modulated gene expression pairs(MGEPs)TheMGEPs are further combined to construct the119898 modulated gene regulation network for a systematic andcomprehensive view of interaction under modulation
To demonstrate the ability of MGERA we set estrogenreceptor (ER) as the modulator and applied the algorithm toTCGA breast cancer expression data [3] which contains 588expression profiles (461 ER+ and 127 ERminus) By using 119875 valuelt001 and the difference in the absolute Pearson correlationcoefficients gt06 as criteria we identified 2324 putativeER+ MGEPs and a highly connected ER+ modulated generegulation network was constructed (Figure 5) The top tengenes with highest connectivity was show in Table 2 Thecysteinetyrosine-rich 1 gene (CYYR1) connected to 142genes was identified as the top hub gene in the network andthus may serve as a key regulator under ER+ modulation
Table 2 Hub genes derived from modulated gene regulationnetwork (Figure 5)
Gene Number of ER+ MGEPsCYYR1 142MRAS 109C9orf19 95LOC339524 93PLEKHG1 92FBLN5 91BOC 91ANKRD35 89FAM107A 83C16orf77 73
Gene Ontology analysis of CYYR1 and its connected neigh-bor genes revealed significant association with extracellularmatrix epithelial tube formation and angiogenesis
42 TraceRNA To identify the regulationnetwork of ceRNAsfor a GoI we developed a web-based application TraceRNApresented earlier in [50] with extension to regulation networkconstructionThe analysis flow chart of TraceRNAwas shownin Figure 6 For a selected GoI the GoI binding miRNAs(GBmiRs) were derived either validated miRNAs from miR-TarBase [51] or predicted miRNAs from SVMicrO [52]ThenmRNAs (other than the given GoI) also targeted by GBmiRswere identified as the candidates of ceRNAs The relevant (ortumor-specific) gene expression data were used to furtherstrengthen relationship between the ceRNA candidates andGoI The candidate ceRNAs which coexpressed with GoIwere reported as putative ceRNAs To construct the generegulation network via GBmiRs we set each ceRNA as thesecondary GoI and the ceRNAs of these secondary GoIswere identified by applying the algorithm recursively Uponidentifying all the ceRNAs the regulation network of ceRNAsof a given GoI was constructed
Advances in Bioinformatics 7
EDG1GIMAP8
ACVR2A
MEOX1P2RY5
ESAM
KLF6
SOX7GSN
PALMDLYVE1
EGFLAM
OLFML2AGIMAP6
CD93
HSPA12B
SH2D3C
ATP8B4
P2RY14
ELMO3
MED29
GUCY1B3
SPNS2
ELTD1
RGS5
GGTA1
FAAH
KIAA1244
HIVEP2
RBP7
CDR1
GIMAP5
CD36ABCA9
LIMS2FOXO1
CYYR1 EBF3
MALLRSU1
FILIP1
KCNJ2 ROBO4
GIMAP1
CFH
LPPR4
C13orf33
C1QTNF7 EMX2OS
C10orf54
CIDEA
VWF
PODNPALLD
CILP
KGFLP1 LRRC32
OMD
ADH1BPLA2G5
C14orf37
WFWBCL6B
CCIN
ADAMTS16
AANAT
KIR3DL1
TPK1
EMILIN2
PLEKHQ1
ACVRL1F13A1
CCL4 PTGDS
BOC
LAYN
TSHZ2
RARRES1
ENPP6
MMP7
KRT6B
DSC1
ARID5BOLFM4
CX3CL1FLJ45803
GABRPDNM3
KCNMB111FAM3D
CPAMD8
SCPEP1PIK3C2G
BCL6
C12orf54
UTRN
IFNGR1IL33
USHBP1
GPBAR1PTH2R
NDRG2
CASQ2MAB21L1
CCL23CCL15CCCCEMP1
CRTAC1
SKIP
TP63
RRAGC
MMRN1
PLEKHF1
GUCY1A3ADCY4
HSD11B1
EBF2
RASD1 UACA
LGI4
ANKRD6
EGR2
RP5-1054A223ELN
GIMAP7
ADIPOQ
TLR4
HAND2
FAM20A
DPT
APDE1B
HHSOCS3
CD248
LCN6
TLL1
RSPO1
CXorf36
RRAD
CRISPLD2
PLB1
DYSF
RASSF222FAM101A
ST5
HSPG2
FGF1
ITGBL1LGI2
DCHS1
PLXDC2
RCN3DACT3SFRP2SPSB1
KCNJ15SRPX2IGFBP7 COL1A2
OR10A5
ACTA2PRICKLE1
XG
FMO1
C16orf30CLEC14A
TCF4C1S
LMAN1L
FBLN1SH3PXD2B
FMO3
MXRA8D22
HTRA1COL8A2
1DACT1
COL16A1
GLT8D2
LMOD1
SPON2
RFTN2
FILIP1L
MFAP5
MMP23BPRRX1 TNFAIP6
LUMCTSKFIBIN
COL6A3MAFB
SFRP4
KERACLDN5RSPO3FGF7
BHLHB3
C16orf77
BST1
SSPN
COL15A1TIE1
ABCC9
CRYABSV2B
LAMA2
SAA1NKAPL
FAM49AANKRD35CHL1
SAA2
SAV1
FBLN5TCEAL7KIAA0355CCDC80
C2orf40
THSD7APRELP
FRZBMMRN2
ALDH2
RASL12SERPING1
TAGLNLTBP2
CRB3
EPN3
TPD52
AP1M2
KIAA1543
TMC4
TJP3
MB
CDH23ZFP36
CLIC5
KL
FHL5
CCDC24
FAM86A
AVPR1ASTX12
FABP4
PDZRN3
ADH1CPPP1R12B
LASS6
SCARF1
LOC338328
ANKRD47SPARCL1SNOTCH4
JPH2TCF7L2
TM4SF18HECA
BSPRY
RASEF
SHROOM3
PVRL4
SPINT1
PROM2
SH2D3A
CLDN3
FAAH2
RAB25
LOC124220
RBM35A
C2orf15
PKP3
MARVELD3
FAM84B
VIL2
ELF3TACSTD1
PRSS8 C1orf172
C1orf34
OVOL1
ARL6IP1
MAGIX
CDS1
BIK
FXYD3
GSTO2
IGSF9
CAPN13
TRPS1
OCLN
DNASE2
C1orf210
P2RX4
LSR
C9orf7
EPHA10
AVAVAVAVFAM83H
KIAA1688
LRRC8E
FMOD
MMP11 RUNDC3B
CCDC107CLDN7
KRT19
CBLCCGN
C19orf46SLC44A4
KIAA1324
PRRG2
DDR1
NEBL
TMEM125
ATAD4
LRFN2
MARVELD2
IRX5
CLDN4
LRBA
SPDEF
LOC400451
SPINT2
DLG3
CREB3L4GRHL2
TFAP2A
PSD4
IRF6
RGS13
C1orf186
HDC
CPA3
AQP2
LCE1F
CYP11B1
GRAMD2
CLDN8
GPRIN2
TNN
PDGFD
GIMAP4
GJA12
SHANK3
MAGED2
PTCH2
KLK5PIGR
C4orf7
LGR6
SLC34A2
SELP
APOD
KIF19
PACRG
ZNF446
ZNF574
CREB3HHAT
PEX11B
C10orf81
KRT7
PSEN2
SLC25A44
SEC16A
CRHR1CCDC114
SPINK7
LCE3ETMEM95
OTOF
OR4D1
SSTR3
KLK9
OR10H2
ATN1
CDK5R2FLJ32214
KRT3
C1orf175
PROP1GHSR
PLA2G4D
TRGV5
TIGD5
VPS28
KIAA0143
DPY19L4
POLR2K
COMMD5
DYRK1B
PSORS1C2OR5H1
C10orf27
KRTAP11-1
LOC652968
OR2B11
OR5BF1
OR10P1
MGC40574
P2RX2 RASGRP4
Gene
Positive correlation under ER+ modulationNegative correlation under ER+ modulation
Figure 5 ER+ modulated gene regulation network
To identify ceRNA candidates three miRNAs bindingprediction algorithms SiteTest SVMicrO and BCMicrOwere used in TracRNA SiteTest is an algorithm similar toMuTaMe and uses UTR features for target prediction SVMi-crO [52] is an algorithm that uses a large number of sequence-level site as well as UTR features including binding secondarystructure energy and conservation whereas BCMicrO [53]employs a Bayesian approach that integrates predictions from6 popular algorithms including TargetScan miRanda PicTarmirTarget PITA and DIANA-microT Pearson correlationcoefficient was used to test the coexpression between the GoIand the candidate ceRNAs We utilized TCGA breast cancercohort [3] as the expression data by using 60 of GBmiRs
as commonmiRNAs and Pearson correlation coefficient gt09as criteria The final scores of putative ceRNAs (see Table 3last column) were generated by using Bordamergingmethodwhich rerank the sum of ranks from both GBmiR bindingand coexpression 119875 values [54] To illustrate the utility of theTraceRNA algorithm for breast cancer study we also focus onthe genes interacted with the estrogen receptor alpha ESR1with GBmiRs including miR-18a miR-18b miR-193b miR-19a miR-19b miR-206 miR-20b miR-22 miR-221 miR-222miR-29b and miR-302c The regulation network generatedby ESR1 as the initial GoI is shown in Figure 7 and the top18 ceRNAs are provided in Table 3 The TraceRNA algorithmcan be accessed httpcompgenomicsutsaeducerna
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
Advances in Bioinformatics 5
MicroRNAs
Modulator
Regulator
Target3998400UTR
3998400UTRmRNA119898
mRNA 119910
Figure 4 Modulation of gene regulation by competing mRNAs
a partition of the paired expression samples into the cor-related and uncorrelated samples The paired expressionsamples that possess such correlated-uncorrelated partition(03 lt 120587 lt 07 and |120572| gt 08) are determined to bemodulated To identify the modulator of a (or a group of)modulated pair(s) a weighted 119905-test was developed to searchfor the genes whose expressions are differentially expressedin the correlated partition versus the uncorrelated partition
34 GEM (Gene Expression Modulator) GEM [46] improvesover MINDy by predicting how a modulator-TF interactionaffects the expression of the target gene It can detect newtypes of interactions that result in stronger correlation butlow Δ119868 which therefore would be missed by MINDy GEMhypothesizes that the correlation between the expression of amodulator 119898 and a target 119909 must change as that of the TF119909 changes Unlike the previous surveyed algorithms GEMfirst transforms the continuous expression levels to binarystates (up- (1) or down-expression (0)) and then works onlywith discrete expression states To model the hypothesizedrelationship the following model is proposed
119875 (119909 = 1 | 119910119898) = 120572119888+ 120572119898119898 + 120572
119910119910 + 120574119898119910 (10)
where 120572119888is a constant 120572
119898and 120572
119910model the effect of
modulator and TF on the target genes and 120574 represents theeffect of modulator-TF interaction on the target gene If themodulator-TF interaction has an effect on 119909 then 120574 willbe nonzero For a given (119909 119910119898) triplets GEM devised analgorithm to estimate the model coefficients in (10) and a testto determine if 120574 is nonzero or119898 is a modulator of 119909 and 119910
35MuTaMe (Mutually TargetedMREEnrichment) Thegoalof MuTaMe [21] is to identify ceRNA networks of a gene ofinterest (GoI) ormRNA that sharemiRNA response elements(MREs) of samemiRNAs Figure 4 shows twomRNAs whereone is the GoIy and the other is a candidate ceRNA ormodulator 119898 In the figure the miRNA represented in colorred has MREs in both mRNA 119910 and mRNA 119898 in this casethe presence of mRNA 119898 will start the competition with 119910for miRNA represented in color red
Thehypothesis ofMuTaMe is thatmRNAs that havemanyof the same MREs can regulate each other by competingfor miRNAs binding The input of this algorithm is a GoIwhich is targeted by a group of miRNAs known to the userThen from a database of predicted MREs for the entiretranscriptome it is possible to obtain the binding sites andits predicted locations in the 31015840UTR for all mRNAsThis datais used to generate scores for each mRNA based on severalfeatures
(a) the number of miRNAs that an mRNA119898 shares withthe GoI 119910
(b) the density of the predicted MREs for the miRNA itfavors the cases in which more MREs are located inshorter distances
(c) the distribution of the MREs for every miRNA itfavors situations in which theMREs tend to be evenlydistributed
(d) the number of MREs predicted to target 119898 it favorssituationswhere eachmiRNA containsmoreMREs in119898
Then each candidate transcript119898will be assigned a scorethat results from multiplying the scores in (a) to (d) Thisscore indicates the likelihood of the candidates to be ceRNAsand will be used to predict ceRNAs
36 Hermes Hermes [20] is an extension of MINDy thatinfers candidate modulators of miRNA activity from expres-sion profiles of genes and miRNAs of the same samplesHermes makes inferences by estimating the MI and CMIHowever different from MINDy (7) Hermes extracts thedependences of this triplet by studying the difference betweenthe CMI of 119909 expression and 119910 expression conditional on theexpression of119898 and theMI of 119909 and 119910 expressions as follows
119868 = 119868 (119909 119910 | 119898) minus 119868 (119909 119910) (11)
These quantities and their associated statistical signif-icance can be computed from collections of expressionof genes with number of samples 250 or greater Hermesexpands MINDy by providing the capacity to identify can-didate modulator genes of miRNAs activity The presenceof these modulators (119898) will affect the relation betweenthe expression of the miRNAs targeting a gene (119909) and theexpression level of this gene (119909)
In summary we surveyed some of the most popularalgorithms for the inference of modulator Additional mod-ulator identification algorithms are summarized in Table 1It is worth noting that the concept of modulator appliesto cases beyond discussed in this paper Such exampleincludes themultilayer integrated regulatorymodel proposedin Yan et al [49] where the top layer of regulators could bealso considered as ldquomodulatorsrdquo
4 Applications to Breast Cancer GeneExpression Data
Algorithms of utilizing modulator concept have been imple-mented in various software packages Here we will discuss
6 Advances in Bioinformatics
Table 1 Gene regulation network and modulator identification methods
Algorithm Features ReferencesARACNE Interaction network constructed via mutual information (MI) [14 42]
Network profiler A varying-coefficient structural equation model (SEM) to represent themodulator-dependent conditional independence between genes [47]
MINDy Gene-pair interaction dependency on modulator candidates by using theconditional mutual information (CMI) [44]
Mimosa Search for modulator by partition samples with a Gaussian mixture model [45]
GEM A probabilistic method for detecting modulators of TFs that affect the expression oftarget gene by using a priori knowledge and gene expression profiles [46]
MuTaMe Based on the hypothesis that shared MREs can regulate mRNAs by competing formicroRNAs binding [21]
Hermes Extension of MINDy to include microRNAs as candidate modulators by using CMIand MI from expression profiles of genes and miRNAs of the same samples [20]
ER120572 modulator Analyzes the interaction between TF and target gene conditioned on a group ofspecific modulator genes via a multiple linear regression [48]
two new applications MEGRA and TraceRNA implementedin-house specifically to utilize the concept of differentialcorrelation coefficients and ceRNAs to construct a modu-lated GRN with a predetermined modulator In the case ofMGERA we chose estrogen receptor ESR1 as the initialstarting point since it is one of the dominant and systemicfactor in breast cancer in the case of TraceRNAwe also chosegene ESR1 and its modulated gene network Preliminaryresults of applications to TCGA breast cancer data arereported in the following 2 sections
41MGERA TheModulatedGeneRegulationAnalysis algo-rithm (MGERA) was designed to explore gene regulationpairs modulated by the modulator 119898 The regulation pairscan be identified by examining the coexpression of two genesbased on Pearson correlation (similar to (7) in the contextof correlation coefficient) Fisher transformation is adoptedto normalize the correlation coefficients biased by samplesizes to obtain equivalent statistical power among data withdifferent sample sizes Statistical significance of differencein the absolute correlation coefficients between two genes istested by the student 119905-test following Fisher transformationFor the gene pairs with significantly different coefficientsbetween two genes active and deactive statuses are iden-tified by examining the modulated gene expression pairs(MGEPs)TheMGEPs are further combined to construct the119898 modulated gene regulation network for a systematic andcomprehensive view of interaction under modulation
To demonstrate the ability of MGERA we set estrogenreceptor (ER) as the modulator and applied the algorithm toTCGA breast cancer expression data [3] which contains 588expression profiles (461 ER+ and 127 ERminus) By using 119875 valuelt001 and the difference in the absolute Pearson correlationcoefficients gt06 as criteria we identified 2324 putativeER+ MGEPs and a highly connected ER+ modulated generegulation network was constructed (Figure 5) The top tengenes with highest connectivity was show in Table 2 Thecysteinetyrosine-rich 1 gene (CYYR1) connected to 142genes was identified as the top hub gene in the network andthus may serve as a key regulator under ER+ modulation
Table 2 Hub genes derived from modulated gene regulationnetwork (Figure 5)
Gene Number of ER+ MGEPsCYYR1 142MRAS 109C9orf19 95LOC339524 93PLEKHG1 92FBLN5 91BOC 91ANKRD35 89FAM107A 83C16orf77 73
Gene Ontology analysis of CYYR1 and its connected neigh-bor genes revealed significant association with extracellularmatrix epithelial tube formation and angiogenesis
42 TraceRNA To identify the regulationnetwork of ceRNAsfor a GoI we developed a web-based application TraceRNApresented earlier in [50] with extension to regulation networkconstructionThe analysis flow chart of TraceRNAwas shownin Figure 6 For a selected GoI the GoI binding miRNAs(GBmiRs) were derived either validated miRNAs from miR-TarBase [51] or predicted miRNAs from SVMicrO [52]ThenmRNAs (other than the given GoI) also targeted by GBmiRswere identified as the candidates of ceRNAs The relevant (ortumor-specific) gene expression data were used to furtherstrengthen relationship between the ceRNA candidates andGoI The candidate ceRNAs which coexpressed with GoIwere reported as putative ceRNAs To construct the generegulation network via GBmiRs we set each ceRNA as thesecondary GoI and the ceRNAs of these secondary GoIswere identified by applying the algorithm recursively Uponidentifying all the ceRNAs the regulation network of ceRNAsof a given GoI was constructed
Advances in Bioinformatics 7
EDG1GIMAP8
ACVR2A
MEOX1P2RY5
ESAM
KLF6
SOX7GSN
PALMDLYVE1
EGFLAM
OLFML2AGIMAP6
CD93
HSPA12B
SH2D3C
ATP8B4
P2RY14
ELMO3
MED29
GUCY1B3
SPNS2
ELTD1
RGS5
GGTA1
FAAH
KIAA1244
HIVEP2
RBP7
CDR1
GIMAP5
CD36ABCA9
LIMS2FOXO1
CYYR1 EBF3
MALLRSU1
FILIP1
KCNJ2 ROBO4
GIMAP1
CFH
LPPR4
C13orf33
C1QTNF7 EMX2OS
C10orf54
CIDEA
VWF
PODNPALLD
CILP
KGFLP1 LRRC32
OMD
ADH1BPLA2G5
C14orf37
WFWBCL6B
CCIN
ADAMTS16
AANAT
KIR3DL1
TPK1
EMILIN2
PLEKHQ1
ACVRL1F13A1
CCL4 PTGDS
BOC
LAYN
TSHZ2
RARRES1
ENPP6
MMP7
KRT6B
DSC1
ARID5BOLFM4
CX3CL1FLJ45803
GABRPDNM3
KCNMB111FAM3D
CPAMD8
SCPEP1PIK3C2G
BCL6
C12orf54
UTRN
IFNGR1IL33
USHBP1
GPBAR1PTH2R
NDRG2
CASQ2MAB21L1
CCL23CCL15CCCCEMP1
CRTAC1
SKIP
TP63
RRAGC
MMRN1
PLEKHF1
GUCY1A3ADCY4
HSD11B1
EBF2
RASD1 UACA
LGI4
ANKRD6
EGR2
RP5-1054A223ELN
GIMAP7
ADIPOQ
TLR4
HAND2
FAM20A
DPT
APDE1B
HHSOCS3
CD248
LCN6
TLL1
RSPO1
CXorf36
RRAD
CRISPLD2
PLB1
DYSF
RASSF222FAM101A
ST5
HSPG2
FGF1
ITGBL1LGI2
DCHS1
PLXDC2
RCN3DACT3SFRP2SPSB1
KCNJ15SRPX2IGFBP7 COL1A2
OR10A5
ACTA2PRICKLE1
XG
FMO1
C16orf30CLEC14A
TCF4C1S
LMAN1L
FBLN1SH3PXD2B
FMO3
MXRA8D22
HTRA1COL8A2
1DACT1
COL16A1
GLT8D2
LMOD1
SPON2
RFTN2
FILIP1L
MFAP5
MMP23BPRRX1 TNFAIP6
LUMCTSKFIBIN
COL6A3MAFB
SFRP4
KERACLDN5RSPO3FGF7
BHLHB3
C16orf77
BST1
SSPN
COL15A1TIE1
ABCC9
CRYABSV2B
LAMA2
SAA1NKAPL
FAM49AANKRD35CHL1
SAA2
SAV1
FBLN5TCEAL7KIAA0355CCDC80
C2orf40
THSD7APRELP
FRZBMMRN2
ALDH2
RASL12SERPING1
TAGLNLTBP2
CRB3
EPN3
TPD52
AP1M2
KIAA1543
TMC4
TJP3
MB
CDH23ZFP36
CLIC5
KL
FHL5
CCDC24
FAM86A
AVPR1ASTX12
FABP4
PDZRN3
ADH1CPPP1R12B
LASS6
SCARF1
LOC338328
ANKRD47SPARCL1SNOTCH4
JPH2TCF7L2
TM4SF18HECA
BSPRY
RASEF
SHROOM3
PVRL4
SPINT1
PROM2
SH2D3A
CLDN3
FAAH2
RAB25
LOC124220
RBM35A
C2orf15
PKP3
MARVELD3
FAM84B
VIL2
ELF3TACSTD1
PRSS8 C1orf172
C1orf34
OVOL1
ARL6IP1
MAGIX
CDS1
BIK
FXYD3
GSTO2
IGSF9
CAPN13
TRPS1
OCLN
DNASE2
C1orf210
P2RX4
LSR
C9orf7
EPHA10
AVAVAVAVFAM83H
KIAA1688
LRRC8E
FMOD
MMP11 RUNDC3B
CCDC107CLDN7
KRT19
CBLCCGN
C19orf46SLC44A4
KIAA1324
PRRG2
DDR1
NEBL
TMEM125
ATAD4
LRFN2
MARVELD2
IRX5
CLDN4
LRBA
SPDEF
LOC400451
SPINT2
DLG3
CREB3L4GRHL2
TFAP2A
PSD4
IRF6
RGS13
C1orf186
HDC
CPA3
AQP2
LCE1F
CYP11B1
GRAMD2
CLDN8
GPRIN2
TNN
PDGFD
GIMAP4
GJA12
SHANK3
MAGED2
PTCH2
KLK5PIGR
C4orf7
LGR6
SLC34A2
SELP
APOD
KIF19
PACRG
ZNF446
ZNF574
CREB3HHAT
PEX11B
C10orf81
KRT7
PSEN2
SLC25A44
SEC16A
CRHR1CCDC114
SPINK7
LCE3ETMEM95
OTOF
OR4D1
SSTR3
KLK9
OR10H2
ATN1
CDK5R2FLJ32214
KRT3
C1orf175
PROP1GHSR
PLA2G4D
TRGV5
TIGD5
VPS28
KIAA0143
DPY19L4
POLR2K
COMMD5
DYRK1B
PSORS1C2OR5H1
C10orf27
KRTAP11-1
LOC652968
OR2B11
OR5BF1
OR10P1
MGC40574
P2RX2 RASGRP4
Gene
Positive correlation under ER+ modulationNegative correlation under ER+ modulation
Figure 5 ER+ modulated gene regulation network
To identify ceRNA candidates three miRNAs bindingprediction algorithms SiteTest SVMicrO and BCMicrOwere used in TracRNA SiteTest is an algorithm similar toMuTaMe and uses UTR features for target prediction SVMi-crO [52] is an algorithm that uses a large number of sequence-level site as well as UTR features including binding secondarystructure energy and conservation whereas BCMicrO [53]employs a Bayesian approach that integrates predictions from6 popular algorithms including TargetScan miRanda PicTarmirTarget PITA and DIANA-microT Pearson correlationcoefficient was used to test the coexpression between the GoIand the candidate ceRNAs We utilized TCGA breast cancercohort [3] as the expression data by using 60 of GBmiRs
as commonmiRNAs and Pearson correlation coefficient gt09as criteria The final scores of putative ceRNAs (see Table 3last column) were generated by using Bordamergingmethodwhich rerank the sum of ranks from both GBmiR bindingand coexpression 119875 values [54] To illustrate the utility of theTraceRNA algorithm for breast cancer study we also focus onthe genes interacted with the estrogen receptor alpha ESR1with GBmiRs including miR-18a miR-18b miR-193b miR-19a miR-19b miR-206 miR-20b miR-22 miR-221 miR-222miR-29b and miR-302c The regulation network generatedby ESR1 as the initial GoI is shown in Figure 7 and the top18 ceRNAs are provided in Table 3 The TraceRNA algorithmcan be accessed httpcompgenomicsutsaeducerna
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
6 Advances in Bioinformatics
Table 1 Gene regulation network and modulator identification methods
Algorithm Features ReferencesARACNE Interaction network constructed via mutual information (MI) [14 42]
Network profiler A varying-coefficient structural equation model (SEM) to represent themodulator-dependent conditional independence between genes [47]
MINDy Gene-pair interaction dependency on modulator candidates by using theconditional mutual information (CMI) [44]
Mimosa Search for modulator by partition samples with a Gaussian mixture model [45]
GEM A probabilistic method for detecting modulators of TFs that affect the expression oftarget gene by using a priori knowledge and gene expression profiles [46]
MuTaMe Based on the hypothesis that shared MREs can regulate mRNAs by competing formicroRNAs binding [21]
Hermes Extension of MINDy to include microRNAs as candidate modulators by using CMIand MI from expression profiles of genes and miRNAs of the same samples [20]
ER120572 modulator Analyzes the interaction between TF and target gene conditioned on a group ofspecific modulator genes via a multiple linear regression [48]
two new applications MEGRA and TraceRNA implementedin-house specifically to utilize the concept of differentialcorrelation coefficients and ceRNAs to construct a modu-lated GRN with a predetermined modulator In the case ofMGERA we chose estrogen receptor ESR1 as the initialstarting point since it is one of the dominant and systemicfactor in breast cancer in the case of TraceRNAwe also chosegene ESR1 and its modulated gene network Preliminaryresults of applications to TCGA breast cancer data arereported in the following 2 sections
41MGERA TheModulatedGeneRegulationAnalysis algo-rithm (MGERA) was designed to explore gene regulationpairs modulated by the modulator 119898 The regulation pairscan be identified by examining the coexpression of two genesbased on Pearson correlation (similar to (7) in the contextof correlation coefficient) Fisher transformation is adoptedto normalize the correlation coefficients biased by samplesizes to obtain equivalent statistical power among data withdifferent sample sizes Statistical significance of differencein the absolute correlation coefficients between two genes istested by the student 119905-test following Fisher transformationFor the gene pairs with significantly different coefficientsbetween two genes active and deactive statuses are iden-tified by examining the modulated gene expression pairs(MGEPs)TheMGEPs are further combined to construct the119898 modulated gene regulation network for a systematic andcomprehensive view of interaction under modulation
To demonstrate the ability of MGERA we set estrogenreceptor (ER) as the modulator and applied the algorithm toTCGA breast cancer expression data [3] which contains 588expression profiles (461 ER+ and 127 ERminus) By using 119875 valuelt001 and the difference in the absolute Pearson correlationcoefficients gt06 as criteria we identified 2324 putativeER+ MGEPs and a highly connected ER+ modulated generegulation network was constructed (Figure 5) The top tengenes with highest connectivity was show in Table 2 Thecysteinetyrosine-rich 1 gene (CYYR1) connected to 142genes was identified as the top hub gene in the network andthus may serve as a key regulator under ER+ modulation
Table 2 Hub genes derived from modulated gene regulationnetwork (Figure 5)
Gene Number of ER+ MGEPsCYYR1 142MRAS 109C9orf19 95LOC339524 93PLEKHG1 92FBLN5 91BOC 91ANKRD35 89FAM107A 83C16orf77 73
Gene Ontology analysis of CYYR1 and its connected neigh-bor genes revealed significant association with extracellularmatrix epithelial tube formation and angiogenesis
42 TraceRNA To identify the regulationnetwork of ceRNAsfor a GoI we developed a web-based application TraceRNApresented earlier in [50] with extension to regulation networkconstructionThe analysis flow chart of TraceRNAwas shownin Figure 6 For a selected GoI the GoI binding miRNAs(GBmiRs) were derived either validated miRNAs from miR-TarBase [51] or predicted miRNAs from SVMicrO [52]ThenmRNAs (other than the given GoI) also targeted by GBmiRswere identified as the candidates of ceRNAs The relevant (ortumor-specific) gene expression data were used to furtherstrengthen relationship between the ceRNA candidates andGoI The candidate ceRNAs which coexpressed with GoIwere reported as putative ceRNAs To construct the generegulation network via GBmiRs we set each ceRNA as thesecondary GoI and the ceRNAs of these secondary GoIswere identified by applying the algorithm recursively Uponidentifying all the ceRNAs the regulation network of ceRNAsof a given GoI was constructed
Advances in Bioinformatics 7
EDG1GIMAP8
ACVR2A
MEOX1P2RY5
ESAM
KLF6
SOX7GSN
PALMDLYVE1
EGFLAM
OLFML2AGIMAP6
CD93
HSPA12B
SH2D3C
ATP8B4
P2RY14
ELMO3
MED29
GUCY1B3
SPNS2
ELTD1
RGS5
GGTA1
FAAH
KIAA1244
HIVEP2
RBP7
CDR1
GIMAP5
CD36ABCA9
LIMS2FOXO1
CYYR1 EBF3
MALLRSU1
FILIP1
KCNJ2 ROBO4
GIMAP1
CFH
LPPR4
C13orf33
C1QTNF7 EMX2OS
C10orf54
CIDEA
VWF
PODNPALLD
CILP
KGFLP1 LRRC32
OMD
ADH1BPLA2G5
C14orf37
WFWBCL6B
CCIN
ADAMTS16
AANAT
KIR3DL1
TPK1
EMILIN2
PLEKHQ1
ACVRL1F13A1
CCL4 PTGDS
BOC
LAYN
TSHZ2
RARRES1
ENPP6
MMP7
KRT6B
DSC1
ARID5BOLFM4
CX3CL1FLJ45803
GABRPDNM3
KCNMB111FAM3D
CPAMD8
SCPEP1PIK3C2G
BCL6
C12orf54
UTRN
IFNGR1IL33
USHBP1
GPBAR1PTH2R
NDRG2
CASQ2MAB21L1
CCL23CCL15CCCCEMP1
CRTAC1
SKIP
TP63
RRAGC
MMRN1
PLEKHF1
GUCY1A3ADCY4
HSD11B1
EBF2
RASD1 UACA
LGI4
ANKRD6
EGR2
RP5-1054A223ELN
GIMAP7
ADIPOQ
TLR4
HAND2
FAM20A
DPT
APDE1B
HHSOCS3
CD248
LCN6
TLL1
RSPO1
CXorf36
RRAD
CRISPLD2
PLB1
DYSF
RASSF222FAM101A
ST5
HSPG2
FGF1
ITGBL1LGI2
DCHS1
PLXDC2
RCN3DACT3SFRP2SPSB1
KCNJ15SRPX2IGFBP7 COL1A2
OR10A5
ACTA2PRICKLE1
XG
FMO1
C16orf30CLEC14A
TCF4C1S
LMAN1L
FBLN1SH3PXD2B
FMO3
MXRA8D22
HTRA1COL8A2
1DACT1
COL16A1
GLT8D2
LMOD1
SPON2
RFTN2
FILIP1L
MFAP5
MMP23BPRRX1 TNFAIP6
LUMCTSKFIBIN
COL6A3MAFB
SFRP4
KERACLDN5RSPO3FGF7
BHLHB3
C16orf77
BST1
SSPN
COL15A1TIE1
ABCC9
CRYABSV2B
LAMA2
SAA1NKAPL
FAM49AANKRD35CHL1
SAA2
SAV1
FBLN5TCEAL7KIAA0355CCDC80
C2orf40
THSD7APRELP
FRZBMMRN2
ALDH2
RASL12SERPING1
TAGLNLTBP2
CRB3
EPN3
TPD52
AP1M2
KIAA1543
TMC4
TJP3
MB
CDH23ZFP36
CLIC5
KL
FHL5
CCDC24
FAM86A
AVPR1ASTX12
FABP4
PDZRN3
ADH1CPPP1R12B
LASS6
SCARF1
LOC338328
ANKRD47SPARCL1SNOTCH4
JPH2TCF7L2
TM4SF18HECA
BSPRY
RASEF
SHROOM3
PVRL4
SPINT1
PROM2
SH2D3A
CLDN3
FAAH2
RAB25
LOC124220
RBM35A
C2orf15
PKP3
MARVELD3
FAM84B
VIL2
ELF3TACSTD1
PRSS8 C1orf172
C1orf34
OVOL1
ARL6IP1
MAGIX
CDS1
BIK
FXYD3
GSTO2
IGSF9
CAPN13
TRPS1
OCLN
DNASE2
C1orf210
P2RX4
LSR
C9orf7
EPHA10
AVAVAVAVFAM83H
KIAA1688
LRRC8E
FMOD
MMP11 RUNDC3B
CCDC107CLDN7
KRT19
CBLCCGN
C19orf46SLC44A4
KIAA1324
PRRG2
DDR1
NEBL
TMEM125
ATAD4
LRFN2
MARVELD2
IRX5
CLDN4
LRBA
SPDEF
LOC400451
SPINT2
DLG3
CREB3L4GRHL2
TFAP2A
PSD4
IRF6
RGS13
C1orf186
HDC
CPA3
AQP2
LCE1F
CYP11B1
GRAMD2
CLDN8
GPRIN2
TNN
PDGFD
GIMAP4
GJA12
SHANK3
MAGED2
PTCH2
KLK5PIGR
C4orf7
LGR6
SLC34A2
SELP
APOD
KIF19
PACRG
ZNF446
ZNF574
CREB3HHAT
PEX11B
C10orf81
KRT7
PSEN2
SLC25A44
SEC16A
CRHR1CCDC114
SPINK7
LCE3ETMEM95
OTOF
OR4D1
SSTR3
KLK9
OR10H2
ATN1
CDK5R2FLJ32214
KRT3
C1orf175
PROP1GHSR
PLA2G4D
TRGV5
TIGD5
VPS28
KIAA0143
DPY19L4
POLR2K
COMMD5
DYRK1B
PSORS1C2OR5H1
C10orf27
KRTAP11-1
LOC652968
OR2B11
OR5BF1
OR10P1
MGC40574
P2RX2 RASGRP4
Gene
Positive correlation under ER+ modulationNegative correlation under ER+ modulation
Figure 5 ER+ modulated gene regulation network
To identify ceRNA candidates three miRNAs bindingprediction algorithms SiteTest SVMicrO and BCMicrOwere used in TracRNA SiteTest is an algorithm similar toMuTaMe and uses UTR features for target prediction SVMi-crO [52] is an algorithm that uses a large number of sequence-level site as well as UTR features including binding secondarystructure energy and conservation whereas BCMicrO [53]employs a Bayesian approach that integrates predictions from6 popular algorithms including TargetScan miRanda PicTarmirTarget PITA and DIANA-microT Pearson correlationcoefficient was used to test the coexpression between the GoIand the candidate ceRNAs We utilized TCGA breast cancercohort [3] as the expression data by using 60 of GBmiRs
as commonmiRNAs and Pearson correlation coefficient gt09as criteria The final scores of putative ceRNAs (see Table 3last column) were generated by using Bordamergingmethodwhich rerank the sum of ranks from both GBmiR bindingand coexpression 119875 values [54] To illustrate the utility of theTraceRNA algorithm for breast cancer study we also focus onthe genes interacted with the estrogen receptor alpha ESR1with GBmiRs including miR-18a miR-18b miR-193b miR-19a miR-19b miR-206 miR-20b miR-22 miR-221 miR-222miR-29b and miR-302c The regulation network generatedby ESR1 as the initial GoI is shown in Figure 7 and the top18 ceRNAs are provided in Table 3 The TraceRNA algorithmcan be accessed httpcompgenomicsutsaeducerna
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
Advances in Bioinformatics 7
EDG1GIMAP8
ACVR2A
MEOX1P2RY5
ESAM
KLF6
SOX7GSN
PALMDLYVE1
EGFLAM
OLFML2AGIMAP6
CD93
HSPA12B
SH2D3C
ATP8B4
P2RY14
ELMO3
MED29
GUCY1B3
SPNS2
ELTD1
RGS5
GGTA1
FAAH
KIAA1244
HIVEP2
RBP7
CDR1
GIMAP5
CD36ABCA9
LIMS2FOXO1
CYYR1 EBF3
MALLRSU1
FILIP1
KCNJ2 ROBO4
GIMAP1
CFH
LPPR4
C13orf33
C1QTNF7 EMX2OS
C10orf54
CIDEA
VWF
PODNPALLD
CILP
KGFLP1 LRRC32
OMD
ADH1BPLA2G5
C14orf37
WFWBCL6B
CCIN
ADAMTS16
AANAT
KIR3DL1
TPK1
EMILIN2
PLEKHQ1
ACVRL1F13A1
CCL4 PTGDS
BOC
LAYN
TSHZ2
RARRES1
ENPP6
MMP7
KRT6B
DSC1
ARID5BOLFM4
CX3CL1FLJ45803
GABRPDNM3
KCNMB111FAM3D
CPAMD8
SCPEP1PIK3C2G
BCL6
C12orf54
UTRN
IFNGR1IL33
USHBP1
GPBAR1PTH2R
NDRG2
CASQ2MAB21L1
CCL23CCL15CCCCEMP1
CRTAC1
SKIP
TP63
RRAGC
MMRN1
PLEKHF1
GUCY1A3ADCY4
HSD11B1
EBF2
RASD1 UACA
LGI4
ANKRD6
EGR2
RP5-1054A223ELN
GIMAP7
ADIPOQ
TLR4
HAND2
FAM20A
DPT
APDE1B
HHSOCS3
CD248
LCN6
TLL1
RSPO1
CXorf36
RRAD
CRISPLD2
PLB1
DYSF
RASSF222FAM101A
ST5
HSPG2
FGF1
ITGBL1LGI2
DCHS1
PLXDC2
RCN3DACT3SFRP2SPSB1
KCNJ15SRPX2IGFBP7 COL1A2
OR10A5
ACTA2PRICKLE1
XG
FMO1
C16orf30CLEC14A
TCF4C1S
LMAN1L
FBLN1SH3PXD2B
FMO3
MXRA8D22
HTRA1COL8A2
1DACT1
COL16A1
GLT8D2
LMOD1
SPON2
RFTN2
FILIP1L
MFAP5
MMP23BPRRX1 TNFAIP6
LUMCTSKFIBIN
COL6A3MAFB
SFRP4
KERACLDN5RSPO3FGF7
BHLHB3
C16orf77
BST1
SSPN
COL15A1TIE1
ABCC9
CRYABSV2B
LAMA2
SAA1NKAPL
FAM49AANKRD35CHL1
SAA2
SAV1
FBLN5TCEAL7KIAA0355CCDC80
C2orf40
THSD7APRELP
FRZBMMRN2
ALDH2
RASL12SERPING1
TAGLNLTBP2
CRB3
EPN3
TPD52
AP1M2
KIAA1543
TMC4
TJP3
MB
CDH23ZFP36
CLIC5
KL
FHL5
CCDC24
FAM86A
AVPR1ASTX12
FABP4
PDZRN3
ADH1CPPP1R12B
LASS6
SCARF1
LOC338328
ANKRD47SPARCL1SNOTCH4
JPH2TCF7L2
TM4SF18HECA
BSPRY
RASEF
SHROOM3
PVRL4
SPINT1
PROM2
SH2D3A
CLDN3
FAAH2
RAB25
LOC124220
RBM35A
C2orf15
PKP3
MARVELD3
FAM84B
VIL2
ELF3TACSTD1
PRSS8 C1orf172
C1orf34
OVOL1
ARL6IP1
MAGIX
CDS1
BIK
FXYD3
GSTO2
IGSF9
CAPN13
TRPS1
OCLN
DNASE2
C1orf210
P2RX4
LSR
C9orf7
EPHA10
AVAVAVAVFAM83H
KIAA1688
LRRC8E
FMOD
MMP11 RUNDC3B
CCDC107CLDN7
KRT19
CBLCCGN
C19orf46SLC44A4
KIAA1324
PRRG2
DDR1
NEBL
TMEM125
ATAD4
LRFN2
MARVELD2
IRX5
CLDN4
LRBA
SPDEF
LOC400451
SPINT2
DLG3
CREB3L4GRHL2
TFAP2A
PSD4
IRF6
RGS13
C1orf186
HDC
CPA3
AQP2
LCE1F
CYP11B1
GRAMD2
CLDN8
GPRIN2
TNN
PDGFD
GIMAP4
GJA12
SHANK3
MAGED2
PTCH2
KLK5PIGR
C4orf7
LGR6
SLC34A2
SELP
APOD
KIF19
PACRG
ZNF446
ZNF574
CREB3HHAT
PEX11B
C10orf81
KRT7
PSEN2
SLC25A44
SEC16A
CRHR1CCDC114
SPINK7
LCE3ETMEM95
OTOF
OR4D1
SSTR3
KLK9
OR10H2
ATN1
CDK5R2FLJ32214
KRT3
C1orf175
PROP1GHSR
PLA2G4D
TRGV5
TIGD5
VPS28
KIAA0143
DPY19L4
POLR2K
COMMD5
DYRK1B
PSORS1C2OR5H1
C10orf27
KRTAP11-1
LOC652968
OR2B11
OR5BF1
OR10P1
MGC40574
P2RX2 RASGRP4
Gene
Positive correlation under ER+ modulationNegative correlation under ER+ modulation
Figure 5 ER+ modulated gene regulation network
To identify ceRNA candidates three miRNAs bindingprediction algorithms SiteTest SVMicrO and BCMicrOwere used in TracRNA SiteTest is an algorithm similar toMuTaMe and uses UTR features for target prediction SVMi-crO [52] is an algorithm that uses a large number of sequence-level site as well as UTR features including binding secondarystructure energy and conservation whereas BCMicrO [53]employs a Bayesian approach that integrates predictions from6 popular algorithms including TargetScan miRanda PicTarmirTarget PITA and DIANA-microT Pearson correlationcoefficient was used to test the coexpression between the GoIand the candidate ceRNAs We utilized TCGA breast cancercohort [3] as the expression data by using 60 of GBmiRs
as commonmiRNAs and Pearson correlation coefficient gt09as criteria The final scores of putative ceRNAs (see Table 3last column) were generated by using Bordamergingmethodwhich rerank the sum of ranks from both GBmiR bindingand coexpression 119875 values [54] To illustrate the utility of theTraceRNA algorithm for breast cancer study we also focus onthe genes interacted with the estrogen receptor alpha ESR1with GBmiRs including miR-18a miR-18b miR-193b miR-19a miR-19b miR-206 miR-20b miR-22 miR-221 miR-222miR-29b and miR-302c The regulation network generatedby ESR1 as the initial GoI is shown in Figure 7 and the top18 ceRNAs are provided in Table 3 The TraceRNA algorithmcan be accessed httpcompgenomicsutsaeducerna
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
8 Advances in Bioinformatics
The GoI binding miRNA (GBmiRs)
Candidates of ceRNAs
Validated miRNAs from miRTarBaseor predicted by SVMicrO
Gene of interest (GoI)
Putative ceRNAs
Identify mRNAs targeted by the same GBmiRs
Identify ceRNAs with coexpressed patterns
Secondary Gol
Select each ceRNA as a secondary GoI
Regulation network of ceRNAs
Output to TraceRNA website
Figure 6 The analysis flow chart of TraceRNA
CADM1
PAPOLAPTPRD
RAB5B
KIAA1522
SENP1
ZC3H11A
RAP2C
KCNE4
DAG1
KIAA0467CXCL14
MBNL1GFAP
CBFBDAXX
TCF4
FOXA1
KBTBD4
WDFY3
ARID4ARAP1GDS1
LEF1PPP3CB
ZFX
RYBPG3BP2
VAV3
FNIP1SMG1
PKN2
PHTF2TAOK1OTUD4
ZNF654 MED13ZNF800
GRM8PAFAH1B1
SRGAP3POMGNT1
GRIA2FAM120A
BMS1P5
HNRPDLGMFBERBB2IPMAP2K4
TMEM57
COL12A1 COPB2C5orf24
C14orf101PCDHAC1
PCDHAC2C1orf9
TEX2
FNBP1L
BCL2L1
PCDHA1PCDHA5
NAV3
BAZ2BTHBS1
MAP3K1 TP53INP1MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
PCDHA3
CRIM1
PAN3CACNA2D2
RAB8B
PPM1B
JAZF1C20orf194
RAPGEF2
FEM1CPCDHA9SFRS12
TGFBR3
ARRDC3
RPGRIP1L
KIAA0240
MAT2AGLCE
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
BACE1
ANTXR2
FAM70A FLRT3
CHD9SLC12A2
CSNK1G3DCUN1D4
ATPBD4
PRPF38BUBE3ARBM12
FMR1
RND3PAK7
CIT
C18orf25MUTED
MEIS2
ZNF608CLK2P
C7orf60BRMS1L
ZDHHC17
CTNND2
ACSL1FUSIP1
PIK3C2A
FAM135A
NR3C2
USP15
PTGFRNC5orf13VEZF1
RAPGEF5
THSD3
SOBP
TMEM11
MBNL2
NARG1NRXN1
NCOA2PPP2CA
MAPRE1
SLC24A3
ZADH2
PPFIA2
TMEM115HOMER1
B4GALT2
FOXP1
GH2
PRM1PML
GGT3P
DMRT2
DST
LOC442245MAP1LC3A
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1 SNORD8
ZNF238
GIT1
GOSR1CPEB4
PPP1R12A
TNRC6CCSH1
PITPNBCOL4A3BP
DLG2ATP2B1
ZNF292
RICTORRUNX2
BTRC
ARX
STK35CYB561D1
PIP4K2A
SLC6A10PLYCAT
CD2BP2
PUM2
RERERSBN1 PDS5B
CPEB3VEGFA
CUL3RAP1B
CAMTA1
SIRT1DCUN1D3
KCNJ2
TSHZ3
DTNA
CCNT2
TARDBP
ZFHX3
CNOT6NOVA1SEMA6A
BAGE5
ATRX
RCAN2KIAA0232
EBF1DNAJA2
SATB2LEMD3
C10orf26
LCOR
ZBTB4
ARHGEF17ST7LQKI
SLC12A5
MMP24
REEP2
WDR47DIXDC1
MEF2C
PRKG1LHX6
ARL4AANK2
NRP2
SCN2A
ZFHX4
ABCD4
MAGI2MAPT
PPP6C
HEYL
CFL2
ABHD13
JAG1
USP47
NBEA
SPTY2D1
BSN NHS
BZRAP1
MAPKBP1
SEMA6D
YPEL5
PHF15 DDX3X
CAPN3
NOPE
SLC38A2
UBL3MAGI1
ZSWIM6 CADPS
SMARCAD1
ESR1
(a)
KB
WDFY3
ZFX
G
PAFAH1B1POMGN
FAM120A
ERBB2MAP2K4
NAV3
BAZ2BTHBS1
MYST4
KLF9
MED13L
RNF11
ATP7ANFIA
CRIM1
PAN3CACNA2D2
RAB8B
KIAA0240
MAT2AGLCE
ESR1
FNDC3AREEP1
RANBP6
TNRC6A
CPEB2
ESRRG
ZFPM2
SOCS5
ELL2
ARPP-19
ZNF516
RAB14
HIPK1
PTEN MIER3ZFP91
NPAS3 ALCAM
DACH1
GOSR1TNRC
ZNF292
BT
PUM2RSBN1
CPEB3VEGFA
CUL3
RAP1B
SIRT1DCUN1D3 TSHZ3
DTNA
TARDBP
ZFHX3
CNOT6 NOVA1
SEMA6A
BAGE5
ATRX
F2C
HS
(b)
Figure 7 (a) ceRNA network for gene of interest ESR1 generated using TraceRNA (b) Network graph enlarged at ESR1
5 Conclusions
In this report we attempt to provide a unified concept ofmodulation of gene regulation encompassing earlier mRNAexpression based methods and lately the ceRNAmethod Weexpect the integration of ceRNA concept into the gene-geneinteractions and their modulator identification will further
enhance our understanding in gene interaction and theirsystemic influence Applications provided here also representexamples of our earlier attempt to construct modulated net-works specific to breast cancer studies Further investigationwill be carried out to extend our modeling to provide aunified understanding of genetic regulation in an alteredenvironment
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
Advances in Bioinformatics 9
Table 3 Top 18 candidate ceRNAs for ESR1 as GOI obtained from TraceRNA ESR1 is at rank of 174 (not listed in this table)
Gene symbol SVMicrO-based prediction Expression correlationFinal score
Score 119875 value Score 119875 valueFOXP1 1066 00043 0508 0016 1212VEZF1 0942 00060 04868 0020 1179NOVA1 0896 00067 0479 0023 1160CPEB3 0858 00074 0484 0022 1149MAP2K4 0919 00064 0322 0097 1139FAM120A 0885 00069 0341 0082 1130PCDHA3 0983 00054 0170 0215 1125SIRT1 0927 00062 0230 0162 1117PCDHA5 0983 00054 0148 0233 1113PTEN 0898 00067 0221 0168 1104PCDHA1 0983 00054 0140 0239 1103NBEA 0752 00098 0491 0020 1102ZFHX4 0970 00056 0154 0229 1097GLCE 0798 00087 03231 0096 1096MAGI2 0777 00092 0321 0097 1086SATB2 0801 00086 0243 0151 1078LEF1 0753 00098 0291 0112 1065ATPBD4 0819 00082 0170 0215 1060
Authorsrsquo Contribution
MFlores andT-HHsiao are contributed equally to thiswork
Acknowledgments
The authors would like to thank the funding support ofthis work by Qatar National Research Foundation (NPRP09 -874-3-235) to Y Chen and Y Huang National ScienceFoundation (CCF-1246073) to Y Huang The authors alsothank the computational support provided by the UTSAComputational Systems Biology Core Facility (NIH RCMI5G12RR013646-12)
References
[1] M Schena D Shalon R W Davis and P O Brown ldquoQuantita-tive monitoring of gene expression patterns with a complemen-tary DNA microarrayrdquo Science vol 270 no 5235 pp 467ndash4701995
[2] E R Mardis ldquoNext-generation DNA sequencing methodsrdquoAnnual Review of Genomics and Human Genetics vol 9 pp387ndash402 2008
[3] Cancer Genome Atlas Network ldquoComprehensive molecularportraits of human breast tumoursrdquoNature vol 490 pp 61ndash702012
[4] D Bell A Berchuck M Birrer et al ldquoIntegrated genomicanalyses of ovarian carcinomardquo Nature vol 474 no 7353 pp609ndash615 2011
[5] R McLendon A Friedman D Bigner et al ldquoComprehensivegenomic characterization defines human glioblastoma genesand core pathwaysrdquo Nature vol 455 no 7216 pp 1061ndash10682008
[6] C M Perou T Soslashrile M B Eisen et al ldquoMolecular portraits ofhuman breast tumoursrdquoNature vol 406 no 6797 pp 747ndash7522000
[7] J Lapointe C Li J P Higgins et al ldquoGene expression profilingidentifies clinically relevant subtypes of prostate cancerrdquo Pro-ceedings of the National Academy of Sciences of the United Statesof America vol 101 no 3 pp 811ndash816 2004
[8] M B Eisen P T Spellman P O Brown and D Botstein ldquoClus-ter analysis and display of genome-wide expression patternsrdquoProceedings of the National Academy of Sciences of the UnitedStates of America vol 95 no 25 pp 14863ndash14868 1998
[9] T Schlitt and A Brazma ldquoCurrent approaches to gene regula-tory network modellingrdquo BMC Bioinformatics vol 8 supple-ment 6 article S9 2007
[10] H Hache H Lehrach and R Herwig ldquoReverse engineering ofgene regulatory networks a comparative studyrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2009 Article ID617281 2009
[11] W P Lee and W S Tzou ldquoComputational methods for dis-covering gene networks from expression datardquo Briefings inBioinformatics vol 10 no 4 pp 408ndash423 2009
[12] C Sima J Hua and S Jung ldquoInference of gene regulatorynetworks using time-series data a surveyrdquo Current Genomicsvol 10 no 6 pp 416ndash429 2009
[13] J M Stuart E Segal D Koller and S K Kim ldquoA gene-coexpression network for global discovery of conserved geneticmodulesrdquo Science vol 302 no 5643 pp 249ndash255 2003
[14] A A Margolin I Nemenman K Basso et al ldquoARACNE analgorithm for the reconstruction of gene regulatory networksin a mammalian cellular contextrdquo BMC Bioinformatics vol 7supplement 1 article S7 2006
[15] E R Dougherty S Kim and Y Chen ldquoCoefficient of determi-nation in nonlinear signal processingrdquo Signal Processing vol 80no 10 pp 2219ndash2235 2000
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
10 Advances in Bioinformatics
[16] S Kim E R Dougherty Y Chen et al ldquoMultivariate measure-ment of gene expression relationshipsrdquo Genomics vol 67 no 2pp 201ndash209 2000
[17] X Chen M Chen and K Ning ldquoBNArray an R package forconstructing gene regulatory networks from microarray databy using Bayesian networkrdquo Bioinformatics vol 22 no 23 pp2952ndash2954 2006
[18] A V Werhli M Grzegorczyk and D Husmeier ldquoCompar-ative evaluation of reverse engineering gene regulatory net-works with relevance networks graphical gaussian models andbayesian networksrdquo Bioinformatics vol 22 no 20 pp 2523ndash2531 2006
[19] I Shmulevich E R Dougherty S Kim and W Zhang ldquoProb-abilistic Boolean networks a rule-based uncertainty model forgene regulatory networksrdquoBioinformatics vol 18 no 2 pp 261ndash274 2002
[20] P Sumazin X Yang H-S Chiu et al ldquoAn extensiveMicroRNA-mediated network of RNA-RNA interactions regulates estab-lished oncogenic pathways in glioblastomardquo Cell vol 147 no2 pp 370ndash381 2011
[21] Y Tay L Kats L Salmena et al ldquoCoding-independent regula-tion of the tumor suppressor PTEN by competing endogenousmRNAsrdquo Cell vol 147 no 2 pp 344ndash357 2011
[22] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009
[23] D Yue J Meng M Lu C L P Chen M Guo and Y HuangldquoUnderstanding MicroRNA regulation a computational per-spectiverdquo IEEE Signal ProcessingMagazine vol 29 no 1 ArticleID 6105465 pp 77ndash88 2012
[24] MW Jones-Rhoades and D P Bartel ldquoComputational identifi-cation of plant MicroRNAs and their targets including a stress-induced miRNArdquo Molecular Cell vol 14 no 6 pp 787ndash7992004
[25] DHanahan andRAWeinberg ldquoThehallmarks of cancerrdquoCellvol 100 no 1 pp 57ndash70 2000
[26] S Y Chun C Johnson J G Washburn M R Cruz-CorreaD T Dang and L H Dang ldquoOncogenic KRAS modulatesmitochondrial metabolism in human colon cancer cells byinducing HIF-1120572 and HIF-2120572 target genesrdquo Molecular Cancervol 9 article 293 2010
[27] N J Hudson A Reverter and B P Dalrymple ldquoA differentialwiring analysis of expression data correctly identifies the genecontaining the causal mutationrdquo PLoS Computational Biologyvol 5 no 5 Article ID e1000382 2009
[28] I Stelniec-Klotz S Legewie O Tchernitsa et al ldquoReverseengineering a hierarchical regulatory network downstream ofoncogenic KRASrdquoMolecular Systems Biology vol 8 Article ID601 2012
[29] C Shen Y Huang Y Liu et al ldquoA modulated empiricalBayes model for identifying topological and temporal estrogenreceptor 120572 regulatory networks in breast cancerrdquo BMC SystemsBiology vol 5 article 67 2011
[30] C A Wilson and J Dering ldquoRecent translational researchmicroarray expression profiling of breast cancer Beyond classi-fication and prognostic markersrdquo Breast Cancer Research vol6 no 5 pp 192ndash200 2004
[31] H E Cunliffe M Ringner S Bilke et al ldquoThe gene expressionresponse of breast cancer to growth regulators patterns andcorrelation with tumor expression profilesrdquo Cancer Researchvol 63 no 21 pp 7158ndash7166 2003
[32] J Frasor F Stossi J M Danes B Komm C R Lyttle andB S Katzenellenbogen ldquoSelective estrogen receptor modula-tors discrimination of agonistic versus antagonistic activitiesby gene expression profiling in breast cancer cellsrdquo CancerResearch vol 64 no 4 pp 1522ndash1533 2004
[33] L J vanrsquot VeerHDaiM J van deVijver et al ldquoGene expressionprofiling predicts clinical outcome of breast cancerrdquoNature vol415 no 6871 pp 530ndash536 2002
[34] S A Kauffman The Origins of Order Self-Organization andSelection in Evolution Oxford University Press New York NYUSA 1993
[35] J D Allen Y Xie M Chen L Girard and G Xiao ldquoComparingstatistical methods for constructing large scale gene networksrdquoPLoS ONE vol 7 no 1 Article ID e29348 2012
[36] Y Huang I M Tienda-Luna and Y Wang ldquoReverse engineer-ing gene regulatory networks a survey of statistical modelsrdquoIEEE Signal Processing Magazine vol 26 no 1 pp 76ndash97 2009
[37] F Crick ldquoCentral dogma of molecular biologyrdquoNature vol 227no 5258 pp 561ndash563 1970
[38] A Hamilton and M Piccart ldquoThe contribution of molecularmarkers to the prediction of response in the treatment of breastcancer a review of the literature on HER-2 p53 and BCL-2rdquoAnnals of Oncology vol 11 no 6 pp 647ndash663 2000
[39] C Sotiriou S Y Neo L M McShane et al ldquoBreast cancerclassification and prognosis based on gene expression profilesfrom a population-based studyrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 100 no18 pp 10393ndash10398 2003
[40] T Soslashrlie C M Perou R Tibshirani et al ldquoGene expressionpatterns of breast carcinomas distinguish tumor subclasses withclinical implicationsrdquo Proceedings of the National Academy ofSciences of theUnited States of America vol 98 no 19 pp 10869ndash10874 2001
[41] J S Carroll C A Meyer J Song et al ldquoGenome-wide analysisof estrogen receptor binding sitesrdquo Nature Genetics vol 38 no11 pp 1289ndash1297 2006
[42] K Basso A A Margolin G Stolovitzky U Klein R Dalla-Favera and A Califano ldquoReverse engineering of regulatorynetworks in human B cellsrdquo Nature Genetics vol 37 no 4 pp382ndash390 2005
[43] K C Liang andXWang ldquoGene regulatory network reconstruc-tion using conditional mutual informationrdquo Eurasip Journalon Bioinformatics and Systems Biology vol 2008 Article ID253894 2008
[44] K Wang B C Bisikirska M J Alvarez et al ldquoGenome-wideidentification of post-translational modulators of transcriptionfactor activity in human B cellsrdquo Nature Biotechnology vol 27no 9 pp 829ndash837 2009
[45] M Hansen L Everett L Singh and S Hannenhalli ldquoMimosamixture model of co-expression to detect modulators of regu-latory interactionrdquo Algorithms for Molecular Biology vol 5 no1 article 4 2010
[46] O Babur E Demir M Gonen C Sander and U DogrusozldquoDiscovering modulators of gene expressionrdquo Nucleic AcidsResearch vol 38 no 17 Article ID gkq287 pp 5648ndash5656 2010
[47] T Shimamura S Imoto Y Shimada et al ldquoA novel net-work profiling analysis reveals system changes in epithelial-mesenchymal transitionrdquo PLoS ONE vol 6 no 6 Article IDe20804 2011
[48] HYWuet al ldquoAmodulator based regulatory network for ERal-pha signaling pathwayrdquo BMC Genomics vol 13 Supplement 6article S6 2012
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
Advances in Bioinformatics 11
[49] K-K Yan W Hwang J Qian et al ldquoConstruction and anal-ysis of an integrated regulatory network derived from High-Throughput sequencing datardquo PLoS Computational Biology vol7 no 11 Article ID e1002190 2011
[50] M Flores and Y Huang ldquoTraceRNA a web based applicationfor ceRNAs predictionrdquo in Proceedings of the IEEE GenomicSignal Processing and Statistics Workshop (GENSIPS rsquo12) 2012
[51] S D Hsu F M Lin W Y Wu et al ldquoMiRTarBase a data-base curates experimentally validated microRNA-target inter-actionsrdquo Nucleic Acids Research vol 39 no 1 pp D163ndashD1692011
[52] H Liu D Yue Y Chen S J Gao andYHuang ldquoImproving per-formance of mammalian microRNA target predictionrdquo BMCBioinformatics vol 11 article 476 2010
[53] Y Dong et al ldquoA Bayesian decision fusion approach formicroRNA target predictionrdquo BMC Genomics vol 13 2012
[54] J A Asm and M Montague ldquoModels for Metasearchrdquo in Pro-ceedings of the 24th annual international ACM SIGIR conferenceon Research and development in information retrieval pp 276ndash284 la New Orleans La USA 2001
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Anatomy Research International
PeptidesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
International Journal of
Volume 2014
Zoology
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Molecular Biology International
GenomicsInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioinformaticsAdvances in
Marine BiologyJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Signal TransductionJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Evolutionary BiologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Biochemistry Research International
ArchaeaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Genetics Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Virolog y
Hindawi Publishing Corporationhttpwwwhindawicom
Nucleic AcidsJournal of
Volume 2014
Stem CellsInternational
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Enzyme Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Microbiology