Research Article A Module Analysis Approach to Investigate ...downloads.hindawi.com/journals/ecam/2013/731370.pdf · action of TCM formula. Hence, applying classic module detection

Hindawi Publishing CorporationEvidence-Based Complementary and Alternative MedicineVolume 2013 Article ID 731370 14 pageshttpdxdoiorg1011552013731370

Research ArticleA Module Analysis Approach to InvestigateMolecular Mechanism of TCM Formula A Trial onShu-feng-jie-du Formula

Jianglong Song1 Fangbo Zhang2 Shihuan Tang2 Xi Liu1 Yibo Gao1 Peng Lu12

Yanping Wang3 and Hongjun Yang2

1 Institute of Automation Chinese Academy of Sciences Beijing 100190 China2 Institute of Chinese Materia Medica China Academy of Chinese Medical Sciences Beijing 100700 China3 Institute of Basic Research in Clinical Medicine China Academy of Chinese Medical Sciences Dongzhimen Beijing 100700 China

Correspondence should be addressed to Yanping Wang wang4816yahoocomcn andHongjun Yang hongjun0420vipsinacom

Received 12 September 2013 Revised 8 October 2013 Accepted 11 October 2013

Academic Editor Aiping Lv

Copyright copy 2013 Jianglong Song et alThis 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

At the molecular level it is acknowledged that a TCM formula is often a complex system which challenges researchers tofully understand its underlying pharmacological action However module detection technique developed from complex networkprovides new insight into systematic investigation of the mode of action of a TCM formula from the molecule perspective Wehere proposed a computational approach integrating the module detection technique into a 2-class heterogeneous network (2-HN) which models the complex pharmacological system of a TCM formula This approach takes three steps construction of a2-HN identification of primary pharmacological units and pathway analysis We employed this approach to study Shu-feng-jie-du (SHU) formula which aimed at discovering its molecular mechanism in defending against influenza infection Actually fourprimary pharmacological units were identified from the 2-HN for SHU formula and further analysis revealed numbers of biologicalpathways modulated by the four pharmacological units 24 out of 40 enriched pathways that were ranked in top 10 correspondingto each of the four pharmacological units were found to be involved in the process of influenza infection Therefore this approachis capable of uncovering the mode of action underlying a TCM formula via module analysis

1 Introduction

With the development and evolution for thousands of yearsTraditional Chinese Medicine (TCM) has become a soundand complete theory based on distinct principles and founda-tion from Western Medicine TCM formulae characterizedby abundant ingredients and vast associated targets areusually effective alternatives to western drugs for variousmultifactorial disorders [1] Influenced by the decreasedefficiency of new drug invention in recent years the patternof drug design has to evolve from traditional ldquoone drug onetargetrdquo to ldquomulticomponent multitargetrdquo drug discovery [23] Asmulticomponent agent with potential treatment effectsTCM formula holds great promise to promote the processof multitarget drug discovery based on molecular networks

[1 4] Thus the investigation of molecular mechanism ofTCM formula plays an important role for better understand-ing the essence of TCM therapies and multicomponent drugdiscovery

Currently network-based approaches become crucial inunveiling and interpreting the mode of action of a TCMformula with the accumulation of volume ldquoomicsrdquo dataand the emerging of network pharmacology So far lotsof researchers have made great effort to acquire and col-lect ldquoomicsrdquo data through advanced in vivo and in vitrotechniques [5ndash7] Among various ldquoomicsrdquo data interactionknowledge such as compound-protein interaction (CPI) andprotein-protein interaction (PPI) as well as Gene Ontology(GO) and pathway annotation make it possible to describeand analyze complex TCM formula in a holistic manner by

2 Evidence-Based Complementary and Alternative Medicine

using computational techniques On the other hand networkpharmacology brought new insight into drug discovery onceit was put forth [8] The research interests of drug discoveryextend from simple disease-drug-gene relations to some newspots such as promiscuity synergistic effect and functionalmodules [9ndash11] Consequently the focus in pharmacologyresearch has shifted to the exploration of multicomponentmultitarget drugs [3 12] In fact plenty of work investigatedthe intrinsic regulating mechanism between many drugs andnumerous targets or synergistic effects of drug combinationsfrom a network perspective [13 14] Meanwhile numerousnetwork-based methods have been developed to decipherthe pathological pattern underlying complex disorders anduncover the mode of action of TCM herbs or formulae [1516]Moreover network targetwas introduced as a new subjectfor studying the pharmacological action of TCMherbs ratherthan individual target or target set [17] By using network-based techniques several TCM formulae such as Liu-wei-di-huang have been particularly observed and studied in orderto discover the underlying mode of action at the molecularlevel [18]Therefore it is essential to investigate themolecularmechanism of TCM formula using network-based methodsespecially in TCM pharmacology research [1]

Notably module analysis technique based on networkmodel holds great promise to deal with most widely-usedTCM formulae of unexpected complexity at the molecularlevel In general a TCM formula contains hundreds ofchemical constituents and may associate with thousandsof potential targets It is a challenging task to identifythe effective bioactive compounds or even discover thepharmacological action of numerous constituents of a TCMformula [1] Thus it is of great importance to capture thedominant modules of the molecular network representing aTCM formula Two common types of dominant modules weare interested in are functional module and pharmacologicalunit A functional module usually represents a group ofgenes or proteins sharing similar molecular functions whilea pharmacological unit is a connected subnetwork in whicha set of compounds with similar physiochemical propertiesmodulate the activities of a group of function-similar geneproducts Typically functional modules or pharmacologicalunits within the molecular network usually hold some sig-nificant properties that are helpful in revealing the modeof action of TCM formula In fact numbers of researchersproposed diversemethods to detect functionalmodules frominteraction networks [11 19 20] On the contrary thereare few researches on identifying pharmacological units formulticomponent drugs [18] On the other hand networkclustering algorithms also known asmodule detectionmeth-ods developed from statistical physics are usually capableof finding significant communities enriched with explicitreal-world meaning [21 22] As a matter of fact severalalgorithms accomplished important tasks in biological fieldsuch as identifying protein complexes [23ndash25] Additionallyto identify functional modules or pharmacological units isobviously another application of network clusteringmethodswhich is crucial in the investigation of pharmacologicalaction of TCM formula Hence applying classic moduledetection algorithms to the molecular network of TCM

formula may contribute to better understanding of its modeof action at the molecular level

We here proposed a computational approach combin-ing clustering algorithm with heterogeneous network toinvestigate the molecular mechanism of TCM formula Thisapproach takes a three-step procedure At first we con-structed a 2-class heterogeneous network (2-HN) comprisedof herbal ingredients and associated targets for a TCM for-mula under studyThen a classicmodule detection algorithmwas applied to the 2-HN and we identified pharmacologicalunits from the 2-HN Finally we finely selected primarypharmacological units and investigated them by pathwayanalysis This approach is apparently applicable for any TCMformula In this paper we use Shu-feng-jie-du formula (SHUformula) as an example to illustrate the procedure of theapproachThepathway analysis of four pharmacological unitsidentified from the 2-HN for SHU formula showed that 24out of 40 enriched pathways that were ranked in top 10corresponding to each of the four pharmacological units weredirectly or indirectly involved in the process of influenzadevelopment

2 Methods

The novel approach is aimed at discovering the molecularmechanism of TCM formula based on a heterogeneousnetwork together with a clustering algorithmThe procedureof this approach mainly consists of three steps the construc-tion of a heterogeneous network module detection fromthe network and the pathway analysis of selected primarypharmacological units In practice we investigated the modeof action of Shu-feng-jie-du formula by using this approach

21 Construct Heterogeneous Network Since our approachtakes advantage of the network to study a TCM formulawe should firstly construct a heterogeneous network com-prised of herbal ingredients and potential targets At thebeginning the specific composition of each herb in a givenTCM formula must be acquired Typical ways to collectthe chemical ingredients of herbs include literature miningTCM database retrieval and identification test By diversemeans we can collect the chemical constituents togetherwith their geometric structure for all herbs in the TCMformula Subsequently the interaction data and potentialtargets could be computed and retrieved respectively basedon the chemical knowledge for the studied TCM formula

First of all we acquire the interaction data betweenherbal compounds by computational chemistry techniquesAlthough various kinds of interaction knowledge is availablecompound pairs with similar chemical structures are widelyused in network-based pharmacology and drug discoveryresearch The rationale is a well-known assumption thatsimilar compounds have similar properties [26] In otherwords similar chemicals may share common targets andare likely to perform synergistic action on complex diseasesThus we evaluate compound pairs by calculating the pairwisechemical similarity using the geometric structure previouslycurated In the field of cheminformatics various methods

Evidence-Based Complementary and Alternative Medicine 3

were proposed to compute the structural similarity betweencompounds Notably fingerprint-based similarity is practi-cally preferable to MaximumCommon Subgraph (MCS) andother methods in dealing with a large number of compoundpairs We here employ Pybel a Python wrapper for theOpenbabel toolkit to calculate fingerprint-based chemicalsimilarity [27] In the similarity measure Tanimoto Coef-ficient is used to evaluate the commonness of fingerprintsderived from two corresponding compounds as follows

119904TC (1198881 1198882) =119888

119886 + 119887 minus 119888

(1)

where 1198881and 119888

2are two compounds 119886 and 119887 are bit

lengths of 1198881and 1198882fingerprints respectively and 119888 is the

number of common bits between 1198881and 1198882fingerprints In

addition a threshold 120579 is predefined to determine whethertwo compounds are similar in structure Compound pairs areconsidered to be similar only if the pairwise similarity is equalto or greater than the threshold In the end similar compoundpairs are collected as one of the sources for the constructionof the heterogeneous network

Next we retrieve potential targets from some authenticdatabases according to the chemical constituents withinthe TCM formula under study When retrieving a specificdatabase such as DrugBank CTD and STITCH we regardthe gene products that interact with herbal compounds aspotential targets Note that only gene products of homosapi-ens (human) will be taken into consideration Once theinitial set of potential targets is achieved the potential targetsshould be carefully selected in order to avoid contingencyIt is understood that ldquohubrdquo targets usually associate withtwo or more chemicals due to the promiscuous propertyof potential target in pharmacological space [9 13] So wehere define Promiscuity Index of a target simply by thenumber of chemicals interacting with that target Similarlythe Promiscuity Index of a chemical can be measured bythe number of its binding targets A threshold 120575 is specifiedbeforehand to eliminate peripheral targets curated for theTCM formula Gene products with Promiscuity Index no lessthan 120575 are eventually selected into the target set for the TCMformula Note that the threshold 120575 is a small integer but isgreater than one for instance 2 or 3

Then we collect interaction relations between geneproducts in the target set from some authentic databasesRecent findings demonstrated that proteins always functionin cooperation with others rather than in isolation insideor out of a cell [13] That is gene products tend to formfunctional modules to participate in certain biological pro-cesses or accomplish specific physiological functions Lots ofdatabases such as HPRD BioGrid IntAct and DIP gatherplenty of acknowledged protein-protein interactions (PPIs)across diverse species We usually select one database as thesource of PPI data due to the diverse reliability of PPIs indifferent databases Therefore the interactome knowledge isintroduced to the heterogeneous network by retrieving PPIsbetween gene products in the target set

Finally we construct an integrated network on the basisof heterogeneous data acquired before Since compoundsand gene products are present in this integrated network

at the same time we consider such a network as a 2-class heterogeneous network (2-HN) In brief 2-class het-erogeneous network (2-HN) is an abstract network modelinvolving two distinct groups of objects As a matter offact heterogeneous network sometimes viewed asmultilayernetwork has been employed in recent work to study complexdrug-target interactions and predict disease genes [18 28 29]In our case the 2-HN describes a complex pharmacologicalsystem relating the TCM formula under study to its treatablediseases From a local point of view the 2-HN can be dividedinto three subnetworks in chemical pharmacological andgenomic space in terms of three types of links in the 2-HN(Figure 1) [30] In most cases it is difficult to investigate andanalyze the 2-HN for the TCM formula due to its complexityMoreover dense modules identified from the 2-HN mayreveal some important pathways enriched in a subset ratherthan the whole set of genes related to the TCM formulaTherefore to identify the pharmacological units from the2-HN by module detection methods is always necessaryto uncover the molecular mechanism of the TCM formula(Figure 1)

22 Detect Significant Modules Since the complex networktheory emerged module detection has become one of themajor techniques to promote the application and develop-ment of complex network A great quantity of algorithmshave been devised and implemented to find significantmodules from connected networks [21ndash23] Among variousclassic methods a well-known method Girvan-Newmanalgorithm is capable to detect communities of a complex sys-tem and identify community structure [22] Girvan-Newmanalgorithm is performed by iteratively removing edges withhighest betweenness from the original network In this waythe community structure could be viewed as a dendrogramWe employ clusterMaker an implementation of Girvan-Newman algorithm in Cytoscape to identify significantmodules within the 2-HN for the TCM formula [31 32]

After the clustering partition is detected from the net-work we need a measure to quantify the significance ofidentified modules Notably modularity is an outstandingquality functionmeasuring the goodness of network partition[33 34] Consequently we use a measure similar to thedefinition of modularity to evaluate whether a module issignificant or not in the original network For an undirectedsimple graph the modularity of a module119862 can be expressedas follows

119876 (119862) =

119897119862

119898

minus (

119889119862

2119898

)

2

(2)

where 119897119862= sum119894119895isin119862

119908119894119895is the summation of weights of edges

in module 119862 119889119862

= sum119899isin119862

deg(119899) is the summation ofdegrees of nodes in module 119862 and 119898 = (12)sum

119894119895isin119866119908119894119895

is the size of the graph 119866 Obviously a significant modulecorresponds to a modularity larger than zero A ldquogoodrdquomodule always has a large modularity otherwise a smallmodularity indicates the ldquopoorrdquo significance of a net-work module Moreover according to the definition above


C5C2 C3

C4C1

C6C7 C8

C9

C2

C3

C4

C6

C9

P1

P3P2

P4 P5

P6P7

P8

P9

P10P11

P2

P3

P4

P5

P9

P7

P10

P11

Pharmacological space

C4

C6

C7

C8C9

P4

P5

P6

P10P9

P11

(b) Pharmacological unit

(a) The 2-HN for TCM formula

Compound-compound interactionCompound-protein interactionProtein-protein interaction

Compound without targets

Compound binding to targets Protein not targeted by any compound

Targeted protein

C10 C10

Chemical space Genomic space

Figure 1 (a) A 2-class heterogeneous network (2-HN) modeling the complex system of a TCM formula and its molecular targets A 2-HN can be simply divided into three subnetworks in chemical pharmacological and genomic space in terms of the type of links (b) Apharmacological unit identified from the 2-HN in (a) A pharmacological unit includes a set of structure-similar herbal compounds and agroup of function-similar target genes indicating that the herbal compounds modulate the activities of gene products

the modularity of a clustering partition of a given network isjust the summation of modularities over all modules in thepartition

23 Analyze Pharmacological Units The significant mod-ules identified from the 2-HN of the TCM formula needto be examined before conducting further analysis Firstmodules should be excluded if they are only comprisedof compounds or gene products Since compound-proteininteractions (CPIs) relate herbal ingredients to potentialtargets modules without any CPI make little contribution touncover the pharmacological action of herbal compounds inthe TCM formula Second modules with small modularityclose to zero should be eliminated Generally a module maynot be significant enough to be considered as a rational phar-macological unit for the TCM formula if it has a fairly smallmodularity Third modules should be paid less attentionif the ratio of preserved compound-protein interactions isparticularly low The ratio of preserved CPIs is defined as thenumber of CPIs in a module divided by the total number of

CPIs in the 2-HNThe ratio for a module 119862 can be expressedas

119877 (119862) =

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119862

10038161003816100381610038161003816

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119866

10038161003816100381610038161003816

(3)

where 119888 is a compound and 119892 is a gene product | sdot | is thenorm of a set that is the number of elements in the set If theratio is low or few CPIs are present in a module the moduleis unlikely to represent the primary interacting pattern thatlinks herbal compounds and potential targets for the TCMformula under studyAfter these examinations the remainingmodules can be simply regarded as primary pharmacologicalunits responsible for the studied TCM formula taking effecton complex diseases

We investigate and analyze the primary pharmacologicalunits by pathway analysis Pathway analysis always play anessential role of discovering possible biological processes thatthe genes in the input list participate in A lot of databasescollect many curated pathways concerning metabolism cel-lular processes and diseases such as KEGG BioCarta Reac-tome GeneGo and Ingenuity Besides Gene Ontology (GO)


another kind of pathways usually reveals the physiologicalfunctions and cellular locations of a group of genes or geneproducts Thus pathway and GO supply us with sufficientknowledge about molecular regulation and gene functionOther analysis methods for instance disease analysis usinggene overlapping and biomarkers could provide new insightto understand the underlying functions of the TCM formulaIn this paper we use MetaDrug a platform of systemspharmacology and toxicity to perform pathway analysis forthe identified primary pharmacological units [35] Then themolecular mechanism underlying the studied TCM formulacould be uncovered through analyzing the enriched pathwaysor GO terms for primary pharmacological units

To illustrate the workflow of the approach in detail weapply the approach to an effective agent for influenza Shu-feng-jie-du formula Instead of Shufeng-jie-du formula weuse SHU formula for short in following sections Followingthe procedure of the approach we can investigate the modeof action underlying SHU formula

3 Results and Discussion

31 2-HN for SHU Formula We firstly acquired the herbcomposition of SHU formula and collected chemical con-stituents within each herb In fact SHU formula mainlyconsists of 8 herbs Bai-Jiang-Cao (Herba Patriniae) Ban-Lan-Gen (Radix Isatidis) Chai-Hu (Radix Bupleuri) Gan-Cao (Radix Glycyrrhizae) Hu-Zhang (Rhizoma PolygoniCuspidati) Lian-Qiao (Fructus Forsythiae) Lu-Gen (Rhi-zoma Phragmitis) and Ma-Bian-Cao (Herba Verbenae)(Table 1) According to the herb composition we collected243 nonredundant chemical constituents for this formula Allconstituents of SHU formula were retrieved from the Chem-istry Database founded by Shanghai Institute of OrganicChemistry (httpwwworganchemcsdbcn) The 2D struc-tures of herbal constituentswere downloaded fromPubChemCompound database according to unique CAS RegistryNumber Then we evaluated the similar compound pairsbased on the fingerprint-based Tanimoto similarity Thethreshold 120579 for similarity score was set to 07 as stated in[27] In this way 562 pairs of compounds were collectedand considered to be similar because they had comparablestructural similarities to the threshold In the next stepwe searched Comparative Toxicogenomics Database (CTD)for potential targets interacting with herbal ingredients inSHU formula [36] The threshold 120575 for Promiscuity Indexof potential targets was set to 3 Namely we only selectedgene products targeted by at least 3 herbal compounds aswell as the interactions between those proteins and chemicalsAs a result 238 potential targets were collected from CTDwhich associatedwith herbal compounds by 1101 interactionsAt last we extracted acknowledged interactions between238 gene products extracted before from BioGRID database[37] There were 718 nonredundant PPIs between the curatedpotential targets Based on these data a 2-HN an integratednetwork for SHU formula was constructed Since we focusedon the largest connected component of the 2-HN for SHU

Table 1 Herb composition of Shu-feng-jie-du formula (SHUformula)

Englishtranslation Pharmaceutical name Simplified

Chinese scriptHu-Zhang Rhizoma Polygoni Cuspidati 虎杖

Lian-Qiao Fructus Forsythiae 连翘

Ban-Lan-Gen Radix Isatidis 板蓝根

Chai-Hu Radix Bupleuri 柴胡

Bai-Jiang-Cao Herba Patriniae 败酱草

Ma-Bian-Cao Herba Verbenae 马鞭草

Lu-Gen Rhizoma Phragmitis 芦根

Gan-Cao Radix Glycyrrhizae 甘草

formula the resultant network contained 171 herbal com-pounds and 238 potential targets after discarding small-sizecomponents (Table 2)

The 2-HN of SHU formula has some interesting proper-ties in topology As shown in Table 2 two groups of nodesin the 2-HN (rectangle for compounds and ellipse for geneproducts) are connected by three types of links It is obviousthat the pharmacological subnetwork is a bipartite which iscomprised of all CPIs (Table 2) So the 2-HN for SHU formulais beyond a bipartite by including compound interactionsandPPIs (Table 2)Thenetwork heterogeneity decreases from2531 of the pharmacological subnetwork to 1588 of the 2-HNfor SHU formulaThis is because compound interactions andPPIs bring many extra links to the ldquononhubrdquo chemicals andgene products respectively [38] In addition the chemicalsubnetwork has 34 connected components of which 17 areisolated compounds (Table 2) Regardless of the isolatednodes each of the remaining connected components has9059 compounds in average That is herbal compounds inSHU formula tend to form multiple components in terms ofsimilar structure As for the genomic subnetwork there are57 connected components among which 55 are comprised ofisolated proteins (Table 2) In fact nearly all of the noniso-lated proteins connect to a giant component with 181 nodesand 717 links in the genomic subnetwork It suggests thatthe giant component determines the mode of action of SHUformula to a large extent Different from the phenomenonin chemical subnetwork target proteins of SHU formulatend to form a single large component instead of multiplecomponents Furthermore only a small fraction (50 out of171) of the involved herbal compounds (blue rectangles) takedirect or indirect actions on the 238 gene products in the 2-HN (Table 2) Apart from the incompleteness of chemical-protein knowledge we could see that only limited numberof compounds have acknowledged therapeutic effects in SHUformula Among these 50 compounds there are several ldquohubrdquocompounds associated with many targets such as quercetinand resveratrol which may exhibit high activities againstinfluenza progression

The ldquohubrdquo compounds usually play an essential role toachieve the excepted effect of SHU formula treating influenzaWe listed four ldquohubrdquo herbal compounds in Table 3 and inves-tigated their pharmacological functions at the same time


Table 2 Topological properties of the 2-HN for SHU formula and its three subnetworks

Property CSN PSN GSN 2-HNNode Compounds 171 50 0 171

Proteins 0 238 238 238Edge CCIs 481 0 0 481

CPIs 0 1101 0 1101PPIs 0 0 718 718

Connected components 34 1 57 1Isolated nodes 17 0 55 0

Clustering coefficient 0662 00 0198 0414Network density 0033 0027 0025 0028

Network heterogeneity 0664 2531 1287 1588lowastCCI is short for compound-compound interaction CPI is compound-protein interaction andPPI is protein-protein interaction CSN represents the chemicalsubnetwork of the 2-HN for SHU formula PSN the pharmacological subnetwork and GSN the genomic subnetworklowastAll the topological properties were calculated using Cytoscape 28 [32]

Table 3 ldquoHubrdquo herbal compounds identified from the pharmacological subnetwork of the 2-HN for SHU formula

Name CAS RN PubChemCID PI Action Reference

Quercetin 117-39-5 5280343 222(i) Quercetin and rutin exhibit prooxidant effect in healthyand antioxidant activity in influenzamdashinfected animals [39]

(ii) Quercetin and oseltamivir exhibited antivirus effect onthe Toll-like receptor 7 (TLR7) signaling pathway whendendritic cells and macrophages were infected with H1N1

[40]

Resveratrol 501-36-0 445154 218 Resveratrol inhibited the replication of influenza virus inMDCK cells [43]

Kaempferol 520-18-3 5280863 67

Kaempferol inhibited influenza A nucleoproteinproduction in human lung epithelial (A549) cells infectedwith the H5N1 virus strain AThailandKan-104 innon-toxic concentrations

[44]

Eugenol 97-53-0 3314 61Eugenol could inhibit autophagy and influenza A virusreplication inhibit the activation of ERK p38MAPK andIKKNF-120581B signal pathways

[45]

lowastPI is Promiscuity Index of individual compound that is the number of binding targets in the 2-HN for SHU formula

Two outstanding compounds are quercetin and resveratrolwith far larger Promiscuity Index (222 and 218 resp) thanother compounds (the third largest is 67 for kaempferol)Previous works revealed the underlying functions of thesefour compounds in defending against influenza For instancequercetin could relieve the oxidative stress caused by exper-imental influenza virus infection in organisms like lungsand liver [39] Another work demonstrated that quercetintogether with oseltamivir exhibited antivirus effect on theToll-like receptor 7 (TLR7) signaling pathway when dendriticcells and macrophages were infected with H1N1 [40] Severalquercetin derivatives such as quercetin-3-rhamnoside andisoquercetin also served as anti-influenza agents by inhibitingthe replication of influenza virus [41 42] Additionallyresveratrol was found to inhibit the replication of influenzavirus in MDCK cells which involved the blockade of thenuclear-cytoplasmic translocation of viral ribonucleopro-teins [43] Moreover kaempferol could inhibit the influenzaA nucleoprotein production in human lung epithelial cellsinfected by the H5N1 virus [44] and eugenol could inhibitautophagy and influenza A virus replication by suppressing

the activation of ERK p38MAPK and IKKNF-120581B signalpathways [45] Therefore these four ldquohubrdquo herbal com-pounds characterized by large Promiscuity Index indeedtake effect to defend against influenza

Although the general effect of SHU formula could beobserved by studying the ldquohubrdquo herbal compounds in the 2-HN we still neededmodule analysis to further investigate thebiological pathways that SHU formula actually influences andregulatesWe firstly identified primary pharmacological unitsfrom the 2-HN for SHU formula and then investigated theparticular mode of action of SHU formula treating influenza

32 Pharmacological Units from the 2-HN Through detect-ing modules using Girvan-Newman algorithm 12 significantmodules were identified from the 2-HN for SHU formulaHowever not all themodules are fairly important and need tobe analyzed in detail We selected primary pharmacologicalunits from the 12 modules according to three principlesexplained before As shown in Table 4 module 11 is onlycomprised of compounds and thus excluded because it is


Table 4 Metrics of detected modules from the 2-HN for SHU formula

Module Compounds Proteins Valid Modularity Ratio of preserved CPIs1 20 121 Yes 0121375 02570392 37 58 Yes 0075361 0151683 31 2 Yes 0040522 00036334 3 30 Yes 0037876 00236155 17 14 Yes 0021214 00145326 19 1 Yes 0030336 00018177 12 4 Yes 0014417 00036338 9 5 Yes 0013261 00045419 11 1 Yes 0009457 000090810 7 1 Yes 0006564 000090811 3 0 No 0001104 0012 2 1 Yes 0000873 0000908

not a valid pharmacological unit (including compounds andgene products)We chose 002 as the threshold formodularityand consequently five more modules 7 8 9 10 and 12 werediscarded due to the low significance in the original networkThe threshold for the ratio of preserved CPIs was set to 001and another two modules 3 and 6 were eliminated as theyincluded too few CPIs In the end four modules 1 2 4 and5 were selected and considered as primary pharmacologicalunits From the topological perspective modules 1 2 4 and5 are highly connected in the background network of the 2-HN characterized by relatively large modularities Besidesthese four pharmacological units are of great importance torepresent the pharmacological essence of SHU formula dueto the large amount of preserved CPIs from the originalsystem So we made great effort to investigate these fourpharmacological units by pathway analysis

We analyzed the underlying biology by performingenrichment analysis with pathways from GeneGo databaseFor each primary pharmacological unit we employed thegenes within the module as input gene list to search forenriched pathways in GeneGo database The top 10 enrichedpathways corresponding to each module were illustrated inFigure 2 The pathways were sorted according to the 119875 valuewhichmeasured the significance of a given pathway enrichedin the gene list of a pharmacological unit The bioactivecompounds in every pharmacological unit potentially actingon the enriched pathways were also highlighted in Figure 2The associated herbal compounds were ranked by Promis-cuity Index which was defined as the number of targetsconnected to a given compound by the preserved CPIs inan identified module (Materials and Methods) From theviewpoint of pathway category the bioactive compounds inevery primary pharmacological unit seemed to particularlyinterfere with pathways from one or two specific categoriesFor example compounds in module 1 generally participatein the processes of cell cycle (4 pathways) and development(4 pathways) the highly enriched pathways of module 2exhibit high relevance tometabolism (9 pathways) especiallythe estradiol metabolism (3 pathways) module 4 mostlyinfluence the biological processes related to apoptosis andsurvival (10 pathways) andmodule 5 interfere in the activities

of cell adhesion (4 pathways) and cytoskeleton remodeling (3pathways) as well as immune response (3 pathways) Despiteof the redundancy of GeneGo pathways we could see thateach of the four pharmacological units tends to regulaterelevant pathways from specific categories which impliesthat SHU formula carries out pharmacological efficacy bysimultaneously intervening pathological activities from dis-tinct aspects at the pathway level Since the module analysisapproach was applied to SHU formula generated explicitresults as exhibited in Figure 2 we should verify the reliabilityof the prediction and evaluate the relevance of SHU formulato influenza infection

According to Figure 2 we could find that compoundsin all four pharmacological units had potential effects oninfluenza infection At first 40 enriched pathways in Figure 2were regulated to some extent by corresponding herbalcompounds in each module which can be explained by theacknowledged regulatory relations between compounds andpathway components from CTD For example resveratrolinfluences the EGFR signaling pathway through binding toEGFR protein and thus decreasing the phosphorylation ofEGFR protein [46] However since not all enriched pathwayswere involved in the activities of influenza infection weparticularly focused on those related to influenza progressionand the regulatory relations between SHU formula andthose pathways As shown in Table 5 24 of the 40 enrichedpathways were found to directly or indirectly participatein the processes of influenza virus invasion productionproliferation and transition and to account for the influenza-induced syndromes as well such as inflammation Here weprimarily studied the specific action of herbal compounds ineach pharmacological unit on 24 influenza-related pathwayswhile the participation of these pathways in the progressionof influenza would be analyzed in following section Formodule 1 resveratrol togetherwith other compounds blockedthe G1S-phase transition [47] inhibited the EGFRHER2signaling pathway [46] and regulated the PTENAKT path-way [46] Quercetin and kaempferol together with otherbioactive compounds in module 2 showed inhibitory effecton the in vitro hepatic metabolism of 17120573-estradiol [48] andon the hydroxylation of benzo[a]pyrene [49] Additionally


Gamma-aminobutyric acid

ActeosidePalmitic acid

Dicumarol Citral

GlycerolCoumarin

Pelargonic acidOctanoic acid

Beta-carotene

Resveratrol Eugenol

Alpha-tocopherolLinalool

Histamine [(6)]responseimmuneinsignalingreceptorH1

Influence proteinsRhoandRasofon [(1)(2)]transitionG1S

ESR1 [(1)(3)]transitionG1Sofregulation

Nucleocytoplasmic [(1)]CDKcyclinsoftransport

EGFR [(3)]pathwaysignaling

TGF-beta-dependent [(3)]MAPKviaEMTofinduction

Gastrin [(3)]proliferationandgrowthcellin

AKT [(4)]signaling

Brca1 [(5)]regulatortranscriptionaas

Regulation [(1)]2)(parttransitionG1Sof

100 10 1

PI

1 1e minus 10 1e minus 20 1e minus 30

P value

(a)

Gamma-decalactone

Ferulic acidPolydatinScopoletin

p-Coumaric acidGenkwanin

IndigotinChrysophanic acid

Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin

PGE2 [(3)(6)]responseimmuneinsignaling

Acetaminophen [(7)]metabolism

Androstenedione and testosterone biosynthesis and [(7)]p2metabolism

Androstenedione and testosterone biosynthesis and [(7)]version)p2(Rodentmetabolism

1-Naphthylamine and 1-nitronaphtalene metabolism [(7)]

Estradiol [(7)]version)(humanmetabolism

Estradiol [(7)]version)(rodentmetabolism

Estradiol [(7)]metabolism

Benzo[a]pyrene [(7)]metabolism


100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid

Ceramides [(8)]pathwaysignaling

Role [(8)]apoptosisinIAP-proteinsof

FAS [(8)]cascadessignaling

Caspase [(8)]cascade

TNFR1 [(6)(8)]pathwaysignaling

Regulation [(8)]proteinsmitochondrialbyapoptosisof

Inhibition [(3)(8)]PEDFbyangiogenesisof

Cytoplasmicmitochondrial transport of proapoptotic proteins [(8)]BimandBmfBid

Granzyme [(8)(9)]signalingB

Apoptotic [(8)]pathwaysTNF-family

100 10 1

PI


P value

(1) Cell cycle(2) Cytoskeleton remodeling(3) Development(4) Protein function(5) DNA-damage(6) Immune response

(7) Metabolism(8) Apoptosis survival(9) Proteolysis(10) Cell adhesion(11) Blood coagulation

(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin

MIF-mediated [(6)]regulationglucocorticoid

Chemokines [(10)]adhesionand

PLAU [(10)]signaling

VEGF [(3)]cascadesVEGFR2-genericviasignaling

Cytoskeleton [(2)]remodeling

HMGB1RAGE [(2)(3)(6)(10)(11)]pathwaysignaling

ECM [(10)]remodeling

Blood [(11)]coagulation

HSP60 [(6)]pathwaysignalingHSP70TLRand

TGF [(2)]remodelingcytoskeletalandWNT

100 10 1

PI


P value



(d)

Figure 2 (a) (b) (c) and (d) Top 10 enriched pathways and associated herbal compounds corresponding tomodule 1 2 4 and 5 respectivelyThe herbal compounds are ranked by Promiscuity Index (PI) which is defined as the number of targets connected to a given compound bythe preserved CPIs in a detected module Note that only compounds with PI greater than zero are listed in this figureThe enriched pathwaysare ranked by the 119875 values calculated in MetaDrug The circled numbers in brackets after pathway name indicate the major category thatpathway belongs to For example ldquoESR1 regulation of G1S transitionrdquo belongs to category 1 and 3 that is cell cycle and development Thecategory knowledge is curated from the classification tree of GeneGo pathways in MetaDrug All pathways in this figure are significant with119875 values lower than 0001

quercetin also suppressed COX-2 expression and PGE2production [50] Herbal compounds in module 4 such asadenosine phenol and betulinic acid tended to inhibit IL-12 and TNF-120572 production [51] downregulate the expressionof IAP2 [52] and trigger CD95 (APO-1Fas)- and p53-independent apoptosis [53] Compounds in module 5 likecatechin could inhibit the endotoxin-inducedHMGB1 release[54] and block the TLR signaling pathway [55] Moreoverthe remaining 16 pathways were also likely to correlate withinfluenza infection although there has been no literaturesupport for those pathways so far In brief 24 influenza-related pathways elucidated the potential effects of SHUformula against influenza infection from diverse aspects atthe pathway level

Moreover by exploring the development of influenza wecould explicitly see how the enriched pathways modulatedby bioactive components in SHU formula led human phys-iological system to a serious disease state These pathwayseither promoted the production and replication of viralRNAs or proteins or induced host immune response andinflammation The participation of these pathways in thepathological process of influenza infection discussed in

the next section explained how SHU formula treated againstinfluenza infection by intervening various pathways in differ-ent stages and cellular locations

33 SHU Formula Treating Influenza When Influenza Avirus (H1N1) enters host cells it induces host cell cyclearrest in G(0)G(1) phase and creates favorable conditionsfor viral replication The nonstructural protein 1 (NS1) ofinfluenza A virus induces G(0)G(1) cell cycle arrest mainlythrough interfering with the RhoApRb signaling pathwaythus providing beneficial conditions for viral protein replica-tion and accumulation [56] The concentration and activityof RhoA protein is pivotal for G(1)S phase transition whichwere decreased with overexpressing NS1 [56] When viralmacromolecules interact with host proteins High-mobility-group box (HMGB) proteins bind to the nucleoprotein(NP) component of viral ribonucleoproteins (vRNPs) in theabsence of viral RNA and HMGB1 protein plays a significantrole in intranuclear replication of influenza viruses [74]PI3KAkt signaling pathway is activated by NS1 protein andinhibition of the PI3KAkt pathway is an anti-influenza


Table 5 Literature-verified pathways related to influenza infection corresponding to four pharmacological units

Module Enriched pathways 119875-value Rank Reference

1

Regulation of G1S transition (part 2) 4137119890 minus 24 1[56]Influence of Ras and Rho proteins on G1S Transition 2156119890 minus 23 2

EGFR signaling pathway 2803119890 minus 20 3 [57]TGF-beta-dependent induction of EMT via MAPK 2603119890 minus 18 5 [58]AKT signaling 5258119890 minus 16 7 [59]Brca1 as a transcription regulator 1710119890 minus 15 9 [60]Histamine H1 receptor signaling in immune response 3503119890 minus 15 10 [61]

2

Estradiol metabolism (human version) 4213119890 minus 9 5[62]Estradiol metabolism 1293119890 minus 7 7

Estradiol metabolism (rodent version) 1832119890 minus 7 8Benzo[a]pyrene metabolism 4024119890 minus 7 9 [63]PGE2 signaling in immune response 6146119890 minus 7 10 [64]

4

Apoptotic TNF-family pathways 8253119890 minus 32 1 [65]Role of IAP-proteins in apoptosis 6132119890 minus 27 2 [66]FAS signaling cascades 6374119890 minus 20 4 [67]Inhibition of angiogenesis by PEDF 2792119890 minus 13 8 [68]Granzyme B signaling 3712119890 minus 13 9 [69]Ceramides signaling pathway 2652119890 minus 12 10 [70]

5

TGF WNT and cytoskeletal remodeling 3303119890 minus 9 1 [71]Chemokines and adhesion 1360119890 minus 7 2 [72]Cytoskeleton remodeling 1502119890 minus 7 3 [73]HMGB1RAGE signaling pathway 5901119890 minus 7 5 [74]HSP60 and HSP70TLR signaling pathway 4805119890 minus 5 9 [75]MIF-mediated glucocorticoid regulation 3981119890 minus 4 10 [76]

lowastThe rank is the order of ascending 119875 values of enriched pathways corresponding to each primary pharmacological unit

strategy which is still in an early phase of preclinical devel-opment [59] In addition influenza virus infection activatesthree distinct MAPKs ERK p38 MAPK and JNK to partic-ipate to various extents in the induction of PGE2 synthesisfrom arachidonic acid in human bronchial epithelial cells[64] Metabolized benzo[a]pyrene (BaP) reduced viral IFNinduction by approximately 80 assessed in LLC-MK2 cell[63]

Airway epitheliumplay an important role in host immuneresponse Many diverse viruses target a polarized epithelialmonolayer during host invasion The polarized epitheliumrestrict the movement of pathogens across the mucosa Thisregulation can be attributed to the presence of a junctionalcomplex between adjacent cells and to an intricate networkof actin filaments [73] Virus-infected alveolar epitheliumregulate CCL2CCR2-dependent monocyte transepithelialmigration dependent on both classical beta(1) and beta(2)integrins but also junctional adhesion molecule pathwaysduring influenza infection [72] The epithelial response toinhaled pathogens in airway epithelium that deposit on theairway epithelial surface includes EGFR signaling cascades[57]

Influenza virus invasion is associated with host immunityand inflammation Inflammatory cytokines such as TNF-120572 IFN-120574 and ET-1 may trigger the occurrence of AMI[65] Toll-like receptors (TLRs) play an important role inearly innate viral inhibition in naturally occurring influenza

with inflammatory cytokine responses [75] Histaminemedi-ates the acute inflammatory and immediate hypersensitivityresponses and it has also been demonstrated to affectchronic inflammation and regulate several essential eventsin the immune response [61] Type V collagen [col(V)]overexpression and IL-17-mediated anti-col(V) immunity arekey contributors to obliterative bronchiolitis pathogenesisIL-17 is shown to induce EMT TGF-120573 mRNA expressionand SMAD3 activation whereas downregulating SMAD7expression in vitro [58] Macrophage migration inhibitoryfactor (MIF) is involved in inflammatory responses to H5N1influenza virus infections by induction of pulmonary inflam-matory cytokines and chemokines [76] BRCA1 regulatesinflammation-induced endothelial cell function and limitsendothelial cell apoptosis and dysfunction [60] Pigmentepithelial-derived factor (PEDF) suppresses inflammation byinhibiting lipopolysaccharide-driven macrophage activationin vitro and in vivo [68] GzmB deficiency associated withpathology morbidity andmortality results in exacerbation oflymphocytic inflammation during bleomycin-induced acutelung injury [69] Ceramide is the core of sphingolipidmetabolism and phosphorylation of ceramide by ceramidekinase gives rise to ceramide-1-phosphatewhich has also beenshown to participate in inflammation [70]

Besides immune responses in host defence influenza Avirus infection induces endoplasmic reticulum stress Fas-dependent apoptosis and TGF-120573 production in a variety of


Eugenol

Influence of Ras and Rho proteins

on G1S transition

Brca1 as a transcription

regulator

ESR1 regulation of G1S transition

AKT signaling

Resveratrol

TGF-beta-dependentinduction of EMT

via MAPK

Acteoside

Nucleocytoplasmictransport of CDKcyclins

Regulation of G1S transition (part 2)

Dicumarol

Palmitic acidAlpha-tocopherol

EGFR signaling pathway

Beta-carotene

Histamine H1 receptor signaling

in immune responseGastrin in cell

growth and proliferation

Linalool

Figure 3 An illustration of SHU formula intervening the influenza development through multiple pathways The blue rectangle is bioactiveherbal compounds derived from SHU formula The ellipse represents biological pathways that the compounds modulate The red ones areliterature-verified pathways that participate in the process of influenza infection while the gray ones are not verified yet A thick edgeindicates many common hits (pathway components that are also associated targets of herbal compounds) between two pathways or betweena compound and a pathway

cells [71] Inhibitor of apoptosis proteins (IAPs) influenceubiquitin-dependent pathways thatmodulate innate immunesignaling via activation of nuclear factor 120581B (NF-120581B) [66]Multiple influenza virus factors have been identified that canactivate intrinsic or extrinsic apoptotic induction pathwaysdsRNA NS1 NA and PB1-F2 are influenza virus inducersof apoptosis dsRNA and NA act via an extrinsic mecha-nism involving proapoptotic host-defensemolecules PKR byinduction of Fas-Fas ligand and NA by activation of TGF-beta PB1-F2 act intrinsically by localization and interactionwith the mitochondrial-dependent apoptotic pathway [67]

The symptoms of influenza virus infection are relatedto gender Females suffer a worse outcome from influenzaA virus infection than males which can be reversed byadministration of estradiol to females and reflects differencesin the induction of proinflammatory responses [62]

34 Discussion According to the results of pathway analysiswe built a simple network to illustrate the pharmacologicalaction of SHU formula against influenza infection (Figure 3)This networkwas constructed based onmodule 1 identified byGirvan-Newman algorithm from the 2-HN of SHU formulaThe edge connecting a compound and a pathway indicatesthe cooccurrence of associated targets of the compound andpathway components while the edge between two pathwaysrepresents the commonness of hits (pathway componentsthat are also associated targets of herbal compounds) cor-responding to both pathways As shown in Figure 3 8

bioactive compounds of module 1 modulate 10 enrichedpathways related to influenza infection From the perspectiveof topology resveratrol is the most important to regulatethe involved pathways compared to other compounds Itis obvious that resveratrol is connected to all 10 pathwaysthrough strong links indicating that resveratrol mediatesmultiple gene products in these pathways Besides resveratrolis found to modulate the 1198661119878-phase transition (119875 value41119890 minus 24) [47] the EGFRHER2 signaling pathway (119875 value28119890minus20) [46] and the PTENAKTpathway (119875 value 53119890minus16)[46] Other compounds like Acteoside also perform similarfunctions on the involved pathways [77] Of the top 10enriched pathways 7 (red ellipse) are found to participatein the development of influenza and its induced symptomsillustrated in Table 5Thus the herbal compounds in Figure 3are likely to intervene in the invasion production prolifer-ation and transition of influenza virus through mediatingmultiple relevant pathways Three pathways (grey ellipse)regulated by the compounds in Figure 3 hold great promiseto influence the influenza development while such predictionneeds further work to test and verify

In this paper we presented a computational approachbased onmodule analysis to investigate themolecularmecha-nism of TCM formulaThis approach has several advantagesOn one hand we employed a precise model 2-class hetero-geneous network (2-HN) to represent the pharmacologicalsystem of a TCM formula Since a 2-HN is structurallymore complete than a bipartite by incorporating interactions


within the same categories so additional information isintegrated into such a comprehensive model In case of the2-HN for SHU formula besides the regulatory relationsbetween chemicals and gene products similar compoundswithin SHU formula and interactions between gene productsare also taken into consideration when studying the modeof action of SHU formula This additional information rep-resented by compound-compound interactions (CCIs) andPPIs is critical to systematic investigation of multicomponentdrugs while traditionalmethods always disregard knowledgelike this or use it separately [15] On the other hand theapproach presented in this paper takes advantage of moduledetection technique to uncover themolecularmechanismof aTCM formula Different from conventionalmethods we ana-lyze small-size yet topologically significant pharmacologicalunits rather than the whole drug-target system of unexpectedcomplexity Generally the pharmacological units identifiedby module detection methods are more reliable in topologythan the original systemThis is because the pharmacologicalunits are significantly components in the original networkfeatured by dense intraconnections So a 2-HN together withmodule detection technique could deal with the challengingtask of discovering the molecular mechanism of a TCMformula from its pharmacological system with hundredsof herbal compounds and thousands of targets as well asunpredictable amount of interactions

Although the approach provides new insight into molec-ular mechanism of TCM formula it can be improved in threeaspects First the compound interaction is not limited tostructurally similar compound pair The derivative or iso-metric relation similarity in physicochemical property andontology similarity between compounds may outperformstructural similarity to some extent Second the moduledetection methods could be improved in order to (i) identifymodules with overlapping nodes and edges and (ii) take intoaccount the differences of interactions in a 2-HN Generallya compound may have diverse therapeutic functions anda gene may participate in diverse biological processes Inother words a node should be assigned to two or moremodules representing diverse functions or processes Sooverlapping modules detected from a 2-HN may be moreconsistent with reality In addition CPIs in a 2-HN shouldbe paid more attention than CCIs and PPIs when detectingpharmacological unitsThis is because CPIs are indispensablein a pharmacological unit that is a connected subnetworkcontaining compounds and gene products Third we couldadopt improved pathway analysis to uncover the biologyunderlying identified pharmacological units As elaboratedin [78] pathway enrichment analysis has two inevitableshortcomings It treats every gene equally when findingpathways enriched in the input gene list Besides it does nottake the pathway dependence into account which results inthree ldquoEstradiol metabolismrdquo pathways enriched in module2 gene list (Figure 2) So precise pathway techniques arein need to find rational and reliable pathways underlyingeach primary pharmacological units from the 2-HN for agiven TCM formula With these improvements the moduleanalysis-based approach will be more capable of uncoveringexplicit molecular mechanism of TCM formula

4 Conclusion

We here propose a computational approach based onmoduleanalysis to investigate the molecular mechanism underly-ing TCM formula The approach incorporates the moduledetection technique with a 2-class heterogeneous networka precise model to depict the complex system of a TCMformula This approach mainly consists of three steps net-work construction module detection and pathway analysisThe application of this approach to Shu-feng-jie-du formulaoutputs good results which identified four primary phar-macological units uncovering key herbal compounds andessential pathways they modulated 24 out of 40 enrichedpathways that were ranked in top 10 corresponding to eachof the four pharmacological units were found to be relevantto the process of influenza infection and some induced symp-toms like inflammation This demonstrates the effectivenessof our approach in discovering the molecular mechanismof a TCM formula Although effective this approach stillrequires improvement with regard to chemical similaritymodule detection algorithm and accurate pathway analysisof identified modules After all our approach provides newinsight into discovering the molecular basis of TCM formulaand further promotes the large-scale exploration of thepharmacological action of multicomponent drugs in a low-cost manner especially TCM formulae

Conflict of Interests

The authors declare that they do not have a direct financialrelation with any commercial identity including the onementioned in the paper None of the authors have a conflictof interests to declare

Authorsrsquo Contribution

Jianglong Song Fangbo Zhang and Shihuan Tang con-tributed equally to this work

Acknowledgments

This work was supported by the Special Research Foundationfor Traditional Chinese Medicine (Grant no 200907001-5)the National Science Foundation for Post-doctoral Scientistsof China (Grant no 2012M510733) and the National ScienceFoundation of China (Grant no 81303152)

References

[1] J Zhao P Jiang and W Zhang ldquoMolecular networks for thestudy of TCM pharmacologyrdquo Briefings in Bioinformatics vol11 no 4 Article ID bbp063 pp 417ndash430 2009

[2] F Sams-Dodd ldquoTarget-based drug discovery is somethingwrongrdquo Drug Discovery Today vol 10 no 2 pp 139ndash147 2005

[3] G R Zimmermann J Lehar andC TKeith ldquoMulti-target ther-apeutics when the whole is greater than the sum of the partsrdquoDrug Discovery Today vol 12 no 1-2 pp 34ndash42 2007

[4] E L Leung Z W Cao Z H Jiang H Zhou and L Liu ldquoNet-work-based drug discovery by integrating systems biology and


computational technologiesrdquo Briefings in Bioinformatics vol 14no 4 pp 491ndash505 2013

[5] M Ashburner C A Ball J A Blake et al ldquoGene ontology toolfor the unification of biologyrdquoNature Genetics vol 25 no 1 pp25ndash29 2000

[6] C Knox V Law T Jewison et al ldquoDrugbank 30 a comprehen-sive resource for ldquoOmicsrdquo research on drugsrdquo Nucleic AcidsResearch vol 39 no 1 pp D1035ndashD1041 2011

[7] L Salwinski C S Miller A J Smith F K Pettit J U Bowieand D Eisenberg ldquoThe database of interacting proteins 2004updaterdquo Nucleic Acids Research vol 32 pp D449ndashD451 2004

[8] A L Hopkins ldquoNetwork pharmacology the next paradigm indrug discoveryrdquoNature Chemical Biology vol 4 no 11 pp 682ndash690 2008

[9] G V Paolini R H B Shapland W P van Hoorn J S Masonand A L Hopkins ldquoGlobal mapping of pharmacological spacerdquoNature Biotechnology vol 24 no 7 pp 805ndash815 2006

[10] M Cokol H N Chua M Tasan et al ldquoSystematic explorationof synergistic drug pairsrdquo Molecular Systems Biology vol 7article 544 2011

[11] S Suthram J T Dudley A P Chiang R Chen T J Hastieand A J Butte ldquoNetwork-based elucidation of human diseasesimilarities reveals common functional modules enriched forpluripotent drug targetsrdquo PLoS Computational Biology vol 6no 2 Article ID e1000662 2010

[12] J J LuW Pan Y J Hu and Y TWang ldquoMulti-target drugs thetrend of drug research and developmentrdquo PLoS ONE vol 7 no6 Article ID e40262 2012

[13] M A Yildirim K I Goh M E Cusick A L Barabasi and MVidal ldquoDrugmdashtarget networkrdquo Nature Biotechnology vol 25pp 1119ndash1126 2007

[14] J Jia F Zhu X Ma Z W Cao Y X Li and Y Z ChenldquoMechanisms of drug combinations interaction and networkperspectivesrdquo Nature Reviews Drug Discovery vol 8 no 2 pp111ndash128 2009

[15] Y Sun R Zhu H Ye et al ldquoTowards a bioinformatics analysisof anti-alzheimerrsquos herbal medicines from a target networkperspectiverdquo Briefings in Bioinformatics vol 14 no 3 pp 327ndash343 2013

[16] L Wang G-B Zhou P Liu et al ldquoDissection of mechanismsof Chinese medicinal formula realgar-indigo naturalis as aneffective treatment for promyelocytic leukemiardquo Proceedings ofthe National Academy of Sciences of the United States of Americavol 105 no 12 pp 4826ndash4831 2008

[17] S Li B Zhang and N Zhang ldquoNetwork target for screeningsynergistic drug combinations with application to traditionalChinese medicinerdquo BMC Systems Biology vol 5 no 1 articleS10 2011

[18] S Li B Zhang D Jiang Y Wei and N Zhang ldquoHerb net-work construction and co-module analysis for uncovering thecombination rule of traditional Chinese herbal formulaerdquo BMCBioinformatics vol 11 no 11 article S6 2010

[19] S G A Konietzny L Dietz and A C McHardy ldquoInferringfunctional modules of protein families with probabilistic topicmodelsrdquo BMC Bioinformatics vol 12 article 141 2011

[20] M T Dittrich G W Klau A Rosenwald T Dandekar andT Muller ldquoIdentifying functional modules in protein-proteininteraction networks an integrated exact approachrdquo Bioinfor-matics vol 24 no 13 pp i223ndashi231 2008

[21] S Fortunato ldquoCommunity detection in graphsrdquoPhysics Reportsvol 486 no 3ndash5 pp 75ndash174 2010

[22] M Girvan and M E J Newman ldquoCommunity structure insocial and biological networksrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 99 no12 pp 7821ndash7826 2002

[23] G Palla I Derenyi I Farkas and T Vicsek ldquoUncoveringthe overlapping community structure of complex networks innature and societyrdquoNature vol 435 no 7043 pp 814ndash818 2005

[24] A J Enright S van Dongen and C A Ouzounis ldquoAn efficientalgorithm for large-scale detection of protein familiesrdquo NucleicAcids Research vol 30 no 7 pp 1575ndash1584 2002

[25] P Jiang and M Singh ldquoSPICi a fast clustering algorithm forlarge biological networksrdquo Bioinformatics vol 26 no 8 ArticleID btq078 pp 1105ndash1111 2010

[26] M Johnson and G Maggiora Concepts and Applications ofMolecular Similarity Wiley-Interscience 1990

[27] NMOrsquoBoyle CMorley andG RHutchison ldquoPybel a pythonwrapper for the ppenbabel cheminformatics toolkitrdquo ChemistryCentral Journal vol 2 no 1 article 5 2008

[28] X Wu R Jiang M Q Zhang and S Li ldquoNetwork-based globalinference of human disease genesrdquoMolecular Biology of Diseasevol 4 article 189 2008

[29] S Zhao and S Li ldquoNetwork-based relating pharmacological andgenomic spaces for drug target identificationrdquo PLoS ONE vol5 no 7 Article ID e11764 2010

[30] Y YamanishiM Araki A GutteridgeWHonda andM Kane-hisa ldquoPrediction of drug-target interaction networks from theintegration of chemical and genomic spacesrdquo Bioinformaticsvol 24 no 13 pp i232ndashi240 2008

[31] J H Morris L Apeltsin A M Newman et al ldquoClustermakera multi-algorithm clustering plugin for cytoscaperdquo BMC Bioin-formatics vol 12 article 436 2011

[32] M E Smoot K Ono J Ruscheinski P-L Wang and T IdekerldquoCytoscape 28 new features for data integration and networkvisualizationrdquo Bioinformatics vol 27 no 3 Article ID btq675pp 431ndash432 2011

[33] M E J Newman and M Girvan ldquoFinding and evaluatingcommunity structure in networksrdquo Physical Review E vol 69no 2 Article ID 026113 15 pages 2004

[34] M E J Newman ldquoModularity and community structure innetworksrdquoProceedings of theNational Academy of Sciences of theUnited States of America vol 103 no 23 pp 8577ndash8582 2006

[35] S Ekins A Bugrim L Brovold et al ldquoAlgorithms for net-work analysis in systems-ADMETox using the metacore andmetadrug platformsrdquoXenobiotica vol 36 no 10-11 pp 877ndash9012006

[36] A P Davis T C Wiegers R J Johnson et al ldquoText min-ing effectively scores and ranks the literature for improvingchemical-gene-disease curation at the comparative toxicoge-nomics databaserdquo PLoS ONE vol 8 no 4 Article ID e582012013

[37] A Chatr-aryamontri B J Breitkreutz S Heinicke et al ldquoTheBioGRID interaction database 2013 updaterdquo Nucleic AcidsResearch vol 41 pp D816ndashD823 2013

[38] J Dong and S Horvath ldquoUnderstanding network concepts inmodulesrdquo BMC Systems Biology vol 1 article 24 2007

[39] V M Savov A S Galabov L P Tantcheva et al ldquoEffects ofrutin and quercetin on monooxygenase activities in experi-mental influenza virus infectionrdquo Experimental and ToxicologicPathology vol 58 no 1 pp 59ndash64 2006

[40] C Chen Z Y Jiang B Yu et al ldquoStudy on the anti-h1n1 viruseffects of quercetinand oseltamivir and theirmechanism related


to tlr7 pathwayrdquo Journal of Asian Natural Products Research vol14 no 9 pp 877ndash885 2012

[41] H J Choi J H Song K S Park and D H Kwon ldquoInhibitoryeffects of quercetin 3-rhamnoside on influenza A virus replica-tionrdquo European Journal of Pharmaceutical Sciences vol 37 no3-4 pp 329ndash333 2009

[42] Y Kim S Narayanan andK-O Chang ldquoInhibition of influenzavirus replication by plant-derived isoquercetinrdquo AntiviralResearch vol 88 no 2 pp 227ndash235 2010

[43] A T Palamara L Nencioni K Aquilano et al ldquoInhibition ofinfluenzaAvirus replication by resveratrolrdquo Journal of InfectiousDiseases vol 191 no 10 pp 1719ndash1729 2005

[44] P Sithisarn M Michaelis M Schubert-Zsilavecz and J CinatlJr ldquoDifferential antiviral and anti-inflammatorymechanisms ofthe flavonoids biochanin A and baicalein in H5N1 influenza Avirus-infected cellsrdquo Antiviral Research vol 97 no 1 pp 41ndash482013

[45] J P Dai X F Zhao J Zeng et al ldquoDrug screening forautophagy inhibitors based on the dissociation of beclin1-bcl2complex using bifc technique and mechanism of eugenol onanti-influenza A virus activityrdquo PLoS ONE vol 8 no 4 ArticleID e61026 2013

[46] Y Wang T Romigh X He et al ldquoResveratrol regulates thePTENAKT pathway through androgen receptor-dependentand -independent mechanisms in prostate cancer cell linesrdquoHuman Molecular Genetics vol 19 no 22 Article ID ddq354pp 4319ndash4329 2010

[47] M Savio T Coppa L Bianchi et al ldquoThe resveratrol analogue441015840-dihydroxy-trans-stilbene inhibits cell proliferation withhigher efficiency but different mechanism from resveratrolrdquoInternational Journal of Biochemistry and Cell Biology vol 41no 12 pp 2493ndash2502 2009

[48] W Schubert U Eriksson B Edgar G Cullberg and THedner ldquoFlavonoids in grapefruit juice inhibit the in vitrohepatic metabolism of 17120573-estradiolrdquo European Journal of DrugMetabolism and Pharmacokinetics vol 20 no 3 pp 219ndash2241995

[49] M K Buening R L Chang and M T Huang ldquoActivationand inhibition of benzo(a)pyrene and aflatoxin B1 metabolismin human liver microsomes by naturally occurring flavonoidsrdquoCancer Research vol 41 no 1 pp 67ndash72 1981

[50] X Xiao D Shi L Liu et al ldquoQuercetin suppressescyclooxygenase-2 expression and angiogenesis throughinactivation of P300 signalingrdquo PLoS ONE vol 6 no 8 ArticleID e22934 2011

[51] G Hasko D G Kuhel J-F Chen et al ldquoAdenosine inhibitsIL-12 and TNF-120572 production via adenosine A(2a) receptor-dependent and independent mechanismrdquo The FASEB Journalvol 14 no 13 pp 2065ndash2074 2000

[52] D Yang T Yaguchi T Nakano and T Nishizaki ldquoAdenosine-induced caspase-3 activation by tuning Bcl-XLDIABLO IAPexpression in HuH-7 human hepatoma cellsrdquo Cell Biology andToxicology vol 26 no 4 pp 319ndash330 2010

[53] S Fulda C Friesen M Los et al ldquoBetulinic acid triggers CD95(APO-1Fas)- and p53-independent apoptosis via activation ofcaspases in neuroectodermal tumorsrdquo Cancer Research vol 57no 21 pp 4956ndash4964 1997

[54] W Li M Ashok J Li H Yang A E Sama and H Wang ldquoAmajor ingredient of green tea rescues mice from lethal sepsispartly by inhibiting HMGB1rdquo PLoS ONE vol 2 no 11 ArticleID e1153 2007

[55] K-M Lee M Yeo J-S Choue et al ldquoProtective mechanism ofepigallocatechin-3-gallate against Helicobocter pylori-inducedgastric epithelial cytotoxicity via the blockage of TLR-4 signal-ingrdquo Helicobacter vol 9 no 6 pp 632ndash642 2004

[56] W JiangQWang S Chen et al ldquoInfluenzaA virusNS1 inducesG0G1cell cycle arrest by inhibiting the expression and activity

of RhoA proteinrdquo Journal of Virology vol 87 no 6 pp 3039ndash3052 2013

[57] J L Koff M X G Shao I F Ueki and J A Nadel ldquoMultipleTLRs activate EGFR via a signaling cascade to produce innateimmune responses in airway epitheliumrdquo American Journal ofPhysiology Lung Cellular andMolecular Physiology vol 294 no6 pp L1068ndashL1075 2008

[58] R Vittal L Fan D S Greenspan E A Mickler and BGopalakrishnan ldquoIl-17 induces type V collagen overexpressionand EMT via TGF-120573-dependent pathways in obliterative bron-chiolitisrdquo American Journal of Physiology Lung Cellular andMolecular Physiology vol 304 pp L401ndashL414 2013

[59] W Li G Wang H Zhang et al ldquoInability of NS1 proteinfroman H5N1 influenza virus to activate pi3kakt signalingpathway correlates to the enhanced virus replication upon pi3kinhibitionrdquo Veterinary Research vol 43 article 36 2012

[60] K K Singh P C Shukla A Quan et al ldquoBrca1 is a novel targetto improve endothelial dysfunction and retard atherosclerosisrdquoThe Journal ofThoracic and Cardiovascular Surgery vol 146 no4 pp 949ndash960 2013

[61] M Jutel M Akdis and C A Akdis ldquoHistamine histaminereceptors and their role in immune pathologyrdquo Clinical andExperimental Allergy vol 39 no 12 pp 1786ndash1800 2009

[62] D P Robinson M E Lorenzo W Jian and S L KleinldquoElevated 17120573-estradiol protects females from influenza A viruspathogenesis by suppressing inflammatory responsesrdquo PLoSPathogens vol 7 no 7 Article ID e1002149 2011

[63] N Hahon and J A Booth ldquoBenzo[a]pyrene metabolites effectson viral interferon inductionrdquo Journal of Interferon Researchvol 6 no 5 pp 591ndash602 1986

[64] K Mizumura S Hashimoto S Maruoka et al ldquoRole ofmitogen-activated protein kinases in influenza virus induction of prostaglandin E2 from arachidonic acid in bronchialepithelial cellsrdquo Clinical and Experimental Allergy vol 33 no9 pp 1244ndash1251 2003

[65] X Guan W Yang X Sun et al ldquoAssociation of influenza virusinfection and inflammatory cytokines with acute myocardialinfarctionrdquo Inflammation Research vol 61 no 6 pp 591ndash5982012

[66] J Silke and P Meier ldquoInhibitor of apoptosis (iap) proteins-modulators of cell death and inflammationrdquoCold SpringHarborPerspectives in Biology vol 5 no 2 Article ID a008730 2013

[67] R J Lowy ldquoInfluenza virus induction of apoptosis by intrinsicand extrinsic mechanismsrdquo International Reviews of Immunol-ogy vol 22 no 5-6 pp 425ndash449 2003

[68] P Zamiri S Masli J W Streilein and A W Taylor ldquoPigmentepithelial growth factor suppresses inflammation by modu-lating macrophage activationrdquo Investigative Ophthalmology ampVisual Science vol 47 no 9 pp 3912ndash3918 2006

[69] J A Hirota P R Hiebert M Gold et al ldquoGranzyme Bdeficiency exacerbates lung inflammation in mice followingacute lung injuryrdquo American Journal of Respiratory Cell andMolecular Biology vol 49 no 3 pp 453ndash462 2013

[70] A Gomez-Munoz P Gangoiti L Arana et al ldquoNew insights onthe role of ceramide 1-phosphate in inflammationrdquo Biochimicaet Biophysica Acta vol 1831 no 6 pp 1060ndash1066 2013


[71] E C Roberson J E Tully A S Guala et al ldquoInfluenza inducesendoplasmic reticulum stress caspase-12-dependent apoptosisand c-Jun N-terminal kinase-mediated transforming growthfactor-120573 release in lung epithelial cellsrdquo American Journal ofRespiratory Cell and Molecular Biology vol 46 no 5 pp 573ndash581 2012

[72] S Herold W von Wulffen M Steinmueller et al ldquoAlveolarepithelial cells direct monocyte transepithelial migration uponinfluenza virus infection impact of chemokines and adhesionmoleculesrdquo Journal of Immunology vol 177 no 3 pp 1817ndash18242006

[73] E Delorme-Axford and C B Coyne ldquoThe actin cytoskeleton asa barrier to virus infection of polarized epithelial cellsrdquo Virusesvol 3 no 12 pp 2462ndash2477 2011

[74] D Moisy S V Avilov Y Jacob et al ldquoHMGB1 protein binds toinfluenza virus nucleoprotein and promotes viral replicationrdquoJournal of Virology vol 86 no 17 pp 9122ndash9133 2012

[75] N Lee C K Wong D S Hui et al ldquoRole of human toll-like receptors in naturally occurring influenza a infectionsrdquoInfluenza and Other Respiratory Viruses vol 7 no 5 pp 666ndash675 2013

[76] X Q Hou Y W Gao S T Yang C Y Wang Z Y Ma andX Z Xia ldquoRole of macrophage migration inhibitory factor ininfluenza H5N1 virus pneumoniardquo Acta Virologica vol 53 no4 pp 225ndash231 2009

[77] K-W Lee H J Kim Y S Lee et al ldquoActeoside inhibitshuman promyelocytic HL-60 leukemia cell proliferation viainducing cell cycle arrest at G

0G1phase and differentiation into

monocyterdquo Carcinogenesis vol 28 no 9 pp 1928ndash1936 2007[78] P Khatri M Sirota and A J Butte ldquoTen years of pathway

analysis current approaches and outstanding challengesrdquo PLoSComputational Biology vol 8 no 2 Article ID 100237 2012

Submit your manuscripts athttpwwwhindawicom

Stem CellsInternational

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


MEDIATORSINFLAMMATION

of


Behavioural Neurology

EndocrinologyInternational Journal of



Disease Markers


BioMed Research International

OncologyJournal of



Oxidative Medicine and Cellular Longevity


PPAR Research

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

ObesityJournal of



Computational and Mathematical Methods in Medicine

OphthalmologyJournal of


Diabetes ResearchJournal of



Research and TreatmentAIDS


Gastroenterology Research and Practice


Parkinsonrsquos Disease

Evidence-Based Complementary and Alternative Medicine

Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom


using computational techniques On the other hand networkpharmacology brought new insight into drug discovery onceit was put forth [8] The research interests of drug discoveryextend from simple disease-drug-gene relations to some newspots such as promiscuity synergistic effect and functionalmodules [9ndash11] Consequently the focus in pharmacologyresearch has shifted to the exploration of multicomponentmultitarget drugs [3 12] In fact plenty of work investigatedthe intrinsic regulating mechanism between many drugs andnumerous targets or synergistic effects of drug combinationsfrom a network perspective [13 14] Meanwhile numerousnetwork-based methods have been developed to decipherthe pathological pattern underlying complex disorders anduncover the mode of action of TCM herbs or formulae [1516]Moreover network targetwas introduced as a new subjectfor studying the pharmacological action of TCMherbs ratherthan individual target or target set [17] By using network-based techniques several TCM formulae such as Liu-wei-di-huang have been particularly observed and studied in orderto discover the underlying mode of action at the molecularlevel [18]Therefore it is essential to investigate themolecularmechanism of TCM formula using network-based methodsespecially in TCM pharmacology research [1]

Notably module analysis technique based on networkmodel holds great promise to deal with most widely-usedTCM formulae of unexpected complexity at the molecularlevel In general a TCM formula contains hundreds ofchemical constituents and may associate with thousandsof potential targets It is a challenging task to identifythe effective bioactive compounds or even discover thepharmacological action of numerous constituents of a TCMformula [1] Thus it is of great importance to capture thedominant modules of the molecular network representing aTCM formula Two common types of dominant modules weare interested in are functional module and pharmacologicalunit A functional module usually represents a group ofgenes or proteins sharing similar molecular functions whilea pharmacological unit is a connected subnetwork in whicha set of compounds with similar physiochemical propertiesmodulate the activities of a group of function-similar geneproducts Typically functional modules or pharmacologicalunits within the molecular network usually hold some sig-nificant properties that are helpful in revealing the modeof action of TCM formula In fact numbers of researchersproposed diversemethods to detect functionalmodules frominteraction networks [11 19 20] On the contrary thereare few researches on identifying pharmacological units formulticomponent drugs [18] On the other hand networkclustering algorithms also known asmodule detectionmeth-ods developed from statistical physics are usually capableof finding significant communities enriched with explicitreal-world meaning [21 22] As a matter of fact severalalgorithms accomplished important tasks in biological fieldsuch as identifying protein complexes [23ndash25] Additionallyto identify functional modules or pharmacological units isobviously another application of network clusteringmethodswhich is crucial in the investigation of pharmacologicalaction of TCM formula Hence applying classic moduledetection algorithms to the molecular network of TCM

formula may contribute to better understanding of its modeof action at the molecular level

We here proposed a computational approach combin-ing clustering algorithm with heterogeneous network toinvestigate the molecular mechanism of TCM formula Thisapproach takes a three-step procedure At first we con-structed a 2-class heterogeneous network (2-HN) comprisedof herbal ingredients and associated targets for a TCM for-mula under studyThen a classicmodule detection algorithmwas applied to the 2-HN and we identified pharmacologicalunits from the 2-HN Finally we finely selected primarypharmacological units and investigated them by pathwayanalysis This approach is apparently applicable for any TCMformula In this paper we use Shu-feng-jie-du formula (SHUformula) as an example to illustrate the procedure of theapproachThepathway analysis of four pharmacological unitsidentified from the 2-HN for SHU formula showed that 24out of 40 enriched pathways that were ranked in top 10corresponding to each of the four pharmacological units weredirectly or indirectly involved in the process of influenzadevelopment

2 Methods

The novel approach is aimed at discovering the molecularmechanism of TCM formula based on a heterogeneousnetwork together with a clustering algorithmThe procedureof this approach mainly consists of three steps the construc-tion of a heterogeneous network module detection fromthe network and the pathway analysis of selected primarypharmacological units In practice we investigated the modeof action of Shu-feng-jie-du formula by using this approach

21 Construct Heterogeneous Network Since our approachtakes advantage of the network to study a TCM formulawe should firstly construct a heterogeneous network com-prised of herbal ingredients and potential targets At thebeginning the specific composition of each herb in a givenTCM formula must be acquired Typical ways to collectthe chemical ingredients of herbs include literature miningTCM database retrieval and identification test By diversemeans we can collect the chemical constituents togetherwith their geometric structure for all herbs in the TCMformula Subsequently the interaction data and potentialtargets could be computed and retrieved respectively basedon the chemical knowledge for the studied TCM formula

First of all we acquire the interaction data betweenherbal compounds by computational chemistry techniquesAlthough various kinds of interaction knowledge is availablecompound pairs with similar chemical structures are widelyused in network-based pharmacology and drug discoveryresearch The rationale is a well-known assumption thatsimilar compounds have similar properties [26] In otherwords similar chemicals may share common targets andare likely to perform synergistic action on complex diseasesThus we evaluate compound pairs by calculating the pairwisechemical similarity using the geometric structure previouslycurated In the field of cheminformatics various methods



119904TC (1198881 1198882) =119888

119886 + 119887 minus 119888

(1)

where 1198881and 119888











119876 (119862) =

119897119862

119898

minus (

119889119862

2119898

)

2

(2)

where 119897119862= sum119894119895isin119862


in module 119862 119889119862

= sum119899isin119862


119894119895isin119866119908119894119895



C5C2 C3

C4C1

C6C7 C8

C9

C2

C3

C4

C6

C9

P1

P3P2

P4 P5

P6P7

P8

P9

P10P11

P2

P3

P4

P5

P9

P7

P10

P11


C4

C6

C7

C8C9

P4

P5

P6

P10P9

P11






Targeted protein

C10 C10






119877 (119862) =

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119862

10038161003816100381610038161003816

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119866

10038161003816100381610038161003816

(3)

























CPIs 0 1101 0 1101PPIs 0 0 718 718








[40]


Kaempferol 520-18-3 5280863 67


[44]


[45]
















Dicumarol Citral

GlycerolCoumarin


Beta-carotene

Resveratrol Eugenol









AKT [(4)]signaling



100 10 1

PI


P value

(a)

Gamma-decalactone




Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin











100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid











100 10 1

PI


P value



(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin











100 10 1

PI


P value



(d)









1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of


















119904TC (1198881 1198882) =119888

119886 + 119887 minus 119888

(1)

where 1198881and 119888











119876 (119862) =

119897119862

119898

minus (

119889119862

2119898

)

2

(2)

where 119897119862= sum119894119895isin119862


in module 119862 119889119862

= sum119899isin119862


119894119895isin119866119908119894119895



C5C2 C3

C4C1

C6C7 C8

C9

C2

C3

C4

C6

C9

P1

P3P2

P4 P5

P6P7

P8

P9

P10P11

P2

P3

P4

P5

P9

P7

P10

P11


C4

C6

C7

C8C9

P4

P5

P6

P10P9

P11






Targeted protein

C10 C10






119877 (119862) =

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119862

10038161003816100381610038161003816

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119866

10038161003816100381610038161003816

(3)

























CPIs 0 1101 0 1101PPIs 0 0 718 718








[40]


Kaempferol 520-18-3 5280863 67


[44]


[45]
















Dicumarol Citral

GlycerolCoumarin


Beta-carotene

Resveratrol Eugenol









AKT [(4)]signaling



100 10 1

PI


P value

(a)

Gamma-decalactone




Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin











100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid











100 10 1

PI


P value



(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin











100 10 1

PI


P value



(d)









1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















C5C2 C3

C4C1

C6C7 C8

C9

C2

C3

C4

C6

C9

P1

P3P2

P4 P5

P6P7

P8

P9

P10P11

P2

P3

P4

P5

P9

P7

P10

P11


C4

C6

C7

C8C9

P4

P5

P6

P10P9

P11






Targeted protein

C10 C10






119877 (119862) =

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119862

10038161003816100381610038161003816

10038161003816100381610038161003816119890119888119892| 119888 119892 isin 119866

10038161003816100381610038161003816

(3)

























CPIs 0 1101 0 1101PPIs 0 0 718 718








[40]


Kaempferol 520-18-3 5280863 67


[44]


[45]
















Dicumarol Citral

GlycerolCoumarin


Beta-carotene

Resveratrol Eugenol









AKT [(4)]signaling



100 10 1

PI


P value

(a)

Gamma-decalactone




Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin











100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid











100 10 1

PI


P value



(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin











100 10 1

PI


P value



(d)









1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of






































CPIs 0 1101 0 1101PPIs 0 0 718 718








[40]


Kaempferol 520-18-3 5280863 67


[44]


[45]
















Dicumarol Citral

GlycerolCoumarin


Beta-carotene

Resveratrol Eugenol









AKT [(4)]signaling



100 10 1

PI


P value

(a)

Gamma-decalactone




Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin











100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid











100 10 1

PI


P value



(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin











100 10 1

PI


P value



(d)









1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of




















CPIs 0 1101 0 1101PPIs 0 0 718 718








[40]


Kaempferol 520-18-3 5280863 67


[44]


[45]
















Dicumarol Citral

GlycerolCoumarin


Beta-carotene

Resveratrol Eugenol









AKT [(4)]signaling



100 10 1

PI


P value

(a)

Gamma-decalactone




Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin











100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid











100 10 1

PI


P value



(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin











100 10 1

PI


P value



(d)









1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of


























Dicumarol Citral

GlycerolCoumarin


Beta-carotene

Resveratrol Eugenol









AKT [(4)]signaling



100 10 1

PI


P value

(a)

Gamma-decalactone




Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin











100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid











100 10 1

PI


P value



(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin











100 10 1

PI


P value



(d)









1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of



















Dicumarol Citral

GlycerolCoumarin


Beta-carotene

Resveratrol Eugenol









AKT [(4)]signaling



100 10 1

PI


P value

(a)

Gamma-decalactone




Carvacrol

Caffeic acid

QuercetinKaempferol

EmodinIndirubin











100 10 1

PI


P value

(b)

Adenosine

Phenol

Betulinic acid











100 10 1

PI


P value



(c)

Gentisic acid

Protocatechuic acid

Pinocembrin

Catechin











100 10 1

PI


P value



(d)









1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of



















1



2



4


5









Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















Eugenol


on G1S transition


regulator


AKT signaling

Resveratrol


via MAPK

Acteoside



Dicumarol



Beta-carotene




Linalool










4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of



















4 Conclusion






Acknowledgments


References




























































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of







































































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of





















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















Documents

Research Article A Module Analysis Approach to Investigate ...downloads.hindawi.com/journals/ecam/2013/731370.pdf · action of TCM formula. Hence, applying classic module detection