research article pathway analysis based on attractor...

Research ArticlePathway Analysis Based on Attractor and Cross Talk inColon Cancer

Yanxia Liu12 Lin Wang1 Bingping Wang2 Meng Yue3 and Yufeng Cheng1

1Department of Radiation Oncology Qilu Hospital Shandong University Jinan Shandong 250012 China2Department of Oncology Shengli Oil Central Hospital Dongying Shandong 257034 China3Department of Gastroenterology Jinan Central Hospital Shandong University Shandong 250013 China

Correspondence should be addressed to Yufeng Cheng chengyufengmedyeahnet

Received 17 February 2016 Revised 5 May 2016 Accepted 12 July 2016

Academic Editor Szilard Nemes

Copyright copy 2016 Yanxia Liu 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

Colon cancer is the third and second most common cancer form in men and women worldwide It is generally accepted that coloncancer mainly results from diet The aim of this study was to identify core pathways which elucidated the molecular mechanismsin colon cancer The microarray data of E-GEOD-44861 was downloaded from ArrayExpress database All human pathwayswere obtained from Kyoto Encyclopedia of Genes and Genomes database In total 135 differential expressed genes (DEG) wereidentified using Linear Models for Microarray Data package Differential pathways were identified with the method of attractorafter overlapping with DEG Pathway cross talk network (PCN) was constructed by combining protein-protein interactions anddifferential pathways Cross talks of all pathways were obtained in PCNThere were 65 pathways with RankProd (RP) values lt 005and 16 pathways with Impact Factors (IF) values gt 100 Five pathways were satisfied with 119875 value lt 005 RP values lt 005 andIF values gt 100 which were considered to be the most important pathways in colon cancer In conclusion the five pathways wereidentified in the center status of colon cancer which may contribute to understanding the mechanism and development of coloncancer

1 Introduction

Colon and rectal cancer are the third most common formsof cancer in the United States [1] Colon cancer is the thirdand second most common cancer form in men and womenworldwide [1] causing appropriate 640 thousand deaths eachyear It is commonly known as colorectal cancer or largebowel cancer It is generally accepted that colon cancermainlyresulted from diet in one way or another [2] Besides it isalso correlated with genetic factors such as family historyof colorectal cancer and familial adenomatous polyposis [3]Also old age [4] gender [5] and presence of adenomatouspolyps [6] are risk factors related to colon cancer Recently aSNP rs5995355 in NCF4 was found significantly associatedwith risk of colorectal cancer after adjustment for bothpotential confounders and multiple comparisons but thechange of expressionwas not found in either tumor or normaltissue [7] Thus elucidating the molecular mechanisms iscritical to clinical diagnosis and treatment for colon cancer

Our purpose of this study is to explore important pathwaysthat reflectedmechanism of the occurrence and developmentof colon cancer by screening differential expressed genes(DEGs) between colon cancer tissues and normal tissues andanalyzing the pathways using biological information

Modernmolecular biology indicates that selective expres-sion of genes controls the regulating mechanism in the biol-ogy DEG that has significant difference at expression levelbetween cancer tissues and normal tissues could conduceto analyzing cancer mechanism Altered pathways betweencancer tissue and normal tissue may help to understandthe disease status and suggest anticancer therapies Basedon microarray data and Kyoto Encyclopedia of Genes andGenomes (KEGG) database numerous researches have ana-lyzed biological processes with genes and pathways by usinga variety of statistical analysis strategies [8ndash10]

Attractor is an analytical approach for identifying andannotating the gene sets that best discriminate betweencell phenotypes [11] It can identify core pathways which

Hindawi Publishing CorporationDisease MarkersVolume 2016 Article ID 2619828 7 pageshttpdxdoiorg10115520162619828

2 Disease Markers

best contrasted the cell types of interest With this methoddifferential pathways between cancer group and normalgroup can be identified

Protein-protein interactions (PPIs) provide valuableinformation about how genes perform functions Network-based methods have been applied to gain insight into themechanism from the interaction data [12] The pathwaysoverlapped with interactional genes are also considered tointeract with each other known as cross talk Pathwayscan affect each other through cross talk rather than work-ing along Cross talk is valuable in understanding diseaseespecially cancers and may play an important role in theinvasion and proliferation of cancer cells [13] Pathway crosstalk network (PCN) constructed with pathways and proteininteractions according to Li et al which was first developed tosearch for colorectal cancer progression andmetastasis basedon transcriptional data can be utilized to analyze genome-wide expression profiling data by analyzing how pathwaysaffect each other and the difference between clusters crosstalk [14] But the results were not sufficient since they didnot present pathway aberrance By combining differentialpathway analysis and PCN the analysis can be used forpathways that not only are significantly altered but alsoinfluence other pathways And this has been applied inbreast cancer to analyze pathways conducted by Sun et al[15]

In this study we tried to explore colon cancer mechanismby analyzing pathways which not only were dysregulated incolon tissues when compared with normal group but alsointeracted with other pathways To achieve this goal geneexpression profiles were downloaded from ArrayExpressdatabase to detect differential expressed genes Human-related pathways were downloaded fromKEGGdatabase andPPIs were downloaded from search tool for the retrievalof interacting genesproteins (STRING) database to identifydifferential pathways and construct PCN

2 Materials and Methods

21 Gene Expression Data

211 Data Resource Microarray data of E-GEOD-44861 [7]alongwith its annotation file was downloaded fromArrayEx-press databaseTherewere 56 colon tumor tissues and 55 adja-cent noncancerous tissues The platform in this study was A-AFFY-113-Affymetrix GeneChip HT Human Genome U133AHT HG-U133A and the title was ldquoAffymetrix expression datafrom colon cancer patient tissuesrdquo

212 Gene Expression Data Preprocessing Microarray ex-pression data should be preprocessed because they aremeasured as intensities Linear Models for Microarray Data(LIMMA) package was chosen to reprocess data with thefunction of expresso [16] And the background was correctedwith robust multichip average (RMA) [17] Normalizationwas performed with quantiles function Then we used MASfor corrected perfect match (PM)mismatch (MM) [18]Medianpolish function was used for summarizing expression

data After probe filtration with featureFilter function 12493genes were obtained

213 Differential Expressed Genes (DEGs) Screening DEGshave become an importantmethod in studying tumor-relatedgenesThey contribute to illuminatingmechanismof a tumorIn this study LIMMA method was applied to screen DEGThe values of |log FC| ge 15 and 119875 value le 001 were selectedas the cut-off criteria

22 Pathway Data The common databases regarding biol-ogy are Kyoto Encyclopedia of Genes and Genomes (KEGG)database (httpwwwkeggjp) [19] and Reactome database(httpwwwreactomeorg) [20] In this study we down-loaded pathways from KEGG database After deleting thepathways containing no genes 294 pathways were obtained

23 Protein-Protein Interaction (PPI) Data PPI data hasbecome an important source of protein function and rela-tionship information in microbiology molecular biologycomputational biology and medicine And it can providevaluable information regarding how genes carry out theirbiological functions [14] The PPI data can be downloadedfrom search tool for the retrieval of interacting genesproteins(STRING) database [21] In total 787896 PPIs were obtained

24 Differential Pathways Analysis To screen differentialpathways attractor method was applied [11]

Genes in normal group and disease group were treatedwith KEGG enrichment analysis Attractors were obtainedwith GSEA-ANOVA an analysis of variance-based imple-mentation of a gene set enrichment algorithm

From this model 119865-statistic of gene was figured out with

119865(119894)=

MSS119894RSS119894 (1)

where MSS119894 denotes the mean treatment sum of squares

MSS119894 =1

119870 minus 1

119870

sum

119896=1

119903119896 [119910(119894)

sdot119896minus 119910(119894)

sdotsdot ]2

(2)

and RSS119894 denotes the residual sum of squares

RSS119894 =1

119873 minus 119870

119870

sum

119896=1

119903119895

sum

119895=1

[119910(119894)

119895119896minus 119910(119894)

sdotsdot ]2 (3)

For pathway 119901 consisting of 119892119901 genes the119879-statistic takes thefollowing form

119879119901 =

[(1119892119901)sum119892119901

119894=1119865(119894)] minus [(1119866)sum

119866

119895=1 119865(119894)]

radic(1198781199012119892119901) + (119878119866

2119866)

(4)

where 119866 denotes the total number of genes in a pathway and1198781199012 and 119878119866

2 were defined as sample aberrancesAfter performing 119879-test and adjusting by false discovery

rate (FDR) of Benjamini-Hochberg [22] 119875 values of eachpathway were obtained

Disease Markers 3

25 Construction of Pathway Cross Talk Network (PCN) Toanalyze interactions between pathways a PCN was con-structed as described by Li et al [14]

251 Interactions It is assumed that cancer-associated genersquosdysregulation of expression can lead to the differential ex-pression of its interacting genes when there is no networkrewiringWe used gene expression correlation tomeasure thedynamic action of the PPIs

In both disease group and background group Spearmancorrelation coefficient of each PPI was calculated with thefollowing formula [23]

119903119864119894119895=

sum119896 (119909119894119896 minus 119909119894) (119909119895119896 minus 119909119895)

radicsum119896 (119909119894119896 minus 119909119894)sum119896 (119909119895119896 minus 119909119895)

(5)

where 119864119894119895 is the PPI between gene119881119894 and gene119881119895 119896 is the 119896thsample 119909119895119896 is the rank of 119881119894 of 119896th sample 119909119894119896 is the rank of119881119895 of 119896th sample 119909119894 and 119909119895 are the average ranks of 119881119894 and 119881119895in the samples respectively

100381610038161003816100381610038161003816119903119864119894119895

100381610038161003816100381610038161003816=

1

2

1003816100381610038161003816100381610038161199031198641198941198951+ 1199031198641198941198952

100381610038161003816100381610038161003816

100381610038161003816100381610038161003816Δ119903119864119894119895

100381610038161003816100381610038161003816=

1003816100381610038161003816100381610038161199031198641198941198951minus 1199031198641198941198952

100381610038161003816100381610038161003816

(6)

where 1199031198641198941198951and 1199031198641198941198952 represent the Spearman coefficients of 119864119894119895in compared samples respectively

The gene pairs are considered to interact intensively if|Δ119903119864119894119895| ge 05 If |Δ119903119864119894119895 | lt 05 but the two genes belong to

differential expressed genes they are also considered to havestrong interactions and should be reserved

252 Weight Weight represents the number of PPI in thenetwork [24] Only gene pairs with weight gt 5 which areconsidered to have strong interactions are recorded and usedto construct a disease-related PCN

253 Degree By analysis of the topological characteristics inthe network all node degrees were obtainedThe degree ratiowas defined as the degree of a node in disease-related PCN tothat in background PCN

We introduced a concept of pathway score to examine thepathway status in the disease The formula was shown as

Pathway score =Degree in diseaseDegree in normal

(7)

26 Comprehensive Analysis of Pathways For a comprehen-sive analysis of disease-related genes and pathways we intro-duced nonparametric rank product (RankProd) approach[25] to find important pathways The formula was shown as

RP = (Rank interTotal

) times (

Rank outerTotal

) (8)

where inter indicated attracting 119875 value of a pathway andouter indicated degree of a pathway

Table 1 Differential pathways with 119875 value lt 005

KEGG ID Term 119875 value05219 Bladder cancer 223119864 minus 05

04080 Neuroactive ligand-receptorinteraction 708119864 minus 05

00071 Fatty acid degradation 000016304740 Olfactory transduction 0000239

04932 Nonalcoholic fatty liver disease(NAFLD) 0000372

00190 Oxidative phosphorylation 000217305322 Systemic lupus erythematosus 000402203010 Ribosome 000873901212 Fatty acid metabolism 000965400920 Sulfur metabolism 001181205012 Parkinsonrsquos disease 001283605033 Nicotine addiction 001293605034 Alcoholism 001293604728 Dopaminergic synapse 002014105206 MicroRNAs in cancer 0025966

00860 Porphyrin and chlorophyllmetabolism 0032464

00520 Amino sugar and nucleotidesugar metabolism 0035002

04110 Cell cycle 0035002

An impact factor (IF) concept was also introduced toexamine the significance of a pathway which was shown as

IF = outer times (1 minus inter) (9)

3 Results

31 DEGAnalysis Based on themicroarray data of E-GEOD-44861 135 DEGs between colon cancer tissues and normaltissues were screened out with |log FC| ge 15 and 119875 value le001 by using the method of LIMMA

32 Differential Pathways Analysis Compared with normalgroup a total of 18 differential pathways with 119875 value lt005 were obtained in the cancer group as shown in Table 1These pathways were colon cancer-related pathways such asldquobladder cancerrdquo pathway and ldquoneuroactive ligand-receptorinteractionrdquo pathway and five pathways were correlated withmetabolism The top nine ranked pathway with significantlevels (119875 value lt 001) were identified which were signifi-cantly different between normal group and cancer group

33 Pathway Cross Talk Network Analysis Pathways andprotein interactions were integrated to a global PCN Inthe network nodes represented as pathways and edgesdenoted cross talk between pathways We analyzed nodedegrees of the network which indicated the connectionsamong pathways Cross talks in background were shown inSupplemental Data 1 (in Supplementary Material available

4 Disease Markers

Table 2 Top ten pathways ranked by pathway scores

Pathways BGdegree

Testdegree

Pathwayscore

PI3K-Akt signalingpathway 291 192 0659794

Pathways in cancer 293 193 0658703Cytokine-cytokine receptorinteraction 289 172 0595156

Proteoglycans in cancer 290 172 0593103HTLV-I infection 292 171 0585616Viral carcinogenesis 289 167 0577855Focal adhesion 291 167 0573883Tuberculosis 291 165 0567010Rap1 signaling pathway 291 163 0560137MAPK signaling pathway 292 163 0558219BG background pathways in normal group degree ratio the ratio of degreein test group to that in BG group HTLV human T-cell leukemia virusMAPK mitogen-activated protein kinase

online at httpdxdoiorg10115520162619828) and crosstalks test groups were shown in Supplemental Data 2

Pathways owning more connections with others indicatethat they are more important in the network In the cancer-related PCN the pathways with large degree indicated theywere more important in the case of cancer

Degree ratio of node degree in test group to that in normalgroup named as pathway score was computed and the scoreswere ranged from 0 to 07 There were 80 pathways withdegree ratios lt 001 of which 58 pathways were with degreeratios = 0 which indicated that the pathway connectionswith others were vanished when people were suffering fromcolon cancerThe important pathways were those with higherpathway scores Top ten ranked pathways with pathwayscores and pathway degree were listed in Table 2

Degrees of pathways were compared between test groupand BG group as shown in Figure 1 Degrees of pathways intest group were much less than degrees of pathways in BGgroup

34 Comprehensive Analysis of Pathways There were twodefinitions that illuminate the status of a pathway in coloncancer RP value and IF value

According to the ranks of degree and119875 value in a pathwayRP value was computed ranging from 0 to 1 There were 65pathways with RP values lt 005 On the basis of 119875 value lt005 18 pathways were eligible as shown in Figure 2 and theywere considered to be important pathways

IF value indicates the significance of a pathway thatperforms in colon cancer Figure 3 displayed all the IF valuesof pathways There were 16 pathways with IF values gt 100whichwere considered to be important pathwaysThe top fiveranked pathways with large values were ldquoneuroactive ligand-receptor interactionrdquo pathway ldquomicroRNAs in cancerrdquo path-way ldquopathways in cancerrdquo pathway ldquocell cyclerdquo pathway andldquohuman T-cell leukemia virus (HTLV-I) infectionrdquo pathway

300

300

250

250 350

200

200

150

150

100

100

50

50

0

0

Pathway

(Deg

ree)

Degree of BGDegree of test

Figure 1 Degrees of pathways in background group (BG) and testgroup Red color indicates test group and black color indicates BG

Pathway191817161514131211109876543210

Valu

e

P valueRP value

005

004

003

002

001

000

Figure 2 RankProd (RP) values of the 18 pathways with 119875 values lt005 Color in red indicates RP value and color in black indicates 119875values

After comparing pathways with RP values 119875 values andIF values five pathways were obtained which were identi-fied important in all three factors ldquobladder cancerrdquo path-way ldquoalcoholismrdquo pathway ldquodopaminergic synapserdquo pathwayldquomicroRNAs in cancerrdquo pathway and ldquocell cyclerdquo pathway

4 Discussion

Due to the fast development of bioinformatics network-based approaches have become increasingly important tosearch for cancer mechanisms [26] such as coexpressionnetwork and PPI PCN based on pathways and PPI playsa key role in identifying important pathways To explore

Disease Markers 5Im

pact

fact

or

Pathway350300250200150100500

160

140

120

100

80

60

40

20

0

Figure 3 Impact factor (IF) values of the 300 pathways

mechanism of colon cancer PCN was applied to search forcore pathways

The microarray data of E-GEOD-44861 was selected toexplore mechanism of colon cancer since it was a represen-tative study in recent years In the study of Ryan et al [7]the focus point was NADPH-related pathways while in thisstudy we mainly focused on searching for core dysregulatedpathways in colon cancer from all human-related pathways

With gene expression profiles of 56 cancer tissues and 55normal adjacent tissues 18 differential pathways were iden-tified by attractor procedure These pathways were greatlychanged in cancer tissues compared with normal tissues

As pathways function with each other and do not workalone cross talk in pathways is needed for analysis In thepathway-based PCN degree indicates connections betweenpathways RankProd method was applied to rank pathwaysin degree and 119875 value generating two factors RP value andimpact factor After comprehensive analyses of 119875 value RPvalue and IF values we identified 5 pathways as impor-tant pathways They were not only with significant changesbetween cancer and normal tissues but also with manyconnections with other pathways Once 1 of the 5 pathwayschanged pathways connected with it will be influenced

ldquoBladder cancerrdquo pathway was identified to be asso-ciated with colon cancer This pathway mainly partici-pated in urothelial carcinoma Urothelial tumors arise andevolve through divergent phenotypic pathways Some tumorsprogress from urothelial hyperplasia to low-grade noninva-sive superficial papillary tumors More aggressive variantseither arise from flat high-grade carcinoma in situ (CIS) andprogress to invasive tumors or arise as invasive tumors Low-grade papillary tumors frequently show a constitutive activa-tion of the receptor tyrosine kinase-Ras pathway exhibitingactivating mutations in the HRAS and fibroblast growthfactor receptor 3 (FGFR3) genes In contrast CIS and invasivetumors frequently show alterations in the TP53 and RBgenes and pathways Invasion and metastases are promotedby several factors that alter the tumor microenvironmentincluding the aberrant expression of E-cadherins (E-cad)

matrix metalloproteinases (MMPs) and angiogenic factorssuch as vascular endothelial growth factor (VEGF) [27]

ldquoAlcoholismrdquo pathway is a chronic relapsing disorderrelated pathway Alcoholism is progressive and has seriousdetrimental health outcomes As one of the primary media-tors of the rewarding effects of alcohol dopaminergic ventraltegmental area (VTA) projections to the nucleus accumbens(NAc) have been identified Acute exposure to alcohol stim-ulates dopamine release into the NAc which activates D1receptors stimulating PKA signaling and subsequent CREB-mediated gene expression whereas chronic alcohol exposureleads to an adaptive downregulation of this pathway inparticular of CREB function The decreased CREB functionin the NAc may promote the intake of drugs of abuse toachieve an increase in reward and thus may be involved inthe regulation of positive affective states of addiction PKAsignaling also affects NMDA receptor activity and may playan important role in neuroadaptation in response to chronicalcohol exposure [28]

ldquoDopaminergic synapserdquo pathway is correlated with ner-vous system Dopamine (DA) is an important and prototyp-ical slow neurotransmitter in the mammalian brain where itcontrols a variety of functions including locomotor activitymotivation and reward learning andmemory and endocrineregulation Once released from presynaptic axonal terminalsDA interacts with at least five receptor subtypes in thecentral nervous system (CNS) Through diverse cAMP- andCa2+-dependent and Ca2+-independent mechanisms DAinfluences neuronal activity synaptic plasticity and behaviorPresynaptically localized D2Rs regulate synthesis and releaseof DA as the main autoreceptor of the dopaminergic system[29 30]

ldquoMicroRNAs in cancerrdquo pathway is involved in a clusterof small nonencoding RNAmolecules of 21ndash23 nucleotides inlength which controls gene expression posttranscriptionallyvia either the degradation of target mRNAs or the inhibitionof protein translation Using high-throughput profiling dys-regulation of miRNAs has been widely observed in differentstages of cancer The upregulation of specific miRNAs couldresult in the repression of tumor suppressor gene expres-sion and conversely the downregulation of specific miRNAscould lead to an increase of oncogene expression boththese situations result in tumor growth and progress ThemiRNA signatures of cancer observed in various studies differsignificantlyThese inconsistencies result from the differencesin the study populations and methodologies [31 32]

ldquoCell cyclerdquo pathway functions in response to DNA dam-age by activating signaling pathways that promote cell cyclearrest and DNA repair When responding to DNA damagethe checkpoint kinase ATM phosphorylates and activatesChk2which in turn directly phosphorylates and activates p53tumor suppressor protein p53 and its transcriptional targetsplay an important role in both G1 and G2 checkpoints [33]ATR-Chk1-mediated protein degradation of Cdc25A proteinphosphatase is also a mechanism conferring intra-S-phasecheckpoint activation [34]

Besides some identified pathways were involved in theregulation of NADPH such as ldquofatty acid metabolismrdquo and

6 Disease Markers

ldquooxidative phosphorylationrdquo which was similar to the resultsof Ryan et al [7]

Still there are limitations in the paper The results weregenerated from bioinformatics analysis which still needclinical data to be verified

5 Conclusion

In this study we concentrated on exploring important path-ways that reflected mechanism of the occurrence and devel-opment of colon cancer We identified DEG and differentialpathways by comparing colon cancer tissues and normaltissues After comparing PCN between cancer and normalwe identified 5 important pathways which may give newinsights into the underlying biological mechanisms drivingthe progression of colon cancer and should be paid closeattention to in further research

Competing Interests

The authors declare no competing interests

Acknowledgments

The authors thank Lijie Wang for the help in the preparationof the paper

References

[1] R L Siegel K D Miller and A Jemal ldquoCancer statistics 2016rdquoCA ACancer Journal for Clinicians vol 66 no 1 pp 7ndash30 2016

[2] CMettlin ldquoRecent developments in colorectal cancer epidemi-ology and early detectionrdquo Current Opinion in Oncology vol 2no 4 pp 731ndash737 1990

[3] M B Terry A I Neugut R M Bostick et al ldquoRisk factorsfor advanced colorectal adenomas a pooled analysisrdquo CancerEpidemiology Biomarkers and Prevention vol 11 no 7 pp 622ndash629 2002

[4] D Lieberman ldquoScreening surveillance and prevention ofcolorectal cancerrdquo Gastrointestinal Endoscopy Clinics of NorthAmerica vol 18 no 3 pp 595ndash605 2008

[5] R E van Gelder J Florie and J Stoker ldquoColorectal cancerscreening and surveillance with CT colonography currentcontroversies and obstaclesrdquo Abdominal Imaging vol 30 no 1pp 5ndash12 2005

[6] J F Noguera Aguilar J C Vicens Arbona RMorales Soriano etal ldquoLiver resection inmetastatic colorectal cancer amultidisci-plinary approachrdquoRevista Espanola de EnfermedadesDigestivasvol 97 no 11 pp 786ndash793 2005

[7] B M Ryan K A Zanetti A I Robles et al ldquoGermlinevariation in NCF4 an innate immunity gene is associated withan increased risk of colorectal cancerrdquo International Journal ofCancer vol 134 no 6 pp 1399ndash1407 2014

[8] S Setia B Nehru and S N Sanyal ldquoUpregulation ofMAPKErk and PI3KAkt pathways in ulcerative colitis-associated colon cancerrdquo Biomedicine and Pharmacotherapyvol 68 no 8 pp 1023ndash1029 2014

[9] E-P I Chiang S-Y Tsai Y-H Kuo et al ldquoCaffeic acidderivatives inhibit the growth of colon cancer involvement of

the PI3-KAkt and AMPK signaling pathwaysrdquo PLoS ONE vol9 no 6 Article ID e99631 2014

[10] M T Do M Na H G Kim et al ldquoIlimaquinone induces deathreceptor expression and sensitizes human colon cancer cells toTRAIL-induced apoptosis through activation of ROS-ERKp38MAPK-CHOP signaling pathwaysrdquo Food and Chemical Toxicol-ogy vol 71 pp 51ndash59 2014

[11] J C Mar N A Matigian J Quackenbush and C A WellsldquoAttract a method for identifying core pathways that definecellular phenotypesrdquo PLoS ONE vol 6 no 10 Article IDe25445 2011

[12] A-L Barabasi and Z N Oltvai ldquoNetwork biology understand-ing the cellrsquos functional organizationrdquo Nature Reviews Geneticsvol 5 no 2 pp 101ndash113 2004

[13] E W Bradley M M Ruan A Vrable and M J OurslerldquoPathway crosstalk between RasRAF and PI3K in promotionof M-CSF-induced MEKERK-mediated osteoclast survivalrdquoJournal of Cellular Biochemistry vol 104 no 4 pp 1439ndash14512008

[14] Y Li P Agarwal and D Rajagopalan ldquoA global pathwaycrosstalk networkrdquo Bioinformatics vol 24 no 12 pp 1442ndash14472008

[15] Y Sun K Yuan P Zhang R Ma Q-W Zhang and X-STian ldquoCrosstalk analysis of pathways in breast cancer using anetwork model based on overlapping differentially expressedgenesrdquo Experimental and Therapeutic Medicine vol 10 no 2pp 743ndash748 2015

[16] M E Ritchie B Phipson D Wu et al ldquoLimma powers differ-ential expression analyses for RNA-sequencing and microarraystudiesrdquo Nucleic Acids Research vol 43 no 7 article e47 2015

[17] L Ma L N Robinson and H C Towle ldquoChREBPlowastMlx is theprincipal mediator of glucose-induced gene expression in theliverrdquo The Journal of Biological Chemistry vol 281 no 39 pp28721ndash28730 2006

[18] B M Bolstad R A Irizarry M Astrand and T P Speed ldquoAcomparison of normalizationmethods for high density oligonu-cleotide array data based on variance and biasrdquo Bioinformaticsvol 19 no 2 pp 185ndash193 2003

[19] M Kanehisa S Goto Y Sato M Furumichi and M TanabeldquoKEGG for integration and interpretation of large-scale molec-ular data setsrdquo Nucleic Acids Research vol 40 no 1 pp D109ndashD114 2012

[20] D Croft G OrsquoKelly G Wu et al ldquoReactome a database ofreactions pathways and biological processesrdquo Nucleic AcidsResearch vol 39 no 1 pp D691ndashD697 2011

[21] A Franceschini D Szklarczyk S Frankild et al ldquoSTRING v91protein-protein interaction networks with increased coverageand integrationrdquoNucleic Acids Research vol 41 no 1 pp D808ndashD815 2013

[22] Y Benjamini D Drai G Elmer N Kafkafi and I Golani ldquoCon-trolling the false discovery rate in behavior genetics researchrdquoBehavioural Brain Research vol 125 no 1-2 pp 279ndash284 2001

[23] C Wu J Zhu and X Zhang ldquoIntegrating gene expressionand protein-protein interaction network to prioritize cancer-associated genesrdquo BMC Bioinformatics vol 13 article 182 2012

[24] J-M Wang J-T Wu D-K Sun P Zhang and L WangldquoPathway crosstalk analysis based on protein-protein networkanalysis in prostate cancerrdquo European Review for Medical andPharmacological Sciences vol 16 no 9 pp 1235ndash1242 2012

[25] F Hong R Breitling C W McEntee B S Wittner J LNemhauser and J Chory ldquoRankProd a bioconductor package

Disease Markers 7

for detecting differentially expressed genes in meta-analysisrdquoBioinformatics vol 22 no 22 pp 2825ndash2827 2006

[26] A del Sol R Balling L Hood and D Galas ldquoDiseases asnetwork perturbationsrdquo Current Opinion in Biotechnology vol21 no 4 pp 566ndash571 2010

[27] A P Mitra R H Datar and R J Cote ldquoMolecular pathwaysin invasive bladder cancer new insights into mechanisms pro-gression and target identificationrdquo Journal of Clinical Oncologyvol 24 no 35 pp 5552ndash5564 2006

[28] R Spanagel ldquoAlcoholism a systems approach from molecularphysiology to addictive behaviorrdquo Physiological Reviews vol 89no 2 pp 649ndash705 2009

[29] J-M Beaulieu and R R Gainetdinov ldquoThe physiology signal-ing and pharmacology of dopamine receptorsrdquo Pharmacologi-cal Reviews vol 63 no 1 pp 182ndash217 2011

[30] K A Neve J K Seamans and H Trantham-DavidsonldquoDopamine receptor signalingrdquo Journal of Receptors and SignalTransduction vol 24 no 3 pp 165ndash205 2004

[31] S Babashah and M Soleimani ldquoThe oncogenic and tumoursuppressive roles of microRNAs in cancer and apoptosisrdquoEuropean Journal of Cancer vol 47 no 8 pp 1127ndash1137 2011

[32] Q Wang S Wang H Wang P Li and Z Ma ldquoMicroRNAsnovel biomarkers for lung cancer diagnosis prediction andtreatmentrdquo Experimental Biology and Medicine vol 237 no 3pp 227ndash235 2012

[33] M L Smith I-T Chen Q Zhan et al ldquoInteraction of thep53-regulated protein GADD45 with proliferating cell nuclearantigenrdquo Science vol 266 no 5189 pp 1376ndash1380 1994

[34] G Lolli and L N Johnson ldquoCAK-cyclin-dependent activatingkinase a key kinase in cell cycle control and a target for drugsrdquoCell Cycle vol 4 no 4 pp 572ndash577 2005

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

2 Disease Markers

best contrasted the cell types of interest With this methoddifferential pathways between cancer group and normalgroup can be identified

Protein-protein interactions (PPIs) provide valuableinformation about how genes perform functions Network-based methods have been applied to gain insight into themechanism from the interaction data [12] The pathwaysoverlapped with interactional genes are also considered tointeract with each other known as cross talk Pathwayscan affect each other through cross talk rather than work-ing along Cross talk is valuable in understanding diseaseespecially cancers and may play an important role in theinvasion and proliferation of cancer cells [13] Pathway crosstalk network (PCN) constructed with pathways and proteininteractions according to Li et al which was first developed tosearch for colorectal cancer progression andmetastasis basedon transcriptional data can be utilized to analyze genome-wide expression profiling data by analyzing how pathwaysaffect each other and the difference between clusters crosstalk [14] But the results were not sufficient since they didnot present pathway aberrance By combining differentialpathway analysis and PCN the analysis can be used forpathways that not only are significantly altered but alsoinfluence other pathways And this has been applied inbreast cancer to analyze pathways conducted by Sun et al[15]

In this study we tried to explore colon cancer mechanismby analyzing pathways which not only were dysregulated incolon tissues when compared with normal group but alsointeracted with other pathways To achieve this goal geneexpression profiles were downloaded from ArrayExpressdatabase to detect differential expressed genes Human-related pathways were downloaded fromKEGGdatabase andPPIs were downloaded from search tool for the retrievalof interacting genesproteins (STRING) database to identifydifferential pathways and construct PCN

2 Materials and Methods

21 Gene Expression Data

211 Data Resource Microarray data of E-GEOD-44861 [7]alongwith its annotation file was downloaded fromArrayEx-press databaseTherewere 56 colon tumor tissues and 55 adja-cent noncancerous tissues The platform in this study was A-AFFY-113-Affymetrix GeneChip HT Human Genome U133AHT HG-U133A and the title was ldquoAffymetrix expression datafrom colon cancer patient tissuesrdquo

212 Gene Expression Data Preprocessing Microarray ex-pression data should be preprocessed because they aremeasured as intensities Linear Models for Microarray Data(LIMMA) package was chosen to reprocess data with thefunction of expresso [16] And the background was correctedwith robust multichip average (RMA) [17] Normalizationwas performed with quantiles function Then we used MASfor corrected perfect match (PM)mismatch (MM) [18]Medianpolish function was used for summarizing expression

data After probe filtration with featureFilter function 12493genes were obtained

213 Differential Expressed Genes (DEGs) Screening DEGshave become an importantmethod in studying tumor-relatedgenesThey contribute to illuminatingmechanismof a tumorIn this study LIMMA method was applied to screen DEGThe values of |log FC| ge 15 and 119875 value le 001 were selectedas the cut-off criteria

22 Pathway Data The common databases regarding biol-ogy are Kyoto Encyclopedia of Genes and Genomes (KEGG)database (httpwwwkeggjp) [19] and Reactome database(httpwwwreactomeorg) [20] In this study we down-loaded pathways from KEGG database After deleting thepathways containing no genes 294 pathways were obtained

23 Protein-Protein Interaction (PPI) Data PPI data hasbecome an important source of protein function and rela-tionship information in microbiology molecular biologycomputational biology and medicine And it can providevaluable information regarding how genes carry out theirbiological functions [14] The PPI data can be downloadedfrom search tool for the retrieval of interacting genesproteins(STRING) database [21] In total 787896 PPIs were obtained

24 Differential Pathways Analysis To screen differentialpathways attractor method was applied [11]

Genes in normal group and disease group were treatedwith KEGG enrichment analysis Attractors were obtainedwith GSEA-ANOVA an analysis of variance-based imple-mentation of a gene set enrichment algorithm

From this model 119865-statistic of gene was figured out with

119865(119894)=

MSS119894RSS119894 (1)

where MSS119894 denotes the mean treatment sum of squares

MSS119894 =1

119870 minus 1

119870

sum

119896=1

119903119896 [119910(119894)

sdot119896minus 119910(119894)

sdotsdot ]2

(2)

and RSS119894 denotes the residual sum of squares

RSS119894 =1

119873 minus 119870

119870

sum

119896=1

119903119895

sum

119895=1

[119910(119894)

119895119896minus 119910(119894)

sdotsdot ]2 (3)

For pathway 119901 consisting of 119892119901 genes the119879-statistic takes thefollowing form

119879119901 =

[(1119892119901)sum119892119901

119894=1119865(119894)] minus [(1119866)sum

119866

119895=1 119865(119894)]

radic(1198781199012119892119901) + (119878119866

2119866)

(4)

where 119866 denotes the total number of genes in a pathway and1198781199012 and 119878119866

2 were defined as sample aberrancesAfter performing 119879-test and adjusting by false discovery

rate (FDR) of Benjamini-Hochberg [22] 119875 values of eachpathway were obtained

Disease Markers 3




119903119864119894119895=

sum119896 (119909119894119896 minus 119909119894) (119909119895119896 minus 119909119895)


(5)


100381610038161003816100381610038161003816119903119864119894119895

100381610038161003816100381610038161003816=

1

2

1003816100381610038161003816100381610038161199031198641198941198951+ 1199031198641198941198952

100381610038161003816100381610038161003816

100381610038161003816100381610038161003816Δ119903119864119894119895

100381610038161003816100381610038161003816=

1003816100381610038161003816100381610038161199031198641198941198951minus 1199031198641198941198952

100381610038161003816100381610038161003816

(6)








(7)



) times (

Rank outerTotal

) (8)










04110 Cell cycle 0035002



3 Results




4 Disease Markers


Pathways BGdegree

Testdegree

Pathwayscore











300

300

250

250 350

200

200

150

150

100

100

50

50

0

0

Pathway

(Deg

ree)



Pathway191817161514131211109876543210

Valu

e

P valueRP value

005

004

003

002

001

000



4 Discussion


Disease Markers 5Im

pact

fact

or

Pathway350300250200150100500

160

140

120

100

80

60

40

20

0













6 Disease Markers



5 Conclusion


Competing Interests


Acknowledgments


References



























Disease Markers 7
















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















Disease Markers 3




119903119864119894119895=

sum119896 (119909119894119896 minus 119909119894) (119909119895119896 minus 119909119895)


(5)


100381610038161003816100381610038161003816119903119864119894119895

100381610038161003816100381610038161003816=

1

2

1003816100381610038161003816100381610038161199031198641198941198951+ 1199031198641198941198952

100381610038161003816100381610038161003816

100381610038161003816100381610038161003816Δ119903119864119894119895

100381610038161003816100381610038161003816=

1003816100381610038161003816100381610038161199031198641198941198951minus 1199031198641198941198952

100381610038161003816100381610038161003816

(6)








(7)



) times (

Rank outerTotal

) (8)










04110 Cell cycle 0035002



3 Results




4 Disease Markers


Pathways BGdegree

Testdegree

Pathwayscore











300

300

250

250 350

200

200

150

150

100

100

50

50

0

0

Pathway

(Deg

ree)



Pathway191817161514131211109876543210

Valu

e

P valueRP value

005

004

003

002

001

000



4 Discussion


Disease Markers 5Im

pact

fact

or

Pathway350300250200150100500

160

140

120

100

80

60

40

20

0













6 Disease Markers



5 Conclusion


Competing Interests


Acknowledgments


References



























Disease Markers 7
















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















4 Disease Markers


Pathways BGdegree

Testdegree

Pathwayscore











300

300

250

250 350

200

200

150

150

100

100

50

50

0

0

Pathway

(Deg

ree)



Pathway191817161514131211109876543210

Valu

e

P valueRP value

005

004

003

002

001

000



4 Discussion


Disease Markers 5Im

pact

fact

or

Pathway350300250200150100500

160

140

120

100

80

60

40

20

0













6 Disease Markers



5 Conclusion


Competing Interests


Acknowledgments


References



























Disease Markers 7
















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















Disease Markers 5Im

pact

fact

or

Pathway350300250200150100500

160

140

120

100

80

60

40

20

0













6 Disease Markers



5 Conclusion


Competing Interests


Acknowledgments


References



























Disease Markers 7
















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















6 Disease Markers



5 Conclusion


Competing Interests


Acknowledgments


References



























Disease Markers 7
















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















Disease Markers 7
















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of





















of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















research article pathway analysis based on attractor...

Documents