research article integrating gene expression and protein...

Research ArticleIntegrating Gene Expression and Protein Interaction Data forSignaling Pathway Prediction of Alzheimerrsquos Disease

Wei Kong1 Jingmao Zhang1 Xiaoyang Mou2 and Yang Yang3

1 Information Engineering College Shanghai Maritime University Shanghai 201306 China2DNJ Pharma and Rowan University Glassboro NJ 08028 USA3Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai 200240 China

Correspondence should be addressed to Wei Kong weikongshmtueducn

Received 23 February 2014 Accepted 18 March 2014 Published 9 April 2014

Academic Editor Tao Huang

Copyright copy 2014 Wei Kong et alThis is an open access article distributed under theCreativeCommonsAttribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Discovering the signaling pathway and regulatory network would provide significant advance in genome-wide understanding ofpathogenesis of human diseases Despite the rich transcriptome data the limitation for microarray data is unable to detect changesbeyond transcriptional level and insufficient in reconstructing pathways and regulatory networks In our study protein-proteininteraction (PPI) data is introduced to addmolecular biological information for predicting signaling pathway of Alzheimerrsquos disease(AD) Combining PPI with gene expression data significant genes are selected by modified linear regression model firstly Thenaccording to the biological researches that inflammation reaction plays an important role in the generation and deterioration ofAD NF-120581B (nuclear factor-kappa B) as a significant inflammatory factor has been selected as the beginning gene of the predictingsignaling pathway Based on that integer linear programming (ILP) model is proposed to reconstruct the signaling pathwaybetween NF-120581B and AD virulence gene APP (amyloid precursor protein) The results identify 6 AD virulence genes included inthe predicted inflammatory signaling pathway and a large amount of molecular biological analysis shows the great understandingof the underlying biological process of AD

1 Introduction

Alzheimerrsquos disease (AD) is a progressive and fatal neurode-generative disorder manifested by cognitive and memorydeteriorationThe characteristic pathology changes inAD arefibrin deposition in cerebral cortex it is the deposition ofbeta-amyloid (A120573) in cell space and poly-Tau protein in cellIn pathomorphism the expression is senile plaques (SP) andneurofibrillary tangles (NFT)

Many studies have investigated the mechanism of ADfromvarious perspectives of its complexity Recent researchesshow that a more accepted hallmark of AD is brain inflam-mation Inflammation clearly occurs in pathologically vul-nerable regions of AD brain and it does so with the fullcomplexity of local peripheral inflammatory responses [1ndash3]In the periphery degenerating tissue and the deposition ofhighly insoluble abnormal materials are classical stimulantsof inflammation Likewise in the AD brain damaged neurons

and neurites and highly insoluble A120573 peptide deposits andneurofibrillary tangles provide obvious stimuli for inflamma-tion [4ndash7]

To give insight to the AD mechanisms high-throughputgene expression data has received extensive attention andmade substantial progress in reconstructing the gene regu-latory network However due to the underlying shortcom-ings of microarray technology such as small sample sizemeasurement error and information insufficiency unveilingdiseasemechanismhas remained amajor challenge to theADresearch community To overcome these problems pathwayinformation and network-based approaches [8] have beenapplied and become more informative and powerful fordiscovering disease mechanism

Protein-protein interaction (PPI) networks are recon-structed from protein domain characteristics gene expres-sion data and structure-based information with other evi-dence for example gene homology function annotations

Hindawi Publishing CorporationComputational and Mathematical Methods in MedicineVolume 2014 Article ID 340758 7 pageshttpdxdoiorg1011552014340758

2 Computational and Mathematical Methods in Medicine

and sequence motifs [9] PPI data contain structure informa-tion among different genes while gene expression data do notIn our study PPI network data as a priori pathway informa-tion is introduced for predicting the inflammatory signalingpathway in AD Many literatures have given outstandingachievements by integrating gene expression data and PPIdata such as identification of protein complexes [10] smallsubnetworks [11] and biomarkers [12] Zhao et al presentedan integer linear programming (ILP) method to uncoverpathways among the given starting proteins ending proteinsand some transduction factor proteins [13] However howto select the transduction factor proteins is a great problemIn our study a modified network-constrained regularizationanalysismethod [14] is proposed for linear regression analysisto select appropriate number of significant genes Simulationresults show that this method can lead to an efficiently globalsmoothness of regression coefficients

Based on that ILP model is presented to reconstructthe inflammatory signaling pathway by integrating PPI datawith the AD gene expression data In the ILP model thestarting and ending proteins of the predicting pathway needto be arranged in advance Nuclear transcription factor NF-120581B (nuclear factor-kappa B) as one of the most importantinflammatory factors is selected as the starting gene of thesignaling pathway As we know that NF-120581B plays a key rolein regulating the immune response to infection thereforeincorrect regulation of NF-120581B has been linked to cancerinflammatory and autoimmune diseases septic shock viralinfection and improper immune development NF-120581B hasalso been implicated in processes of synaptic plasticity andmemory [15] On the other hand APP (amyloid precursorprotein) as the most important AD virulence gene andprecursor protein of A120573 is arranged as the ending protein ofthe predicting pathway

The experiment results show that 6 AD virulence genesare identified being included in the predicted inflammatorysignaling pathway and a large amount of inflammationrelated genes and pathways has been found by molecularbiological analysis and they show the great understanding ofthe pathogenesis of AD

2 Methods

21 Linear Regression Model Linear regression model iswidely used in estimation and variable selection In ourstudy the model is applied to selected subset of significantgenes which are important for AD and are going to be thetransduction factors of reconstructing pathway In the nextprediction step gene expression data and PPI data will beintegrated by ILP model After all of the above a pathwaycould be identified between NF-120581B and APPThe usual linearregression model can be expressed as

120583 =

119901

sum

119895=1

119909119895120573119895 = 11990911205731 + 11990921205732 + sdot sdot sdot + 119909119901120573119901 (1)

where 120583 = (1205831 1205832 120583119899)119879 is response vector 119899 is the

sample number 119909119895 = (1199091 1199092 119909119899)119879 119895 = 1 2 119901 is the

predictor represented by the 119895th genersquos expression data in allsamples and 120573119895 is the 119895th genersquos weight vector Assume thatthe predictors are standardized and the response is centeredwe get

119899

sum

119894=1

120583119894 = 0

119899

sum

119894=1

119909119894119895 = 0

119899

sum

119894=1

1199092

119894119895= 0

for 119895 = 1 2 119901

(2)

Gene expression data has the characteristic of less sampleand great noise As a simple model linear regression modelhas significant performance in handling less sample and greatnoise data The significant genes will get a larger coefficientwhile the nonsignificant genes will get a smaller coefficient

22 Network-Constrained Regularization for the LinearRegression Model Before using linear regression modelcoefficient 120573 needs to be estimated Manymethods have beenproposed which focused on addressing high-dimensionalitygenomic data such as LASSO LA-SEN and LARS Here amodified network-constrain regularization analysis by CLi and H Li [14] is applied to estimate the coefficient sinceit has been proved to perform better than other methodsThis method is a lasso-type problem It defines a normalizedLaplacian matrix 119871 as

119871 =

1 minus 119908 (119906 V)119889119906

if 119906 = V and 119889119906 = 0

minus119908 (119906 V)radic119889119906119889V

if 119906 and V are adjacent

0 otherwise

(3)

where 119908(119906 V) represents the weight of edge between linkedgenes 119906 and V 119889V = sum119906simV 119908(119906 V) represents all the adjacentgenes of V on the networkThen the definition of the network-constrained regularization criterion is

119871 (1205821 1205822 120573) = (120583 minus 119883120573)119879(120583 minus 119883120573) + 1205821

100381610038161003816100381612057310038161003816100381610038161+ 1205822120573

119879119871120573

(4)

where119883 = (1199091|119870|119909119901) |1205731| = sum119901

119895=1|120573119895|12058211205822 are nonnegative

turning parameters And then we estimate 120573 by minimizing(4)

120573 = argmin120573

119871 (1205821 1205822 120573) (5)

Minimizing (4) is equivalent to solving a lasso-typeoptimization problem Turning parameters are estimatedby 10-fold cross-validation (CV) Genes in gene interactionnetwork are selected by PubGene we chose genes related toAlzheimer

23 Integer Linear Programming (ILP) The ILP model for-mulates signaling network detection as an optimizationproblem and treats a signaling network as a whole entityas described in its original publication [13] PPI networkis a weighted undirected graph that can be described as

Computational and Mathematical Methods in Medicine 3

G(VEW) where V is vertices in the graph representingprotein E is edge between proteins and W represents theweight of edgesW can be calculated by gene expression dataThe ILP model can be described as follows

Minmize119909119894 119910119894119895

119878 = minus

|119881|

sum

119894=1

|119881|

sum

119895=1

119908119894119895119910119894119895 + 120582

|119881|

sum

119894=1

|119881|

sum

119895=1

119910119894119895

Subject to 119910119894119895 le 119909119894

119910119894119895 le 119909119895

|119881|

sum

119895=1

119910119894119895 ge 1 if 119894 is either a starting

or ending protein

|119881|

sum

119895=1

119910119894119895 ge 2119909119894 if 119894 is not a starting

or ending protein

119909119894 = 1 if 119894 is a protein known in STN

119909119894 isin 0 1 119894 = 1 2 |119881|

119910119894119895 isin 0 1 119894 119895 = 1 2 |119881|

(6)

where 119908119894119895 is the weight between proteins 119894 and 119895 in weightedundirected graph119866 119910119894119895 is a binary variable to denote whetherthe edge119864(119894 119895) is a part of the STN 119909119894 is also a binary variableto denote whether protein 119894 is a component of the STN 120582 isa positive penalty parameter |119881| includes all proteins in thePPI network 119910119894119895 le 119909119894 and 119910119894119895 le 119909119895 mean that only if proteins119894 and 119895 are both components of STN the edge 119864(119894 119895) shouldbe considered sum|119881|

119895=1119910119894119895 ge 1 represents at least one protein

contact with starting protein or ending proteinsum|119881|119895=1

119910119894119895 ge 2119909119894

makes sure that if 119909119894 is selected as a component of STN thereare at least two proteins link to the vertex

The starting protein and ending protein have confirmedabove that the genes selected by linear model were treated astransduction factors The method of chosen parameter 120582 canbe found in its original publicationThen we detected proteinpathway by the ILP model

3 Results and Discussion

To evaluate ILP model AD dataset series GSE1297 wasusedwhichwere humanhippocampal gene expression down-loaded from GEO DataSets from the National Center forBiotechnology Information (NCBI) offered by Blalock et al[16] The hippocampal specimens they used are obtainedthrough the Brain Bank of the Alzheimerrsquos Disease ResearchCenter at the University of KentuckyThe humanGene Chips(HG-U133A) of Affymetrix and Microarray Suite 5 are usedin analyzing the microarray data There are a total of 9control 7 incipient 8 moderate and 7 severe AD samples

included in this dataset with 22283 gene expressions in eachsample The PPI data we used is downloaded from websiteBioGRID (httpthebiogridorg) with 12466 proteins and40323 interactions in total

The file format of microarray data downloaded fromNCBI is CEL The probe data needs data processing likebackground correct normalization probe correct and soon Then ANOVAs were used on preliminary select genesand removed all genes whose 119875 value was less than005 After processing 7030 genes for each sample wereleft Then taking linear regression model with modifiednetwork-constrained regularization and AD biological infor-mation the coefficients 120573 = (1205731 1205732 120573119901) and 119901 =

7030 were obtained Among them 7017 values of 120573

were zeroes and the other 13 120573119904 were nonzero valuesSo these 13 genes with nonzero values of 120573 were consid-ered as significant genes to AD phenotype and they aredenoted in green circles with ldquo(1)rdquo and gene names inFigure 1

The 13 selected genes can be mapped to PPI network toget the corresponding proteins and the interactions betweenthem and other proteins Each selected gene was connectedwith some other genes in PPI network by the edges In ILPalgorithm edge between 119894 and 119895was represented by 119910119894119895WhenILP chose this edge 119910119894119895 = 1 otherwise 119910119894119895 = 0

ILP tries to assign 0 or 1 for119910119894119895 to ensure the result networkhas the largest weight For the weight of edge 119908119894119895 here weuse the Pearson coefficient of the gene expression values torepresent the weight between proteins 119894 and 119895

Then using NF-120581B as starting protein and APP as endingprotein ILP model was applied to formulate the signalingnetwork In the ILP algorithm penalty parameter 120582 is a sizecontrol parameter that needs to be adapted manually If itsvalue is too large the predicted signaling network will beenormous otherwise it will be too small to catch the usefulbiological information In our simulation experiment afteradapting from small value to large value 120582 was determinedas 065

We finally got a signaling pathway with several smallsubnetworks This network is reconstructed by 45 genesincluding 13 selected significant genes and is shown inFigure 1

In Figure 1 ldquo(0) NF-120581Brdquo represents the starting pro-tein NF-120581B ldquo(3) APPrdquo represents the ending protein APPldquo(1)rdquo with the protein names denote the correspondingselected genes by the regress model and ldquo(2)rdquo with pro-tein names are selected by ILP to reconstruct the sig-naling pathways between NF-120581B and APP In order toanalyze the biological functions of the pathways and sub-networks the predicted result was mapped into its cod-ing gene pathway network and the online analysis web-site DAVID (httpdavidabccncifcrfgovhomejsp) was uti-lized to further understand their molecular biological func-tions to AD Table 1 shows the KEGG pathway analysisresult

First of all among the prediction results there are 5 genesthat have been confirmed as the AD virulence genes such asSNCA CALM1 GSK3B PSEN1 and APP which have beenbiologically demonstrated playing crucial roles in AD Based


(2) WWP1

(2) KLF5

(1) GSK3B

(2) DAZAP2

(2) SEPT2

(2) SNCA

(2) EEF1A1

(2) SRSF5

(1) CALM1

(1) TGM2

(2) NOC4L

(1) FASN

(2) NDRG1

(1) CREB1

(2) RBBP4

(1) E2F1

(2) CDK2(1) OGT

(3) APP

(0) NFKB1

(2) NOTCH1

(1) PSEN1

(2) CTNNB1

(2) COPS5

(1) MAPK14

(2) CENPC1

(2) UBC

(1) PRKAB1

(2) SNRNP200

(2) HNRNPU

(2) NR3C1

(1) RELA

(2) NCOR2

(2) SNW1

(2) KAT2B

(2) CDC25B

(2) DUSP1

(1) MAPK3

(2) UBR5

(2) KATNA1

(2) DYRK2

(2) CEBPA

(2) ESR1

(1) SIRT1

(2) HDAC2

Figure 1 Protein signaling pathway predicting result between NF-120581B and APP

on Table 1 T cell receptor signaling pathway B cell recep-tor signaling pathway the Notch signaling pathway NOD-like receptor signaling pathway Toll-like receptor signalingpathway MAPK signaling pathway neurotrophin signalingpathway insulin signaling pathway and so on were foundto include a major part of important genes derived fromthe regression model Specially the main predicted pathwayin Figure 1 includes NFKB1 NOTCH1 PSEN1 CTNNBCOPS5 MAPK14 CENPC1 UBC and APP the molecularbiological analysis shows that they have close correlationbetween inflammatory response and AD

It was found that inflammation is a major mechanismof acute brain injury and chronic neurodegeneration [17]During the onset of an inflammatory response signalingpathways are activated for translating extracellular signalsinto intracellular responses converging to the activation ofNF-120581B the central transcription factor in driving the inflam-matory response [18] NF-120581B has long been considered a pro-totypical proinflammatory signaling pathway largely basedon the activation of NF-120581B by proinflammatory cytokinessuch as interleukin-1 (IL-1) and tumor necrosis factor 120572

(TNF120572) and the role of NF-120581B on the expression of otherproinflammatory genes including cytokines chemokinesand adhesionmolecules which has been extensively reviewedelsewhere [9]

Recent studies have also found that Notch receptorsin Notch signaling pathway regulate cell differentiationand function and Notch1 has been shown to induce glia

in the peripheral nervous system [19 20] NF-120581B NotchMAKP and PSEN1 included in the main pathway ofFigure 1 were observed to have strong regulating functionsbetween each other since interleukin-1 (IL-1) activates NF-120581B via interleukin-1 receptor-associated kinase (IRAK) andmitogen-activated protein kinase (MEKK1(MAP3K)) depen-dent inhibition of NF-120581B inhibitor (I-120581B) [21 22] V-relreticuloendotheliosis viral oncogene homolog (c-Rel (NF-120581B subunit)) can trigger Notch homolog 1 translocation-associated (NOTCH1 receptor) signaling pathway by induc-ing expression of Jagged1 ligand for Notch receptors [23 24]NOTCH1 receptor activated by Jagged1 or Delta-like 1 (DLL1)is cleaved by ADAMmetallopeptidase domain 17 (ADAM17)and PSEN1 to intracellular domain of NOTCH1 NOTCH1is transported to nucleus and participates in recombinationsignal binding protein for immunoglobulin kappa J region(RBP-J kappa (CBF1)) mediated transcription [24 25]

It was also found that 120573-catenin- (CTNNB1-) dependentWNT signaling pathways have crucial roles in the regulationof diverse cell behaviours including cell fate proliferationsurvival differentiation migration and polarity [26 27] It isinteresting to note that loss of TNF120572 function would inhibitWnt120573-catenin signaling [28] Recently studies show thatWnt120573-catenin and NF-120581B are independent pathways cross-regulation between the Wnt and NF-120581B signaling pathwayshas emerged as an important area for the regulation ofa diverse array of genes and pathways active in chronicinflammation immunity development and tumorigenesis


Table 1 KEGG pathway analysis of the predicted pathways andsubnetworks in Figure 1

Pathway Numberof genes

Prostate cancer 9Notch signaling pathway 8Neurotrophin signaling pathway 6Pathways in cancer 6Chronic myeloid leukemia 6Alzheimerrsquos disease 5Melanogenesis 9T cell receptor signaling pathway 5Acute myeloid leukemia 5NOD-like receptor signaling pathway 5Cell cycle 5Insulin signaling pathway 6Pancreatic cancer 4B cell receptor signaling pathway 4Small cell lung cancer 4Progesterone-mediated oocyte maturation 4MAPK signaling pathway 4Toll-like receptor signaling pathway 5Endometrial cancer 6Spliceosome 4Glioma 5Adipocytokine signaling pathway 3Epithelial cell signaling inHelicobacter pylori infection 3RIG-I-like receptor signaling pathway 3Colorectal cancer 3

Both 120573-catenin and NF-120581B activate inducible nitric oxidesynthase (iNOS) gene expression [29]

In addition the regulatory network between COPS5 andCENPC1 has been extracted from our algorithm which isalso observed to be implicated in the pathogenesis of ADTheCOP9 (constitutive photomorphogenesis 9) signalosome(COPS) a large multiprotein complex that resembles the 19Slid of the 26S proteasome plays a central role in the regulationof the E3-cullin RING ubiquitin ligases (CRLs) The catalyticactivity of the COPS complex carried by subunit 5 (COPS5Jab1) COPS-dependent COPS 5 displays isopeptidaseactivity it is intrinsically inactive in other physiologicallyrelevant forms [30] Increased APP and accumulation ofneurotoxic A120573 in the brain are central to the pathogenesisof AD COPS5 is found to be a novel RanBP9-bindingprotein that increases APP processing and A120573 generation[31] COPS5 regulates the stability of the inner kinetochorecomponents CENP-T and CENP-W providing the first directlink betweenCOPS5 and themitotic apparatus and highlight-ing the role of COPS5 as amultifunctional cell cycle regulator[32] CENP-T interacts with both centromeric chromatinand microtubule binding kinetochore complexes Transienttargeting of CENP-C to a noncentromere LacO locus induces

the recruitment of some outer kinetochore proteins similarto CENP-T [33] Our result exhibited that COPS5 regulatesCENP-C in the main pathway and ubiquitin and NF-120581B werefound to be associated with them Ubiquitin can degrade theI120581Bwhich is the inhibitor of NF-120581B processing of precursorsand activation of the I120581Bkinase (IKK) through a degradation-independent mechanism [34] On the other hand COPS5functions through CDK2 to control premature senescencein a novel way depending on cyclin E in the cytoplasm[35]

4 Conclusions

Although many efforts have been done several decades ofAD it is still difficult to uncover its phenotype-pathwayrelationship and pathogenesis Recent studies show that thepathology of AD has an inflammatory component that ischaracterized by upregulation of proinflammatory cytokinesparticularly in response to A120573 However the signaling path-ways and regulatory networks of the inflammation in ADpathogenesis are very difficult to reconstruct due to thecomplexity

To discover the inflammation signaling pathway andregulatory network of AD in our study protein interactivenetwork data PPI was introduced to overcome the infor-mation insufficiencies of DNA microarray gene expressiondata by integer linear programming (ILP)method Two stageshad been used in predicting inflammatory pathway for ADFirstly significant genes had been selected by linear regres-sion analysis with themodified network-constrained regular-ization analysis Then ILP model was applied to reconstructthe signaling pathway betweenNF-120581B andADvirulence geneAPP since NF-120581B has long been considered a prototypicalproinflammatory signaling pathway From the molecularbiology analysis we found that genes on the main pathway ofthe reconstruction results play crucial roles in inflammatoryresponse and APP which give more biological insight for ADpathogenesis such as NF-120581B NOTCH1 CTNNB1 COPS5and their signaling pathways Evenmore the pathogenic con-tribution of the inflammatory response in AD is supported byour finding of the regulating and functions of the genes andsubnetworks in the predicted signaling pathways In generalour studies on combining PPI and gene expression datadiscover the signaling pathways of inflammatory response onAD and help for deeply understanding the pathogenesis ofAD

Conflict of Interests

The authors declare no conflict of interests

Acknowledgment

This work was supported by the National Natural ScienceFoundation of China (nos 61271446 and 61003093)


References

[1] A Haruhiko S Barger S Barnum et al ldquoInflammation andAlzheimerrsquos diseaserdquoNeurobiology of Aging vol 21 pp 383ndash4212000

[2] M Meyer-Luehmann T L Spires-Jones C Prada et al ldquoRapidappearance and local toxicity of amyloid-120573 plaques in a mousemodel of Alzheimerrsquos diseaserdquo Nature vol 451 no 7179 pp720ndash724 2008

[3] Y-J Lee S B Han S-Y Nam K-W Oh and J T HongldquoInflammation and Alzheimerrsquos diseaserdquo Archives of PharmacalResearch vol 33 no 10 pp 1539ndash1556 2010

[4] D Galimberti and E Scarpini ldquoInflammation and oxidativedamage in Alzheimerrsquos disease friend or foerdquo Frontiers inBioscience vol 3 pp 252ndash266 2011

[5] H Johnston H Boutin and S M Allan ldquoAssessing thecontribution of inflammation inmodels of Alzheimerrsquos diseaserdquoBiochemical Society Transactions vol 39 no 4 pp 886ndash8902011

[6] C Holmes and J Butchart ldquoSystemic inflammation andAlzheimerrsquos diseaserdquo Biochemical Society Transactions vol 39no 4 pp 898ndash901 2011

[7] X Lan R Liu L Sun T Zhang and G Du ldquoMethyl salicylate2-O-120573-D-lactoside a novel salicylic acid analogue acts as ananti-inflammatory agent on microglia and astrocytesrdquo Journalof Neuroinflammation vol 8 article 98 2011

[8] N Bhardwaj and H Lu ldquoCorrelation between gene expressionprofiles and protein-protein interactions within and acrossgenomesrdquo Bioinformatics vol 21 no 11 pp 2730ndash2738 2005

[9] Y Huang X Sun and G Hu ldquoAn integrated genetics approachfor identifying protein signal pathways of Alzheimerrsquos diseaserdquoComputer Methods in Biomechanics and Biomedical Engineer-ing vol 14 no 4 pp 371ndash378 2011

[10] J Feng R Jiang and T Jiang ldquoA max-flow-based approach tothe identification of protein complexes using protein interactionand microarray datardquo IEEEACM Transactions on Computa-tional Biology and Bioinformatics vol 8 no 3 pp 621ndash634 2011

[11] Y Huang J Zhang and Y Huang ldquoComputational iden-tification of proteins sub-network in Parkinsonrsquos diseasestudyrdquo in Proceedings of the International Conference on Anti-Counterfeiting Security and Identification (ASID rsquo12) pp 1ndash42012

[12] M J Jahid and J Ruan ldquoIdentification of biomarkers in breastcancer metastasis by integrating protein-protein interactionnetwork and gene expression datardquo in Proceedings of theIEEE International Workshop on Genomic Signal Processing andStatistics (GENSIPS rsquo11) pp 60ndash63 2011

[13] X-M Zhao R-S Wang L Chen and K Aihara ldquoUncoveringsignal transduction networks from high-throughput data byinteger linear programmingrdquoNucleic Acids Research vol 36 no9 article e48 2008

[14] C Li and H Li ldquoNetwork-constrained regularization andvariable selection for analysis of genomic datardquo Bioinformaticsvol 24 no 9 pp 1175ndash1182 2008

[15] X Zhou L Yuan X Zhao et al ldquoGenistein antagonizes inflam-matory damage induced by 120573-amyloid peptide in microgliathrough TLR4 and NF-120581Brdquo Nutrition vol 30 no 1 pp 90ndash952014

[16] E M Blalock J W Geddes K C Chen N M Porter W RMarkesbery and PW Landfield ldquoIncipient Alzheimerrsquos diseasemicroarray correlation analyses reveal major transcriptional

and tumor suppressor responsesrdquo Proceedings of the NationalAcademy of Sciences of the United States of America vol 101 no7 pp 2173ndash2178 2004

[17] T Rolova L Puli J Magga et al ldquoComplex regulation of acuteand chronic neuroinflammatory responses in mouse modelsdeficient for nuclear factor kappa B p50 subunitrdquo Neurobiologyof Disease vol 15 no 64 pp 16ndash29 2013

[18] P T Foteinou S E Calvano S F Lowry and I P AndroulakisldquoIn silico simulation of corticosteroids effect on an NFkB-dependent physicochemical model of systemic inflammationrdquoPLoS ONE vol 4 no 3 Article ID e4706 2009

[19] S Zanotti and C Ernesto ldquoNotch1 and Notch2 expressionin osteoblast precursors regulates femoral microarchitecturerdquoBone vol 62 pp 22ndash28 2014

[20] K Tanigaki F Nogaki J Takahashi K Tashiro H Kurookaand T Honjo ldquoNotch1 and Notch3 instructively restrict bFGF-responsive multipotent neural progenitor cells to an astroglialfaterdquo Neuron vol 29 no 1 pp 45ndash55 2001

[21] P Renard andMRaes ldquoTheproinflammatory transcription fac-tor NF120581B a potential target for novel therapeutical strategiesrdquoCell Biology and Toxicology vol 15 no 6 pp 341ndash344 1999

[22] J Yao W K Tae J Qin et al ldquoInterleukin-1 (IL-1)-inducedTAK1-dependent Versus MEKK3-dependent NF120581B activationpathways bifurcate at IL-1 receptor-associated kinase modifi-cationrdquo The Journal of Biological Chemistry vol 282 no 9 pp6075ndash6089 2007

[23] J Bash W-X Zong S Banga et al ldquoRelNF-120581B can trigger theNotch signaling pathway by inducing the expression of Jagged1a ligand for Notch receptorsrdquoThe EMBO Journal vol 18 no 10pp 2803ndash2811 1999

[24] C Osipo T E Golde B A Osborne and L A Miele ldquoOff thebeaten pathway the complex cross talk between Notch andNF-120581Brdquo Laboratory Investigation vol 88 no 1 pp 11ndash17 2008

[25] V Bolos J Grego-Bessa and J L De La Pompa ldquoNotchsignaling in development and cancerrdquo Endocrine Reviews vol28 no 3 pp 339ndash363 2007

[26] J N Anastas and R T Moon ldquoWNT signalling pathways astherapeutic targets in cancerrdquo Nature Reviews Cancer vol 13no 1 pp 11ndash26 2013

[27] K N Nejak-Bowen and S P S Monga ldquoBeta-catenin signalingliver regeneration and hepatocellular cancer sorting the goodfrom the badrdquo Seminars in Cancer Biology vol 21 no 1 pp 44ndash58 2011

[28] M Gong C Liu L Zhang et al ldquoLoss of the TNF120572 functioninhibits Wnt120573-catenin signaling exacerbates obesity develop-ment in adolescent spontaneous obese micerdquo Molecular andCellular Biochemistry 2014

[29] QDu andDAGeller ldquoCross-regulation betweenWnt andNF-120581B signaling pathwaysrdquo Forum on Immunopathological DiseaseTherapeutics vol 1 no 3 pp 155ndash181 2010

[30] A Echalier Y Pan M Birol et al ldquoInsights into the regulationof the human COP9 signalosome catalytic subunit CSN5Jab1rdquo Proceeding of National Academy of Science of USA vol 110 no4 pp 1273ndash1278 2013

[31] H Wang D Dey I Carrera et al ldquoCOPS5 (Jab1) proteinincreases 120573 site processing of amyloid precursor protein andamyloid 120573 peptide generation by stabilizing RanBP9 proteinlevelsrdquo The Journal of Biological Chemistry vol 288 no 37 pp26668ndash26677 2013

[32] Y Chun M Lee and B Park ldquoCSN5JAB1 interacts with thecentromeric components CENP-T and CENP-W and regulates


their proteasome-mediated degradationrdquoThe Journal of Biolog-ical Chemistry vol 288 no 38 pp 27208ndash27219 2013

[33] K E Gascoigne K Takeuchi A Suzuki T Hori T Fukagawaand I M Cheeseman ldquoInduced ectopic kinetochore assemblybypasses the requirement for CENP-A nucleosomesrdquo Cell vol145 no 3 pp 410ndash422 2011

[34] Z J Chen ldquoUbiquitin signalling in the NF-120581B pathwayrdquoNatureCell Biology vol 7 no 8 pp 758ndash765 2005

[35] A Yoshida N Yoneda-Kato and J Y Kato ldquoCSN5 specificallyinteracts with CDK2 and controls senescence in a cytoplasmiccyclin E-mediatedmannerrdquo Science Reports vol 3 p 1054 2013

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


and sequence motifs [9] PPI data contain structure informa-tion among different genes while gene expression data do notIn our study PPI network data as a priori pathway informa-tion is introduced for predicting the inflammatory signalingpathway in AD Many literatures have given outstandingachievements by integrating gene expression data and PPIdata such as identification of protein complexes [10] smallsubnetworks [11] and biomarkers [12] Zhao et al presentedan integer linear programming (ILP) method to uncoverpathways among the given starting proteins ending proteinsand some transduction factor proteins [13] However howto select the transduction factor proteins is a great problemIn our study a modified network-constrained regularizationanalysismethod [14] is proposed for linear regression analysisto select appropriate number of significant genes Simulationresults show that this method can lead to an efficiently globalsmoothness of regression coefficients

Based on that ILP model is presented to reconstructthe inflammatory signaling pathway by integrating PPI datawith the AD gene expression data In the ILP model thestarting and ending proteins of the predicting pathway needto be arranged in advance Nuclear transcription factor NF-120581B (nuclear factor-kappa B) as one of the most importantinflammatory factors is selected as the starting gene of thesignaling pathway As we know that NF-120581B plays a key rolein regulating the immune response to infection thereforeincorrect regulation of NF-120581B has been linked to cancerinflammatory and autoimmune diseases septic shock viralinfection and improper immune development NF-120581B hasalso been implicated in processes of synaptic plasticity andmemory [15] On the other hand APP (amyloid precursorprotein) as the most important AD virulence gene andprecursor protein of A120573 is arranged as the ending protein ofthe predicting pathway

The experiment results show that 6 AD virulence genesare identified being included in the predicted inflammatorysignaling pathway and a large amount of inflammationrelated genes and pathways has been found by molecularbiological analysis and they show the great understanding ofthe pathogenesis of AD

2 Methods

21 Linear Regression Model Linear regression model iswidely used in estimation and variable selection In ourstudy the model is applied to selected subset of significantgenes which are important for AD and are going to be thetransduction factors of reconstructing pathway In the nextprediction step gene expression data and PPI data will beintegrated by ILP model After all of the above a pathwaycould be identified between NF-120581B and APPThe usual linearregression model can be expressed as

120583 =

119901

sum

119895=1

119909119895120573119895 = 11990911205731 + 11990921205732 + sdot sdot sdot + 119909119901120573119901 (1)

where 120583 = (1205831 1205832 120583119899)119879 is response vector 119899 is the

sample number 119909119895 = (1199091 1199092 119909119899)119879 119895 = 1 2 119901 is the

predictor represented by the 119895th genersquos expression data in allsamples and 120573119895 is the 119895th genersquos weight vector Assume thatthe predictors are standardized and the response is centeredwe get

119899

sum

119894=1

120583119894 = 0

119899

sum

119894=1

119909119894119895 = 0

119899

sum

119894=1

1199092

119894119895= 0

for 119895 = 1 2 119901

(2)

Gene expression data has the characteristic of less sampleand great noise As a simple model linear regression modelhas significant performance in handling less sample and greatnoise data The significant genes will get a larger coefficientwhile the nonsignificant genes will get a smaller coefficient

22 Network-Constrained Regularization for the LinearRegression Model Before using linear regression modelcoefficient 120573 needs to be estimated Manymethods have beenproposed which focused on addressing high-dimensionalitygenomic data such as LASSO LA-SEN and LARS Here amodified network-constrain regularization analysis by CLi and H Li [14] is applied to estimate the coefficient sinceit has been proved to perform better than other methodsThis method is a lasso-type problem It defines a normalizedLaplacian matrix 119871 as

119871 =

1 minus 119908 (119906 V)119889119906

if 119906 = V and 119889119906 = 0

minus119908 (119906 V)radic119889119906119889V

if 119906 and V are adjacent

0 otherwise

(3)

where 119908(119906 V) represents the weight of edge between linkedgenes 119906 and V 119889V = sum119906simV 119908(119906 V) represents all the adjacentgenes of V on the networkThen the definition of the network-constrained regularization criterion is

119871 (1205821 1205822 120573) = (120583 minus 119883120573)119879(120583 minus 119883120573) + 1205821

100381610038161003816100381612057310038161003816100381610038161+ 1205822120573

119879119871120573

(4)

where119883 = (1199091|119870|119909119901) |1205731| = sum119901

119895=1|120573119895|12058211205822 are nonnegative

turning parameters And then we estimate 120573 by minimizing(4)

120573 = argmin120573

119871 (1205821 1205822 120573) (5)

Minimizing (4) is equivalent to solving a lasso-typeoptimization problem Turning parameters are estimatedby 10-fold cross-validation (CV) Genes in gene interactionnetwork are selected by PubGene we chose genes related toAlzheimer

23 Integer Linear Programming (ILP) The ILP model for-mulates signaling network detection as an optimizationproblem and treats a signaling network as a whole entityas described in its original publication [13] PPI networkis a weighted undirected graph that can be described as



Minmize119909119894 119910119894119895

119878 = minus

|119881|

sum

119894=1

|119881|

sum

119895=1

119908119894119895119910119894119895 + 120582

|119881|

sum

119894=1

|119881|

sum

119895=1

119910119894119895

Subject to 119910119894119895 le 119909119894

119910119894119895 le 119909119895

|119881|

sum

119895=1


or ending protein

|119881|

sum

119895=1


or ending protein


119909119894 isin 0 1 119894 = 1 2 |119881|

119910119894119895 isin 0 1 119894 119895 = 1 2 |119881|

(6)




119910119894119895 ge 2119909119894
















(2) WWP1

(2) KLF5

(1) GSK3B

(2) DAZAP2

(2) SEPT2

(2) SNCA

(2) EEF1A1

(2) SRSF5

(1) CALM1

(1) TGM2

(2) NOC4L

(1) FASN

(2) NDRG1

(1) CREB1

(2) RBBP4

(1) E2F1

(2) CDK2(1) OGT

(3) APP

(0) NFKB1

(2) NOTCH1

(1) PSEN1

(2) CTNNB1

(2) COPS5

(1) MAPK14

(2) CENPC1

(2) UBC

(1) PRKAB1

(2) SNRNP200

(2) HNRNPU

(2) NR3C1

(1) RELA

(2) NCOR2

(2) SNW1

(2) KAT2B

(2) CDC25B

(2) DUSP1

(1) MAPK3

(2) UBR5

(2) KATNA1

(2) DYRK2

(2) CEBPA

(2) ESR1

(1) SIRT1

(2) HDAC2















4 Conclusions





Acknowledgment



References












































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of


















Minmize119909119894 119910119894119895

119878 = minus

|119881|

sum

119894=1

|119881|

sum

119895=1

119908119894119895119910119894119895 + 120582

|119881|

sum

119894=1

|119881|

sum

119895=1

119910119894119895

Subject to 119910119894119895 le 119909119894

119910119894119895 le 119909119895

|119881|

sum

119895=1


or ending protein

|119881|

sum

119895=1


or ending protein


119909119894 isin 0 1 119894 = 1 2 |119881|

119910119894119895 isin 0 1 119894 119895 = 1 2 |119881|

(6)




119910119894119895 ge 2119909119894
















(2) WWP1

(2) KLF5

(1) GSK3B

(2) DAZAP2

(2) SEPT2

(2) SNCA

(2) EEF1A1

(2) SRSF5

(1) CALM1

(1) TGM2

(2) NOC4L

(1) FASN

(2) NDRG1

(1) CREB1

(2) RBBP4

(1) E2F1

(2) CDK2(1) OGT

(3) APP

(0) NFKB1

(2) NOTCH1

(1) PSEN1

(2) CTNNB1

(2) COPS5

(1) MAPK14

(2) CENPC1

(2) UBC

(1) PRKAB1

(2) SNRNP200

(2) HNRNPU

(2) NR3C1

(1) RELA

(2) NCOR2

(2) SNW1

(2) KAT2B

(2) CDC25B

(2) DUSP1

(1) MAPK3

(2) UBR5

(2) KATNA1

(2) DYRK2

(2) CEBPA

(2) ESR1

(1) SIRT1

(2) HDAC2















4 Conclusions





Acknowledgment



References












































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















(2) WWP1

(2) KLF5

(1) GSK3B

(2) DAZAP2

(2) SEPT2

(2) SNCA

(2) EEF1A1

(2) SRSF5

(1) CALM1

(1) TGM2

(2) NOC4L

(1) FASN

(2) NDRG1

(1) CREB1

(2) RBBP4

(1) E2F1

(2) CDK2(1) OGT

(3) APP

(0) NFKB1

(2) NOTCH1

(1) PSEN1

(2) CTNNB1

(2) COPS5

(1) MAPK14

(2) CENPC1

(2) UBC

(1) PRKAB1

(2) SNRNP200

(2) HNRNPU

(2) NR3C1

(1) RELA

(2) NCOR2

(2) SNW1

(2) KAT2B

(2) CDC25B

(2) DUSP1

(1) MAPK3

(2) UBR5

(2) KATNA1

(2) DYRK2

(2) CEBPA

(2) ESR1

(1) SIRT1

(2) HDAC2















4 Conclusions





Acknowledgment



References












































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of























4 Conclusions





Acknowledgment



References












































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of

















References












































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of


























of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















research article integrating gene expression and protein...

Documents