miRNA-Disease Association Prediction with Collaborative...

Research ArticlemiRNA-Disease Association Prediction with CollaborativeMatrix Factorization

Zhen Shen1 You-Hua Zhang2 Kyungsook Han3 Asoke K Nandi4

Barry Honig5 and De-Shuang Huang1

1 Institute of Machine Learning and Systems Biology School of Electronics and Information EngineeringTongji University Shanghai 201804 China2School of Information and Computer Anhui Agricultural University Changjiang West Road 130 Hefei Anhui China3Department of Computer Science and Engineering Inha University Incheon Republic of Korea4Department of Electronic and Computer Engineering Brunel University London Uxbridge UB8 3PH UK5Center for Computational Biology and Bioinformatics Columbia University 1130 St Nicholas AvenueRoom 815 New York NY 10032 USA

Correspondence should be addressed to De-Shuang Huang dshuangtongjieducn

Received 31 March 2017 Accepted 2 May 2017 Published 28 September 2017

Academic Editor Fang-Xiang Wu

Copyright copy 2017 Zhen Shen 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

As one of the factors in the noncoding RNA family microRNAs (miRNAs) are involved in the development and progression ofvarious complex diseases Experimental identification of miRNA-disease association is expensive and time-consuming Thereforeit is necessary to design efficient algorithms to identify novel miRNA-disease association In this paper we developed thecomputational method of Collaborative Matrix Factorization for miRNA-Disease Association prediction (CMFMDA) to identifypotential miRNA-disease associations by integrating miRNA functional similarity disease semantic similarity and experimentallyverified miRNA-disease associations Experiments verified that CMFMDA achieves intended purpose and application values withits short consuming-time and high prediction accuracy In addition we used CMFMDA on Esophageal Neoplasms and KidneyNeoplasms to reveal their potential related miRNAs As a result 84 and 82 of top 50 predicted miRNA-disease pairs for thesetwo diseases were confirmed by experiment Not only this but also CMFMDA could be applied to new diseases and new miRNAswithout any known associations which overcome the defects of many previous computational methods

1 Introduction

MicroRNAs (miRNAs) are a class of short noncoding RNAs(19sim25 nt) which normally regulate gene expression andprotein production by targeting messenger RNAs (mRNAs)at the posttranscriptional level [1ndash9] Since the first twomiRNA lin-4 and let-7 were found in 1993 and 2000 [1011] thousands of miRNAs have been detected in eukaryoticorganisms ranging from nematodes to humans The latestvenison of miRBase contains 26845 entries and more than2000 miRNAs have been detected in human [12ndash14] Withthe development of bioinformatics and the progress ofmiRNA-related projects researches are gradually focused onthe function of miRNAs Existing studies have shown thatmiRNAs are involved inmany important biological processes

[15 16] like cell differentiation [17] proliferation [18] signaltransduction [19] viral infection [20] and so on Thereforeit is easy to find that miRNAs have close relationship withvarious human complex diseases [12 21ndash26] For exampleresearchers found that mir-433 is upregulated in gastriccarcinoma by regulating the expression of GRB2 which isa known tumour-associated protein [27] Mir-126 can notonly function as an inhibitor to suppress the growth ofcolorectal cancer cells by its overexpression but also can helpto differentiate between malignant and normal colorectaltissue [28] Besides the change of mir-17sim92 miRNA clusterexpression has close relationship with kidney cyst growthin polycystic kidney disease [29] Considering the closerelationship between miRNA and disease we should try allmeans to excavate all latent associations between miRNA

HindawiComplexityVolume 2017 Article ID 2498957 9 pageshttpsdoiorg10115520172498957

2 Complexity

and disease and to facilitate the diagnose prevention andtreatment human complex disease [30ndash33] However usingexperimental methods to identify miRNA-disease associa-tion is expensive and time-consuming As themiRNA-relatedtheories are becoming more and more common such as theprediction model about miRNA and disease the function ofmiRNA in biological processes and signaling pathways newtherapies are urgently needed for the treatment of complexdisease it is necessary to develop powerful computationalmethods to reveal potential miRNA-disease associations [1215 20 34ndash40]

Previous studies had shown that functionally similarmiRNAs always appear in similar diseases therefore manycomputational models were proposed to identify novelmiRNA-disease associations [13 41ndash46] For example Jianget al [31] analyzed and improved disease-gene predictionmodel introduced the principle of hypergeometric distri-bution and how to use it and discussed its application inprediction model and its actual effect In order to real-ize the prediction function of the improved model theyused different types of dataset including miRNA functionalsimilarity data disease phenotype similarity data and theknown human disease-miRNA association data Thereforethe prediction accuracy of this method is greatly impactedby miRNA neighbor information and miRNA-target inter-action prediction Chen et al [47] reported a new methodHGIMDA to identify novel miRNA-disease association byusing heterogeneous graph inference This algorithm can getbetter prediction accuracy by integrating known miRNA-disease associations miRNA functional similarity diseasesemantic similarity and Gaussian interaction profile kernelsimilarity for diseases and miRNAs In addition HGIMDAcould be applied for new diseases and new miRNAs whichdo not have any known association Li et al [48] proposed thecomputational model Matrix completion for MiRNA-diseaseassociation prediction (MCMDA) to predict miRNA-diseaseassociations This model only uses known miRNA-diseaseassociations and achieved better prediction performanceThelimitation of MCMDA is that it could not be applied for newdiseases and new miRNAs which do not have any knownassociation You et al [49] developed model Path-BasedMiRNA-Disease Association Prediction (PBMDA) to predictmiRNA-disease associations by integrating known humanmiRNA-disease associations miRNA functional similaritydisease semantic similarity and Gaussian interaction profilekernel similarity formiRNAs and diseases Depth-first searchalgorithm was used in this model to identify novel miRNA-disease associations Benefiting from effective algorithm andreliable biological datasets PBMDA has better predictionperformance Furthermore Xu et al [50] introduced anapproach to identify disease-related miRNAs by the miRNAtarget-dysregulated network (MTDN) Furthermore in orderto distinguish and identify disease-related miRNAs fromcandidate a SVM classifier based on radial basis functionand the lib SVM package had been proposed Researcheshave shown that miRNAs can functionally interact withenvironmental factors (EFs) to affect and determine humancomplex disease Chen [51] proposed model miREFRWRto predict the association between disease and miRNA-EF

interactions Random walks theory was applied on miRNAsimilarity network and EF similarity network In additiondrug chemical structure similarity miRNA function sim-ilarity and networked-based similarity were also used inmiREFRWR Based on these biological datasets and efficientcalculation method miREFRWR could be an effective toolin computational biology What is more Chen et al [52]also proposed a computational model RKNNMDA to predictthe potential associations between miRNA and disease Fourbiological datasets experimentally verified human miRNA-disease associations miRNA functional similarity diseasesemantic similarity and Gaussian interaction profile kernelsimilarity for miRNAs and diseases were integrated intoRKNNMDA It can be found that the prediction accuracyof RKNNMDA is excellent Moreover RKNNMDA could beapplied for newdiseaseswhich do not have any known relatedmiRNA information

Generally speaking current prediction model onmiRNA-disease association is still demonstrating some short-comings For example unreliable datasets have a great influ-ence on the accuracy of prediction model such as miRNA-target interactions and disease-genes associations Inaddition for miRNAs and diseases which do not haveany known associations we cannot use some of the existingmodels to predict its relevant information In other words weneed to design and develop a new effective computationalmodel According to the assumption that functionally similarmiRNAs always appear in similar diseases we introduce themodel of Collaborative Matrix Factorization for MiRNA-Disease Association prediction (CMFMDA) to reveal novelmiRNA-disease association by integrating experimentallyvalidated miRNA-disease associations miRNA functionalsimilarity information and disease semantic similarityinformation For CMFMDA we can obtain its test resultswith three different ways 5-fold CV Local LOOCV andglobal LOOCV The AUCs of these three methods are08697 08318 and 08841 respectively which suggest thatCMFMDA is a reliable and efficient prediction model Andthen we use two cases Esophageal Neoplasms and KidneyNeoplasms to evaluate the performance of CMFMDA Inboth of these two important diseases 42 and 41 out of top 50predicted miRNA-disease associations were confirmed byrecent experimental literatures respectively In additionexperiments show that CMFMDA can be applied for diseasesand miRNAs without any known association

2 Materials and Methods

21 HumanmiRNA-Disease Associations We obtained infor-mation about the associations between miRNA and dis-ease from HMDD including 5430 experimentally confirmedhuman miRNA-diseases associations about 383 diseases and495 miRNAs Adjacency matrix 119860 is proposed to describethe association between miRNA and disease If miRNA 119898(119894)is associated with disease 119889(119895) the entity 119860(119898(119894) 119889(119895)) is1 otherwise 0 Furthermore we declared two variables nmand nd to represent the number of miRNAs and diseasesinvestigated in this paper respectively

Complexity 3

22 MiRNA Functional Similarity Base on the assumptionthat miRNAs with similarity functions are regarded to beinvolved in similar diseasesWang et al [42] present amethodto calculate the miRNA functional similarity score WedownloadedmiRNA functional similarity scores from httpwwwcuilabcnfilesimagescuilabmisimzip and constructedmatrix 119878119872 to represent the miRNA function similaritynetwork where the entity 119878119872(119898(119894) 119898(119895)) represents thefunctional similarity score between miRNA119898(119894) and119898(119895)

23 Disease Semantic Similarity In this paper disease can bedescribed as aDirectedAcyclic Graph (DAG) andDAG(119863) =(119863 119879(119863) 119864(119863)) was used to describe disease119863 where 119879(119863)is the node set including all ancestor nodes of 119863 and 119863 itselfand 119864(119863) is the corresponding links set including the directedges from parent nodes to child nodes The semantic valueof disease119863 in DAG(119863) is defined as follows

1198631198811 (119863) = sum119889isin119879(119863)

1198631119863 (119889)

1198631119863 (119889)

=

1 if 119889 = 119863max Δ lowast 1198631119863 (1198891015840) | 1198891015840 isin children of 119889 if 119889 = 119863

(1)

where Δ is the semantic contribution factor For disease 119863the contribution of itself to the semantic value of disease 119863is 1 However with the growth of the distance between 119863and other disease the contributionwill fallTherefore diseaseterms in the same layer would have the same contribution tothe semantic value of disease119863

If there is much in common between two diseases inDAG their semantic similarity will become largerThereforethe semantic similarity between diseases 119889(119894) and 119889(119895) can bedefined as follows

119878119863 (119889 (119894) 119889 (119895)) =sum119905isin119879(119894)cap119879(119895) (1198631119894 (119905) + 1198631119895 (119905))

1198631198811 (119894) + 1198631198811 (119895) (2)

where 119878119863 is the disease semantic similarity matrix

24 CMFMDA In this study we developed the compu-tational model of Collaborative Matrix Factorization forMiRNA-Disease Association prediction (CMFMDA) to pre-dict novel miRNA-disease associations [53] The flow ofCMFMDA is shown in Figure 1

In the first step in Figure 1 we will get the final miRNAsimilarity matrix 119878119872 and diseases similarity matrix 119878119863 byintegrating miRNA functional similarity network diseasesemantic similarity network and experimentally verifiedmiRNA-disease associations

Then we use WKNKN [54] to estimate the associationprobability for these unknown cases based on their knownneighbors

Thirdly Collaborative Matrix Factorization was used toobtain the final prediction 119865 This step contains three parts

(1) For the input matrix119884 this step adopts singular valuedecomposition to get the initial value of 119860 and 119861

[119880 119878 119881] = 119878119881119863 (119884 119896)

119860 = 11988011987812119896

119861 = 11988111987812119896

(3)

(2) We use 119871 to represent the objection function and use119886119894 and 119887119895 to represent the 119894th and 119895th row vectors of 119860and 119861 Two alternative update rules (one for updatingmatrix119860 and one for updatingmatrix119861) were derivedby setting 120597119871120597119860 = 0 and 120597119871120597119861 = 0 According toalternating least squares these two update rules arerun alternatingly until convergence

min119860119861

10038171003817100381710038171003817119884 minus 119860119861119879100381710038171003817100381710038172

119865+ 120582119897 (1198602119865 + 1198612119865)

+ 12058211989810038171003817100381710038171003817119878119872 minus 11986011986011987910038171003817100381710038171003817

2

119865+ 120582119889

10038171003817100381710038171003817119878119863 minus 119861119861119879100381710038171003817100381710038172

119865

119860 = (119884119861 + 120582119889119878119889119860) (119861119879119861 + 120582119897119868119896 + 120582119889119860119879119860)minus1

119861

= (119884119879119860 + 120582119898119878119898119861) (119860119879119860 + 120582119897119868119896 + 120582119898119861119879119861)minus1

(4)

Finally the predicted matrix for miRNA-disease associa-tions is then obtained by multiplying 119860 and 119861

3 Results

31 Performance Evaluation Based on the known miRNA-disease associations obtained fromHMDDdatabase [55] thepredictive performance of CMFMDA is evaluated throughtwo ways Local and global LOOCV Not only that threecomputational models WBSMDA [4] RLSMDA [12] andNCPMDA [56] were introduced to compare the predictionperformance with CMFMDA To obtain relevant miRNAinformation for the chosen disease 119889 all association relatedto disease 119889 was left out and the rest of the associationsserve as a training set to get prediction association byCMFMDA For cross-validation the difference between localLOOCV and global LOOCV is that all diseases would beinvestigated simultaneously or not Furthermore Receiver-Operating Characteristics (ROC) were used to express thedifference between true positive rate (TPR sensitivity) andfalse positive rate (FPR 1 minus specificity) at different thresholdsIn this case sensitivity indicates that the percentage of thetest miRNA-disease association which obtained ranks higherthan the given threshold Meanwhile specificity indicates thepercentage of miRNA-disease associations below the thresh-old What is more Area under the ROC curve (AUC) couldbe calculated to demonstrate the prediction performance ofCMFMDA AUC = 1 showed that the model has perfectprediction ability AUC = 05 indicates random predictionability

4 Complexity

MiRNA similaritymatrix SM

Disease similarity matrix SD

MiRNA-diseaseassociations

Horizontal direction update get weighted average of the corresponding values in theprofiles of the K knownmiRNAs nearest to

Vertical direction update get weighted average of the corresponding values in the profiles of the K knowndiseases nearest to

decomposition

Convergence

NoYes

(c)

(d)

(b)

(a)

mj Ym di Yd

A = US12k

B = VS12k

Y = AB

Singular value

A = (YB + dSdA) (BTB + lIk + dATA)

minus1

B = (YTA + mSmB) (ATA + lIk + mBTB)minus1

Y = ＧＲ (Y Ydm)

Ydm =(Yd + Ym)

2

Figure 1 Flowchart of potential miRNA-disease associations prediction based on CMFMDA

To illustrate the performance of CMFMDA we com-pare it with the existed computational model NCPMDARLSMDA and WBSMDA The comparison result has beenshown in Figure 2 As a result these four models obtainedAUCs of 08841 08630 08501 and 07799 in the globalLOOCV respectively For local LOOCV these four modelsobtained AUCs of 08318 08198 08068 and 07213 respec-tively In general CMFMDA has not only high predictionperformance but also better ability to identify novel miRNA-disease association

32 Case Studies All diseases in this paper have been inves-tigated by CMFMDA to predict some novel miRNAs whichhave association with the disease Here two case stud-ies Esophageal Neoplasms and Kidney Neoplasms wereproposed to demonstrate the prediction performance ofCMFMDA In addition we use two important miRNA-disease association databases to validate the predictionresults miR2Disease [57] and dbDEMC [58] A final noteabout validation datasets is that only the associations whichwere absent from the HMDD database would be used In

other words validation datasets have no correlation with thedatasets which have been used for prediction

Esophageal Neoplasms is a serious disease in diges-tive system which leads to high death rate [59ndash61] Earlydiagnosis and treatment is essential for improving patientrsquossurvival [62 63] Here we use CMFMDA to identify poten-tial miRNAs associated with Esophageal Neoplasms As aresult 9 out of the top 10 and 42 out of the top 50 pre-dicted related miRNAs were experimentally confirmed to beassociated with Esophageal Neoplasms (See Table 1) Forexample mir-133b can inhibit the cell growth and invasionof esophageal squamous cell carcinoma (ESCC) by targetingFascin homolog 1 [59] The expression level of mir-335 is anindependent prognostic factor in ESCC which might be apotential valuable biomarker for ESCC [64]

As a common urologic malignancy the morbidity andmortality of Kidney Neoplasm have been shown to rise grad-ually [65ndash68] Renal cell carcinoma (RCC) can be dividedinto several different types of cancer [69ndash71] including chro-mophobe RCC (CHRCC) collecting duct carcinoma (CDC)

Complexity 5

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1TP

RGlobal LOOCV

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1

TPR

Local LOOCV

WBSMDA (AUC = 07799)RLSMDA (AUC = 08501)NCPMDA (AUC = 08630)CMFMDA (AUC = 08841)


Figure 2 Performance comparisons between CMFMDA and three state-of-the-art disease-miRNA association prediction models(NCPMDA RLSMDA and WBSMDA) in terms of ROC curve and AUC based on local and global LOOCV respectively As a resultCMFMDA achieved AUCs of 08841 and 08318 in the global and local LOOCV significantly outperforming all the previous classical models

clear cell RCC (CCRCC) and papillary RCC (PRCC) Previ-ous studies have shown that miRNAs play a significant partin Kidney Neoplasm [72ndash74] In this paper CMFMDA wasemployed to identify potential miRNAs associated with Kid-neyNeoplasms As a result 9 out of the top-10 candidates and41 out of the top-50 candidates of Kidney Neoplasm relatedmiRNAs were confirmed by dbDEMC and miR2Ddisease(See Table 2) For example the serum level of mi-210 maybe used as a novel noninvasive biomarker for the detectionof CCRCC [75] Experiment results demonstrate that mir-9 expression is correlative not only with the development ofCCRCC but also with the development of metastatic recur-rence [76]

The results of cross-validation and independent casestudies show that CMFMDA can satisfy the needs to iden-tify potential miRNA-disease associations Furthermore alldiseases in HMDD have been investigated by CMFMDA topredict potential miRNAs (See Supplementary Table 1 inSupplementary Material available online at httpsdoiorg10115520172498957) We hope that potential disease-miRNA association predicted by CMFMDA could be con-firmed by further biological experiments

4 Discussion

According to the assumption that functionally similar miR-NAs are often associated with similar diseases we proposedthe computational model of Collaborative Matrix Factoriza-tion forMiRNA-Disease Association prediction (CMFMDA)to identify potential miRNA-disease associations by integrat-ing miRNA functional similarity disease semantic similarityand experimentally verified miRNA-disease associations Wecompare CMFMDA with the existing computational modelNCPMDA RLSMDA and WBSMDA and concluded thatCMFMDA has better prediction performance from thesefour modelsrsquo obtained AUCs in the global LOOCV orlocal LOOCV respectively There are some reasons for thereliable performance of CMFMDA Firstly several types ofexperimentally confirmed biological datasets are used inCMFMDA including known miRNA-disease associationsmiRNA functional similarity network and disease semanticsimilarity network which help improve the prediction per-formance and reduce variance Then CMFMDA can worknot only for known miRNA-disease association but also fordiseases andmiRNAswithout any known association Finally

6 Complexity

Table 1 We implemented CMFMDA to predict potential Esophag-ealNeoplasms-relatedmiRNAsAs a result 9 out of the top 10 and 42out of the top 50 predicted Esophageal Neoplasms related miRNAswere confirmed based on miR2Disease and dbDEMC (1st columntop 1ndash25 2nd column top 26ndash50)

miRNA Evidence miRNA Evidencehsa-mir-142 dbDEMC hsa-mir-30c dbDEMChsa-mir-1 dbDEMC hsa-mir-212 Unconfirmedhsa-mir-16 dbDEMC hsa-mir-424 dbDEMChsa-mir-127 dbDEMC hsa-mir-429 dbDEMChsa-mir-497 dbDEMC hsa-mir-498 dbDEMChsa-mir-200b dbDEMC hsa-mir-340 Unconfirmedhsa-mir-376c Unconfirmed hsa-mir-222 dbDEMChsa-mir-148b dbDEMC hsa-mir-146b dbDEMChsa-mir-335 dbDEMC hsa-mir-195 dbDEMChsa-mir-93 dbDEMC hsa-mir-10b dbDEMChsa-mir-125b dbDEMC hsa-mir-218 Unconfirmedhsa-mir-18a dbDEMC hsa-mir-181a dbDEMChsa-mir-17 dbDEMC hsa-mir-137 dbDEMChsa-mir-30a dbDEMC hsa-let-7e dbDEMChsa-mir-133b dbDEMC hsa-mir-181b dbDEMChsa-mir-135a dbDEMC hsa-mir-106a dbDEMC

hsa-mir-107 dbDEMCmiR2Disease hsa-mir-19b dbDEMC

hsa-mir-224 dbDEMC hsa-mir-95 dbDEMChsa-mir-199b dbDEMC hsa-mir-122 Unconfirmedhsa-mir-221 dbDEMC hsa-mir-152 dbDEMChsa-mir-18b dbDEMC hsa-mir-370 dbDEMChsa-mir-191 dbDEMC hsa-mir-30d dbDEMChsa-let-7g dbDEMC hsa-mir-15b dbDEMChsa-let-7f Unconfirmed hsa-mir-629 Unconfirmedhsa-mir-494 dbDEMC hsa-mir-204 Unconfirmed

as a global prediction model CMFMDA could be used topredict all disease-related miRNA at the same time

Although CMFMDA has better prediction performancethe limitation still exists in it and needs to be improved inthe future Firstly CMFMDA may cause bias to miRNAswith more known associated diseases Secondly the knownmiRNA-disease associations with experimental evidences arestill insufficient The prediction performance of CMFMDAwill be improved by integrating more reliable biologicalinformation [77ndash86] Finally how to more reasonably extractand integrate information from biological datasets should beinvestigated in the future

5 Conclusions

Research has shown that the abnormal expression of miRNAplays a crucial role in the occurrence and development ofhuman complex diseases The in-depth study and analysisof diseases-related miRNA could help find new biomarkerand therapies and then improve the survival rate of patients

Table 2 We implemented CMFMDA to prioritize candidate miR-NAs for Kidney Neoplasms based on known associations in theHMDD database As a result 9 out of the top 10 and 41 out of thetop 50 predictedKidneyNeoplasms relatedmiRNAswere confirmedbased on miR2Disease and dbDEMC (1st column top 1ndash25 2ndcolumn top 26ndash50)

miRNA Evidence miRNA Evidencehsa-mir-429 dbDEMC hsa-mir-34c dbDEMC

hsa-mir-200b dbDEMCmiR2Disease hsa-mir-10b dbDEMC

hsa-mir-200a dbDEMC hsa-mir-9 dbDEMC


hsa-mir-203 dbDEMC hsa-mir-99b dbDEMC

hsa-mir-218 dbDEMC hsa-mir-126 dbDEMCmiR2Disease

hsa-mir-127 dbDEMC hsa-mir-224 dbDEMChsa-mir-155 dbDEMC hsa-mir-195 dbDEMC

hsa-mir-205 dbDEMCmiR2Disease hsa-mir-145 dbDEMC

hsa-mir-196a dbDEMC hsa-mir-7 dbDEMCmiR2Disease

hsa-mir-143 dbDEMC hsa-mir-142 Unconfirmedhsa-mir-135a Unconfirmed hsa-mir-139 dbDEMChsa-mir-92a Unconfirmed hsa-mir-16 dbDEMChsa-mir-19a dbDEMC hsa-mir-183 dbDEMC

hsa-mir-29c dbDEMCmiR2Disease hsa-mir-296 Unconfirmed

hsa-mir-146a dbDEMC hsa-mir-367 Unconfirmedhsa-mir-99a dbDEMC hsa-mir-15b dbDEMC


hsa-mir-199a dbDEMCmiR2Disease hsa-mir-196b dbDEMC

hsa-mir-27a dbDEMCmiR2Disease hsa-mir-339 dbDEMC

hsa-mir-100 dbDEMC hsa-mir-26a dbDEMCmiR2Disease

hsa-mir-452 dbDEMC hsa-mir-302c Unconfirmedhsa-mir-34b dbDEMC hsa-mir-302b Unconfirmedhsa-mir-93 dbDEMC hsa-mir-107 dbDEMChsa-mir-34a dbDEMC hsa-mir-133b Unconfirmed

Therefore it is necessary to develop more effective compu-tational models to identify potential miRNA-disease associ-ations In this paper we presented a computational modelCMFMDA to identify novel miRNA-disease associationsExcept for disease semantic similarity andmiRNA functionalsimilarity CMFMDA also uses known miRNA-disease asso-ciations to predict miRNA-disease associations LOOCVwaschosen to evaluate the predict performance of CMFMDAThe results of LOOCV and case studies show that CMFMDAhas better prediction performance than other models Inother words as an effective tool CMFMDA can be used notonly to predict potential miRNA-disease associations but

Complexity 7

also to identify new biomarker that gave new direction fordiagnosis and treatment of human complex disease

Disclosure

De-Shuang Huang is the corresponding author of this paperProfessor A K Nandi is a Distinguished Visiting Professor atTongji University Shanghai China

Conflicts of Interest

The authors declare that the research was conducted in theabsence of any commercial or financial relationships thatcould be construed as potential conflicts of interest

Acknowledgments

Thisworkwas supported by the grants of theNational ScienceFoundation of China nos 61520106006 61732012 3157136461532008 U1611265 61672382 61772370 61402334 61472173and 61472282 and China Postdoctoral Science Foundation[Grant nos 2015M580352 2017M611619 and 2016M601646]

References

[1] V Ambros ldquomicroRNAs tiny regulators with great potentialrdquoCell vol 107 no 7 pp 823ndash826 2001

[2] V Ambros ldquoThe functions of animal microRNAsrdquo Nature vol431 no 7006 pp 350ndash355 2004

[3] D P Bartel ldquoMicroRNAs genomics biogenesis mechanismand functionrdquo Cell vol 116 no 2 pp 281ndash297 2004

[4] X Chen C C Yan X Zhang et al ldquoWBSMDA within andbetween score for MiRNA-disease association predictionrdquo Sci-entific Reports vol 6 Article ID 21106 2016

[5] R Ibrahim N A Yousri M A Ismail and N M El-MakkyldquoMiRNA and gene expression based cancer classificationusing self-learning and co-training approachesrdquo in Proceedingsof the IEEE International Conference on Bioinformatics andBiomedicine IEEE BIBM 2013 pp 495ndash498 Shanghai ChinaDecember 2013

[6] Y Katayama M Maeda K Miyaguchi et al ldquoIdentification ofpathogenesis-related microRNAs in hepatocellular carcinomaby expression profilingrdquo Oncology Letters vol 4 no 4 pp 817ndash823 2012

[7] G Meister and T Tuschl ldquoMechanisms of gene silencing bydouble-stranded RNArdquo Nature vol 431 no 7006 pp 343ndash3492004

[8] D S Huang and H-J Yu ldquoNormalized feature vectors a novelalignment-free sequence comparison method based on thenumbers of adjacent amino acidsrdquo IEEEACM Transactions onComputational Biology and Bioinformatics vol 10 no 2 pp457ndash467 2013

[9] S-P Deng and D S Huang ldquoSFAPS an R package forstructurefunction analysis of protein sequences based oninformational spectrum methodrdquo Methods vol 69 no 3 pp207ndash212 2014

[10] R C Lee R L Feinbaum and V Ambros ldquoThe C elegansheterochronic gene lin-4 encodes small RNAs with antisensecomplementarity to lin-14rdquoCell vol 75 no 5 pp 843ndash854 1993

[11] B J Reinhart F J Slack M Basson et al ldquoThe 21-nucleotidelet-7 RNA regulates developmental timing in Caenorhabditiselegansrdquo Nature vol 403 no 6772 pp 901ndash906 2000

[12] X Chen and G-Y Yan ldquoSemi-supervised learning for potentialhuman microRNA-disease associations inferencerdquo ScientificReports vol 4 article 5501 2014

[13] S Bandyopadhyay R Mitra U Maulik and M Q ZhangldquoDevelopment of the human cancer microRNA networkrdquoSilence vol 1 article 6 2010

[14] A Kozomara and S Griffiths-Jones ldquomiRBase annotating highconfidence microRNAs using deep sequencing datardquo NucleicAcids Research vol 42 pp D68ndashD73 2013

[15] X Chen C Clarence Yan X Zhang et al ldquoRBMMMDApredicting multiple types of disease-microRNA associationsrdquoScientific Reports vol 5 Article ID 13877 2015

[16] D P Bartel ldquoMicroRNAs target recognition and regulatoryfunctionsrdquo Cell vol 136 no 2 pp 215ndash233 2009

[17] E A Miska ldquoHow microRNAs control cell division differenti-ation and deathrdquoCurrent Opinion in Genetics and Developmentvol 15 no 5 pp 563ndash568 2005

[18] A M Cheng M W Byrom J Shelton and L P FordldquoAntisense inhibition of human miRNAs and indications for aninvolvement of miRNA in cell growth and apoptosisrdquo NucleicAcids Research vol 33 no 4 pp 1290ndash1297 2005

[19] Q Cui Z Yu E O Purisima and E Wang ldquoPrinciples ofmicroRNA regulation of a human cellular signaling networkrdquoMolecular Systems Biology vol 2 article 46 2006

[20] X ChenM X Liu andG Y Yan ldquoRWRMDA predicting novelhuman microRNA-disease associationsrdquoMolecular BioSystemsvol 8 no 10 pp 2792ndash2798 2012

[21] J Li Y Liu X Xin et al ldquoEvidence for positive selection ona number of microRNA regulatory interactions during recenthuman evolutionrdquo PLoS Genetics vol 8 no 3 Article IDe1002578 2012

[22] K Chen and N Rajewsky ldquoNatural selection on humanmicroRNA binding sites inferred from SNP datardquo NatureGenetics vol 38 no 12 pp 1452ndash1456 2006

[23] M A Saunders H Liang and W-H Li ldquoHuman polymor-phism atmicroRNAs andmicroRNA target sitesrdquoProceedings ofthe National Academy of Sciences of the United States of Americavol 104 no 9 pp 3300ndash3305 2007

[24] P Sethupathy and F S Collins ldquoMicroRNA target site polymor-phisms and human diseaserdquo Trends in Genetics vol 24 no 10pp 489ndash497 2008

[25] S-P Deng L Zhu and D S Huang ldquoPredicting hub genesassociated with cervical cancer through gene co-expressionnetworksrdquo IEEEACM Transactions on Computational Biologyand Bioinformatics vol 13 no 1 pp 27ndash35 2016

[26] S-P Deng L Zhu and D S Huang ldquoMining the bladdercancer-associated genes by an integrated strategy for the con-struction and analysis of differential co-expression networksrdquoBMC Genomics vol 16 no 3 supplement 3 article S4 2015

[27] H Luo H Zhang Z Zhang et al ldquoDown-regulated miR-9 andmiR-433 in human gastric carcinomardquo Journal of Experimentaland Clinical Cancer Research vol 28 article 82 2009

[28] X-M Li A-M Wang J Zhang and H Yi ldquoDown-regulationof miR-126 expression in colorectal cancer and its clinicalsignificancerdquo Medical Oncology vol 28 no 4 pp 1054ndash10572011

[29] V Patel D Williams S Hajarnis et al ldquoMiR-17sim92 miRNAcluster promotes kidney cyst growth in polycystic kidney

8 Complexity

diseaserdquo Proceedings of the National Academy of Sciences of theUnited States of America vol 110 no 26 pp 10765ndash10770 2013

[30] G A Calin and C M Croce ldquoMicroRNA signatures in humancancersrdquo Nature Reviews Cancer vol 6 no 11 pp 857ndash8662006

[31] Q Jiang Y Hao G Wang et al ldquoPrioritization of diseasemicroRNAs through a human phenome-microRNAome net-workrdquo BMC Systems Biology vol 4 supplement 1 article S22010

[32] Q Jiang G Wang S Jin Y Li and Y Wang ldquoPredictinghumanmicroRNA-disease associations based on support vectormachinerdquo International Journal of Data Mining and Bioinfor-matics vol 8 no 3 pp 282ndash293 2013

[33] C-H Zheng D S Huang L Zhang and X-Z Kong ldquoTumorclustering using nonnegative matrix factorization with geneselectionrdquo IEEE Transactions on Information Technology inBiomedicine vol 13 no 4 pp 599ndash607 2009

[34] X Chen ldquoPredicting lncRNA-disease associations and con-structing lncRNA functional similarity network based on theinformation of miRNArdquo Scientific Reports vol 5 Article ID13186 2015

[35] X Chen C C Yan C Luo W Ji Y Zhang and Q Dai ldquoCon-structing lncRNA functional similarity network based onlncRNA-disease associations and disease semantic similarityrdquoScientific Reports vol 5 Article ID 11338 2015

[36] X Chen and G-Y Yan ldquoNovel human lncRNA-disease associ-ation inference based on lncRNA expression profilesrdquo Bioinfor-matics vol 29 no 20 pp 2617ndash2624 2013

[37] X Chen M-X Liu Q-H Cui and G-Y Yan ldquoPrediction ofdisease-related interactions between microRNAs and environ-mental factors based on a semi-supervised classifierrdquo PLoSONE vol 7 no 8 Article ID e43425 2012

[38] X Chen ldquoKATZLDA KATZ measure for the lncRNA-diseaseassociation predictionrdquo Scientific Reports vol 5 Article ID16840 2015

[39] C-H Zheng L Zhang V T-Y Ng S C-K Shiu and D SHuang ldquoMolecular pattern discovery based on penalized ma-trix decompositionrdquo IEEEACMTransactions onComputationalBiology and Bioinformatics vol 8 no 6 pp 1592ndash1603 2011

[40] L Zhu S-P Deng and D S Huang ldquoA two-stage geometricmethod for pruning unreliable links in protein-protein net-worksrdquo IEEE Transactions on Nanobioscience vol 14 no 5 pp528ndash534 2015

[41] M Lu Q Zhang M Deng et al ldquoAn analysis of humanmicroRNA and disease associationsrdquo PLoS ONE vol 3 no 10Article ID e3420 2008

[42] D Wang J Wang M Lu F Song and Q Cui ldquoInferring thehuman microRNA functional similarity and functional net-work based on microRNA-associated diseasesrdquo Bioinformaticsvol 26 no 13 pp 1644ndash1650 2010

[43] K Goh M E Cusick D Valle B Childs M Vidal andA Barabasi ldquoThe human disease networkrdquo Proceedings of theNational Academy of Sciences of the United States of Americavol 104 no 21 pp 8685ndash8690 2007

[44] C Pasquier and J Gardes ldquoPrediction of miRNA-diseaseassociationswith a vector spacemodelrdquo Scientific Reports vol 6Article ID 27036 2016

[45] T D Le J Zhang L Liu and J Li ldquoComputational methods foridentifying miRNA sponge interactionsrdquo Briefings in Bioinfor-matics p bbw042 2016

[46] D S Huang and C H Zheng ldquoIndependent component anal-ysis-based penalized discriminant method for tumor classifica-tion using gene expression datardquo Bioinformatics vol 22 no 15pp 1855ndash1862 2006

[47] X Chen C C Yan X Zhang Z-H You Y-A Huang andG-YYan ldquoHGIMDA Heterogeneous graph inference for miRNA-disease association predictionrdquo Oncotarget vol 7 no 40 pp65257ndash65269 2016

[48] J-Q Li Z-H Rong X Chen G-Y Yan and Z-H YouldquoMCMDA matrix completion for MiRNA-disease associationpredictionrdquo Oncotarget vol 8 pp 21187ndash21199 2017

[49] Z-H You Z-A Huang Z Zhu et al ldquoPBMDA a novel andeffective path-based computational model for miRNA-diseaseassociation predictionrdquo PLOS Computational Biology vol 13no 3 Article ID e1005455 2017

[50] J Xu C-X Li J-Y Lv et al ldquoPrioritizing candidate dis-ease miRNAs by topological features in the miRNA target-dysregulated network case study of prostate cancerrdquoMolecularCancer Therapeutics vol 10 no 10 pp 1857ndash1866 2011

[51] X Chen ldquomiREFRWR a novel disease-related microRNA-environmental factor interactions prediction methodrdquo Molec-ular BioSystems vol 12 no 2 pp 624ndash633 2016

[52] X Chen Q-FWu andG-Y Yan ldquoRKNNMDA ranking-basedKNN for MiRNA-disease association predictionrdquo RNA Biologypp 1ndash11 2017

[53] X Zheng H Ding H Mamitsuka and S Zhu ldquoCollaborativematrix factorization with multiple similarities for predictingdrug-target interactionsrdquo in Proceedings of the 19th ACMSIGKDD International Conference on Knowledge Discovery andData Mining pp 1025ndash1033 Chicago Ill USA August 2013

[54] A Ezzat P Zhao M Wu X Li and C Kwoh ldquoDrug-targetinteraction prediction with graph regularized matrix factoriza-tionrdquo IEEEACM Transactions on Computational Biology andBioinformatics vol 14 no 3 pp 646ndash656 2016

[55] Y Li CQiu J Tu et al ldquoHMDDv20 a database for experimen-tally supported human microRNA and disease associationsrdquoNucleic Acids Research vol 42 pp D1070ndashD1074 2014

[56] C Gu B Liao X Li and K Li ldquoNetwork consistency projectionfor human miRNA-disease associations inferencerdquo ScientificReports vol 6 Article ID 36054 2016

[57] Q Jiang Y Wang Y Hao et al ldquomiR2Disease a manuallycurated database for microRNA deregulation in human dis-easerdquo Nucleic Acids Research vol 37 supplement 1 pp D98ndashD104 2009

[58] Z Yang F Ren C Liu et al ldquodbDEMC a database of differ-entially expressed miRNAs in human cancersrdquo BMC Genomicsvol 11 supplement 4 article S5 2010

[59] M Kano N Seki N Kikkawa et al ldquoMiR-145 miR-133a andmiR-133b tumor-suppressive miRNAs target FSCN1 in eso-phageal squamous cell carcinomardquo International Journal ofCancer vol 127 no 12 pp 2804ndash2814 2010

[60] P C Enzinger and R J Mayer ldquoEsophageal cancerrdquo The NewEngland Journal of Medicine vol 349 no 23 pp 2241ndash22522003

[61] B He B Yin B Wang Z Xia C Chen and J Tang ldquomicroR-NAs in esophageal cancer (Review)rdquo Molecular MedicineReports vol 6 no 3 pp 459ndash465 2012

[62] Z Xie G Chen X Zhang et al ldquoSalivary microRNAs aspromising biomarkers for detection of esophageal cancerrdquo PLoSONE vol 8 no 4 Article ID e57502 2013

Complexity 9

[63] J Wan W Wu Y Che N Kang and R Zhang ldquoInsights intothe potential use of microRNAs as a novel class of biomarkersin esophageal cancerrdquo Diseases of the Esophagus vol 29 no 5pp 412ndash420 2016

[64] B-J Zhang H-Y Gong F Zheng D-J Liu and H-X LiuldquoUp-regulation of miR-335 predicts a favorable prognosis inesophageal squamous cell carcinomardquo International Journal ofClinical and Experimental Pathology vol 7 no 9 pp 6213ndash62182014

[65] A Jemal R Siegel E Ward et al ldquoCancer statistics 2006rdquo CAA Cancer Journal for Clinicians vol 56 no 2 pp 106ndash130 2006

[66] N M A White T T Bao J Grigull et al ldquoMiRNA profilingfor clear cell renal cell carcinoma biomarker discovery andidentification of potential controls and consequences ofmiRNAdysregulationrdquoThe Journal of Urology vol 186 no 3 pp 1077ndash1083 2011

[67] N M A White H W Z Khella J Grigull et al ldquoMiRNAprofiling in metastatic renal cell carcinoma reveals a tumour-suppressor effect formiR-215rdquoBritish Journal of Cancer vol 105no 11 pp 1741ndash1749 2011

[68] R Siegel DNaishadham andA Jemal ldquoCancer statistics 2012rdquoCA A Cancer Journal for Clinicians vol 62 no 1 pp 10ndash292012

[69] W M Linehan ldquoGenetic basis of kidney cancer role ofgenomics for the development of disease-based therapeuticsrdquoGenome Research vol 22 no 11 pp 2089ndash2100 2012

[70] W M Linehan M M Walther and B Zbar ldquoThe genetic basisof cancer of the kidneyrdquo The Journal of Urology vol 170 no 6part 1 pp 2163ndash2172 2003

[71] W M Linehan and B Zbar ldquoFocus on kidney cancerrdquo CancerCell vol 6 no 3 pp 223ndash228 2004

[72] U Senanayake S Das P Vesely et al ldquomiR-192 miR-194 miR-215 miR-200c and miR-141 are downregulated and their com-mon target ACVR2B is strongly expressed in renal childhoodneoplasmsrdquo Carcinogenesis vol 33 no 5 pp 1014ndash1021 2012

[73] H Hidaka N Seki H Yoshino et al ldquoTumor suppressivemicroRNA-1285 regulates novel molecular targets aberrantexpression and functional significance in renal cell carcinomardquoOncotarget vol 3 pp 44ndash57 2012

[74] M Redova A Poprach J Nekvindova et al ldquoCirculating miR-378 andmiR-451 in serum are potential biomarkers for renal cellcarcinomardquo Journal of Translational Medicine vol 10 article 552012

[75] A Zhao G Li M Peocrsquoh C Genin and M Gigante ldquoSerummiR-210 as a novel biomarker for molecular diagnosis ofclear cell renal cell carcinomardquo Experimental and MolecularPathology vol 94 no 1 pp 115ndash120 2013

[76] MA THildebrandt J Gu J Lin et al ldquoHsa-miR-9methylationstatus is associated with cancer development and metastaticrecurrence in patients with clear cell renal cell carcinomardquoOncogene vol 29 no 42 pp 5724ndash5728 2010

[77] Y Liu D A Tennant Z Zhu J K Heath X Yao and S HeldquoDiMEA scalable diseasemodule identification algorithmwithapplication to glioma progressionrdquo PLoS ONE vol 9 no 2Article ID e86693 2014

[78] L Wong Z-H You Z Ming J Li X Chen and Y-A HuangldquoDetection of interactions between proteins through rotationforest and local phase quantization descriptorsrdquo InternationalJournal of Molecular Sciences vol 17 no 1 p 21 2015

[79] Z H You L Zhu C H Zheng H Yu S Deng and Z JildquoPrediction of protein-protein interactions from amino acid

sequences using a novel multi-scale continuous and discontin-uous feature setrdquo BMC Bioinformatics vol 15 supplement 15 pS9 2014

[80] L Zhu W-L Guo S-P Deng and D S Huang ldquoChIP-PITenhancing the analysis of ChIP-Seq data using convex-relaxedpair-wise interaction tensor decompositionrdquo IEEEACM Trans-actions onComputational Biology and Bioinformatics vol 13 no1 pp 55ndash63 2016

[81] D S Huang L Zhang K Han S Deng K Yang andH ZhangldquoPrediction of protein-protein interactions based on protein-protein correlation using least squares regressionrdquo CurrentProtein and Peptide Science vol 15 no 6 pp 553ndash560 2014

[82] L Zhu Z You D S Huang and BWang ldquot-LSE a novel robustgeometric approach for modeling protein-protein interactionnetworksrdquo PLoS ONE vol 8 no 4 Article ID e58368 2013

[83] D S Huang and J-X Du ldquoA constructive hybrid structure opti-mization methodology for radial basis probabilistic neural net-worksrdquo IEEE Transactions on Neural Networks vol 19 no 12pp 2099ndash2115 2008

[84] D S Huang Systematic Theory of Neural Networks for PatternRecognition Publishing House of Electronic Industry of ChinaBeijing China 1996

[85] D S Huang andW Jiang ldquoA general CPL-AdSmethodology forfixing dynamic parameters in dual environmentsrdquo IEEE Trans-actions on Systems Man and Cybernetics Part B Cyberneticsvol 42 no 5 pp 1489ndash1500 2012

[86] D S Huang ldquoRadial basis probabilistic neural networks modeland applicationrdquo International Journal of Pattern Recognitionand Artificial Intelligence vol 13 no 7 pp 1083ndash1101 1999

Submit your manuscripts athttpswwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of


Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of


Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of


Mathematical PhysicsAdvances in

Complex AnalysisJournal of


OptimizationJournal of


CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of


Operations ResearchAdvances in

Journal of


Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences


The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Algebra

Discrete Dynamics in Nature and Society



Decision SciencesAdvances in

Journal of


Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

2 Complexity

and disease and to facilitate the diagnose prevention andtreatment human complex disease [30ndash33] However usingexperimental methods to identify miRNA-disease associa-tion is expensive and time-consuming As themiRNA-relatedtheories are becoming more and more common such as theprediction model about miRNA and disease the function ofmiRNA in biological processes and signaling pathways newtherapies are urgently needed for the treatment of complexdisease it is necessary to develop powerful computationalmethods to reveal potential miRNA-disease associations [1215 20 34ndash40]

Previous studies had shown that functionally similarmiRNAs always appear in similar diseases therefore manycomputational models were proposed to identify novelmiRNA-disease associations [13 41ndash46] For example Jianget al [31] analyzed and improved disease-gene predictionmodel introduced the principle of hypergeometric distri-bution and how to use it and discussed its application inprediction model and its actual effect In order to real-ize the prediction function of the improved model theyused different types of dataset including miRNA functionalsimilarity data disease phenotype similarity data and theknown human disease-miRNA association data Thereforethe prediction accuracy of this method is greatly impactedby miRNA neighbor information and miRNA-target inter-action prediction Chen et al [47] reported a new methodHGIMDA to identify novel miRNA-disease association byusing heterogeneous graph inference This algorithm can getbetter prediction accuracy by integrating known miRNA-disease associations miRNA functional similarity diseasesemantic similarity and Gaussian interaction profile kernelsimilarity for diseases and miRNAs In addition HGIMDAcould be applied for new diseases and new miRNAs whichdo not have any known association Li et al [48] proposed thecomputational model Matrix completion for MiRNA-diseaseassociation prediction (MCMDA) to predict miRNA-diseaseassociations This model only uses known miRNA-diseaseassociations and achieved better prediction performanceThelimitation of MCMDA is that it could not be applied for newdiseases and new miRNAs which do not have any knownassociation You et al [49] developed model Path-BasedMiRNA-Disease Association Prediction (PBMDA) to predictmiRNA-disease associations by integrating known humanmiRNA-disease associations miRNA functional similaritydisease semantic similarity and Gaussian interaction profilekernel similarity formiRNAs and diseases Depth-first searchalgorithm was used in this model to identify novel miRNA-disease associations Benefiting from effective algorithm andreliable biological datasets PBMDA has better predictionperformance Furthermore Xu et al [50] introduced anapproach to identify disease-related miRNAs by the miRNAtarget-dysregulated network (MTDN) Furthermore in orderto distinguish and identify disease-related miRNAs fromcandidate a SVM classifier based on radial basis functionand the lib SVM package had been proposed Researcheshave shown that miRNAs can functionally interact withenvironmental factors (EFs) to affect and determine humancomplex disease Chen [51] proposed model miREFRWRto predict the association between disease and miRNA-EF

interactions Random walks theory was applied on miRNAsimilarity network and EF similarity network In additiondrug chemical structure similarity miRNA function sim-ilarity and networked-based similarity were also used inmiREFRWR Based on these biological datasets and efficientcalculation method miREFRWR could be an effective toolin computational biology What is more Chen et al [52]also proposed a computational model RKNNMDA to predictthe potential associations between miRNA and disease Fourbiological datasets experimentally verified human miRNA-disease associations miRNA functional similarity diseasesemantic similarity and Gaussian interaction profile kernelsimilarity for miRNAs and diseases were integrated intoRKNNMDA It can be found that the prediction accuracyof RKNNMDA is excellent Moreover RKNNMDA could beapplied for newdiseaseswhich do not have any known relatedmiRNA information

Generally speaking current prediction model onmiRNA-disease association is still demonstrating some short-comings For example unreliable datasets have a great influ-ence on the accuracy of prediction model such as miRNA-target interactions and disease-genes associations Inaddition for miRNAs and diseases which do not haveany known associations we cannot use some of the existingmodels to predict its relevant information In other words weneed to design and develop a new effective computationalmodel According to the assumption that functionally similarmiRNAs always appear in similar diseases we introduce themodel of Collaborative Matrix Factorization for MiRNA-Disease Association prediction (CMFMDA) to reveal novelmiRNA-disease association by integrating experimentallyvalidated miRNA-disease associations miRNA functionalsimilarity information and disease semantic similarityinformation For CMFMDA we can obtain its test resultswith three different ways 5-fold CV Local LOOCV andglobal LOOCV The AUCs of these three methods are08697 08318 and 08841 respectively which suggest thatCMFMDA is a reliable and efficient prediction model Andthen we use two cases Esophageal Neoplasms and KidneyNeoplasms to evaluate the performance of CMFMDA Inboth of these two important diseases 42 and 41 out of top 50predicted miRNA-disease associations were confirmed byrecent experimental literatures respectively In additionexperiments show that CMFMDA can be applied for diseasesand miRNAs without any known association

2 Materials and Methods

21 HumanmiRNA-Disease Associations We obtained infor-mation about the associations between miRNA and dis-ease from HMDD including 5430 experimentally confirmedhuman miRNA-diseases associations about 383 diseases and495 miRNAs Adjacency matrix 119860 is proposed to describethe association between miRNA and disease If miRNA 119898(119894)is associated with disease 119889(119895) the entity 119860(119898(119894) 119889(119895)) is1 otherwise 0 Furthermore we declared two variables nmand nd to represent the number of miRNAs and diseasesinvestigated in this paper respectively

Complexity 3



1198631198811 (119863) = sum119889isin119879(119863)

1198631119863 (119889)

1198631119863 (119889)

=


(1)



119878119863 (119889 (119894) 119889 (119895)) =sum119905isin119879(119894)cap119879(119895) (1198631119894 (119905) + 1198631119895 (119905))

1198631198811 (119894) + 1198631198811 (119895) (2)







[119880 119878 119881] = 119878119881119863 (119884 119896)

119860 = 11988011987812119896

119861 = 11988111987812119896

(3)


min119860119861

10038171003817100381710038171003817119884 minus 119860119861119879100381710038171003817100381710038172

119865+ 120582119897 (1198602119865 + 1198612119865)

+ 12058211989810038171003817100381710038171003817119878119872 minus 11986011986011987910038171003817100381710038171003817

2

119865+ 120582119889

10038171003817100381710038171003817119878119863 minus 119861119861119879100381710038171003817100381710038172

119865

119860 = (119884119861 + 120582119889119878119889119860) (119861119879119861 + 120582119897119868119896 + 120582119889119860119879119860)minus1

119861

= (119884119879119860 + 120582119898119878119898119861) (119860119879119860 + 120582119897119868119896 + 120582119898119861119879119861)minus1

(4)


3 Results


4 Complexity






decomposition

Convergence

NoYes

(c)

(d)

(b)

(a)

mj Ym di Yd

A = US12k

B = VS12k

Y = AB

Singular value


minus1


Y = ＧＲ (Y Ydm)

Ydm =(Yd + Ym)

2







Complexity 5

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1TP

RGlobal LOOCV

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1

TPR

Local LOOCV






4 Discussion


6 Complexity







5 Conclusions





















Complexity 7


Disclosure




Acknowledgments


References






























8 Complexity



































Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




Complexity 3



1198631198811 (119863) = sum119889isin119879(119863)

1198631119863 (119889)

1198631119863 (119889)

=


(1)



119878119863 (119889 (119894) 119889 (119895)) =sum119905isin119879(119894)cap119879(119895) (1198631119894 (119905) + 1198631119895 (119905))

1198631198811 (119894) + 1198631198811 (119895) (2)







[119880 119878 119881] = 119878119881119863 (119884 119896)

119860 = 11988011987812119896

119861 = 11988111987812119896

(3)


min119860119861

10038171003817100381710038171003817119884 minus 119860119861119879100381710038171003817100381710038172

119865+ 120582119897 (1198602119865 + 1198612119865)

+ 12058211989810038171003817100381710038171003817119878119872 minus 11986011986011987910038171003817100381710038171003817

2

119865+ 120582119889

10038171003817100381710038171003817119878119863 minus 119861119861119879100381710038171003817100381710038172

119865

119860 = (119884119861 + 120582119889119878119889119860) (119861119879119861 + 120582119897119868119896 + 120582119889119860119879119860)minus1

119861

= (119884119879119860 + 120582119898119878119898119861) (119860119879119860 + 120582119897119868119896 + 120582119898119861119879119861)minus1

(4)


3 Results


4 Complexity






decomposition

Convergence

NoYes

(c)

(d)

(b)

(a)

mj Ym di Yd

A = US12k

B = VS12k

Y = AB

Singular value


minus1


Y = ＧＲ (Y Ydm)

Ydm =(Yd + Ym)

2







Complexity 5

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1TP

RGlobal LOOCV

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1

TPR

Local LOOCV






4 Discussion


6 Complexity







5 Conclusions





















Complexity 7


Disclosure




Acknowledgments


References






























8 Complexity



































Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




4 Complexity






decomposition

Convergence

NoYes

(c)

(d)

(b)

(a)

mj Ym di Yd

A = US12k

B = VS12k

Y = AB

Singular value


minus1


Y = ＧＲ (Y Ydm)

Ydm =(Yd + Ym)

2







Complexity 5

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1TP

RGlobal LOOCV

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1

TPR

Local LOOCV






4 Discussion


6 Complexity







5 Conclusions





















Complexity 7


Disclosure




Acknowledgments


References






























8 Complexity



































Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




Complexity 5

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1TP

RGlobal LOOCV

0 02 04 06 08 1 12FPR

0

01

02

03

04

05

06

07

08

09

1

TPR

Local LOOCV






4 Discussion


6 Complexity







5 Conclusions





















Complexity 7


Disclosure




Acknowledgments


References






























8 Complexity



































Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




6 Complexity







5 Conclusions





















Complexity 7


Disclosure




Acknowledgments


References






























8 Complexity



































Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




Complexity 7


Disclosure




Acknowledgments


References






























8 Complexity



































Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




8 Complexity



































Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




Complexity 9

































Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of











Volume 2014




Journal of











Journal of


Function Spaces






Algebra





Journal of




miRNA-Disease Association Prediction with Collaborative...

Documents

Transcript of miRNA-Disease Association Prediction with Collaborative...