modified weighted similarity in heterogeneous …...associations accurately. in this paper, weighted...

http://www.iaeme.com/IJARET/index.asp 141 [email protected]

International Journal of Advanced Research in Engineering and Technology (IJARET) Volume 11, Issue 9, September 2020, pp. 141-155, Article ID: IJARET_11_09_015

Available online at http://www.iaeme.com/IJARET/issues.asp?JType=IJARET&VType=11&IType=9

ISSN Print: 0976-6480 and ISSN Online: 0976-6499

DOI: 10.34218/IJARET.11.9.2020.015

© IAEME Publication Scopus Indexed

MODIFIED WEIGHTED SIMILARITY IN

HETEROGENEOUS GRAPH FOR PREDICTION

OF miRNA DISEASE ASSOCIATION

Rashmi J R

Research Scholar, Department of Studies in Computer Science,

University of Mysore, Mysore, India

Lalitha Rangarajan

Retired Professor, Department of Studies in Computer Science,

University of Mysore, Mysore, India

ABSTRACT

In the last decade lot of experimental research has witnessed and verified the

important roles of miRNA in the development of complex human diseases. Publicly

available MiRNA data and different analysis methodologies have given rise to the

development of many computational models to predict miRNA disease association.

Predicting accurate miRNA disease association is very essential for the proper

diagnosis and treatment of diseases. During past few years lots of computational

methods have been developed. However, each method has its own limitation and it has

not yet been possible to develop an efficient method that can predict miRNA-disease

associations accurately. In this paper, weighted meta-graph based computational

approach for predicting the association between diseases and miRNAs is proposed.

The proposed algorithm is designed by integrating available miRNA functional

similarity, miRNA similarity based on Environmental factors, miRNA similarity based

on diseases to get the average miRNA similarity, disease semantic similarity and

disease functional similarity are also integrated to get the average disease similarity.

AUC of 0.9617833 on global LOOCV has been achieved using the proposed method.

Key words: Meta-graph, MiRNA functional Similarity, Disease Semantic Similarity,

Heterogeneous Information Network

Cite this Article: Rashmi J R and Lalitha Rangarajan, Modified Weighted Similarity

in Heterogeneous Graph for Prediction of Mirna Disease Association, International

Journal of Advanced Research in Engineering and Technology, 11(9), 2020,

pp. 141-155.

http://www.iaeme.com/IJARET/issues.asp?JType=IJARET&VType=11&IType=9

Modified Weighted Similarity in Heterogeneous Graph for Prediction of Mirna Disease

Association


1. INTRODUCTIONS

MicroRNAs or MiRNA are short non-coding RNA molecules that are involved in different

physiological and developmental processes by controlling the gene expression of target

mRNAs. They play important roles in almost all kinds of cancer where they modulate key

processes during tumor genesis such as metastasis, Apoptosis, proliferation, or angiogenesis

[11]. Depending on the mRNA targets they regulate, they can act as oncogenes or as tumor

suppressor genes. Multiple links between microRNA biogenesis and cancer highlight its

significance for tumor diseases. However, mechanisms of their own regulation on the

transcriptional and posttranscriptional level in health and disease are only beginning to

emerge [11]. MiRNA disease association prediction is one of the emerging and essential

fields of research.

1.1. Background

It has been verified that miRNAs are the positive regulators. Experimental approaches for

identifying the miRNA disease associations are having high precision but they are considered

as time consuming and expensive. In the past few years many experiments have been

conducted lot of data is being collected and stored in the databases such as dbDEMC [5] and

HMDD [6] which contains human miRNA diseases which are experimentally verified. Still

the research in the field of miRNA disease association is ongoing; in order to complement the

biologists in their work many computational approaches have been proposed. In recent years

many computational approaches have given outstanding performance such as most of the

methods are based on the assumption that similar miRNAs are tend to associated with the

similar diseases. Chen et al [1] proposed the RWRMDA to predict disease associated

miRNAs. In this method the global similarity for miRNAs and diseases is taken into account.

The drawback associated with this method is it is not possible to predict the miRNAs for the

diseases which don’t have related miRNAs. Wang et al [3] showed that the relationship

between different diseases can be represented by a structure called DAG (Directed Acyclic

Graph). Disease semantic similarity was calculated first and based on that the functional

similarity between miRNAs is calculated. The Functional similarity has been utilized by

many researchers in their work for predicting the disease related miRNAs. MiRNA functional

similarity is calculated by assigning different weights based on the miRNA family and cluster

in HDMP [7], and then miRNAs are ranked by their final scores. Improved Random walk is

applied to set the scores for the candidate miRNA in MIDP [8], which implies that the

miRNA with targets has the higher possibility of being associated with disease

All the methods mentioned above have shown significant performances, but they were

unable to predict the miRNAs for the diseases which does not have any related miRNAs.

Chen et al [9] proposed a method called HGIMDA to uncover potential miRNA-disease

associations by integrating miRNA functional similarity, disease semantic similarity,

Gaussian interaction profile kernel similarity, and experimentally verified miRNA-disease

associations into a heterogeneous graph. They were able to achieve 0.87 on global LOOCV.

You et al [10] proposed the Path based MiRNA disease association by integrating known

human miRNA-disease associations, miRNA functional similarity, disease semantic

similarity, and Gaussian interaction profile kernel similarity for miRNAs and diseases. In this

model heterogeneous graph of three interlinked three sub graphs is constructed and depth first

search algorithm is adopted to infer potential miRNAs associated with a disease they achieved

the reliable performance 0.916 on global LOOCV. LRSSMDA was proposed by Chen et

al[11] where they extracted miRNA functional similarity, disease semantic similarity,

Gaussian interaction profile kernel similarity and applied the Laplacian Regularized Sparse

Subspace Learning to discover potential association between miRNAs and diseases These

Rashmi J R and Lalitha Rangarajan


methods showed outstanding performances but they faced challenges for integrating multiple

kernels in a better way.

Long et al [3] proposed WMGHMDA: a novel weighted meta-graph-based model for

predicting human microbe-disease association on heterogeneous information network.

Inspired from their method we also proposed a meta-graph model to predict miRNA disease

associations. Here we incorporated multiple sources of prior biological knowledge. This

model is capable of predicting the miRNAs for the disease without known associations. To

implement this method first heterogeneous information network is constructed by connecting

the integrated miRNA and disease similarity with the help of miRNA and disease bipartite

network.

2. MATERIALS AND METHODS

The proposed meta-graph algorithm iteratively enumerates meta-graphs associated with each

miRNA disease pair. Finally, the probability score for each miRNA disease pair is calculated

by summing up the contribution values of relevant weighted meta-graphs and prioritize

candidate miRNAs according to their probability score. In this work we have utilized miRNA

similarity based on environmental factors, miRNA similarity based on diseases, miRNA

functional similarity, disease semantic similarity, disease functional similarity. This

effectively boosts the improvement of prediction accuracy. The workflow for the proposed

method is shown below

Figure 1 Work Flow


Association


2.1. Adjacency Matrix

We have collected 5430 miRNA disease associations from the HMDD (Human MiRNA

Disease Database). These associations are represented in the form of the adjacency matrix

A(di, mj). 383 diseases and 642 miRNAs are taken into account for study. If there is

association between a disease di and miRNA mj, A(i)(j)=1 otherwise 0.

MiRNA Functional Similarity: miRNA–miRNA functional similarity scores are

downloaded from http://cmbi.bjmu.edu.cn/misim as proposed by Wang et al [3]. It is based

on the observation that genes with similar functions are associated with common diseases.

The matrix is denoted by MFS

MiRNA Similarity based on Environmental Factors: J Ha et al [12] proposed a new

method of calculating miRNA-miRNA similarity in biological context. Environmental factors

have been recently utilized as new means of inferring the relation between miRNAs. Inspired

by this work we have calculated the miRNA Similarity based on environmental factors as

provided in MiREnvironment database [13]. The database contains 800 miRNAs and 260

environmental factors such as drugs, cigarettes, alcohol, viruses, stress, radiation etc.

MiRNA similarity based on EFs is calculated using equation 1

√ √ ⁄ (1)

MEFS(i, j) is the number of common EFs between m(i) and m(j) as in [12]. T(i, i) denote

the sum of common EFs between m[i] and all other miRNAs (is nothing but sum of ith

row

elements of MEFS).Similarly, T(j, j) is similarity computed as the jth

column sum of MEFS

2.2. MiRNA Similarity based on Diseases

MiRNA Similarity based on diseases is calculated with the help of HMDD database in a

similar manner as in the case of MEFS. The following equation is used

√ √ ⁄ (2)

All the above 3 matrices are added and average is calculated to get the average Similarity

matrix and named as FMS

Final miRNA similarity [FMS] is calculated by using equation

FMS (i, j) = ((MFS(i, j)+MMEFS(i, j)+MMDS(i, j))/3 (3)

2.3. Disease Semantic Similarity

Disease Semantic similarity DSS is calculated as proposed in Wang et al [3] work. MeSH

database (http://www.ncbi.nlm.nih.gov/) provides a system for disease classification. It is

helpful for studying the relationship between diseases. Diseases can be described using a

DAG (Directed Acyclic Graph) in which the nodes represent different diseases while the link

represents the relationship between nodes. A disease D can be represented as ,

where represents the ancestor nodes of D including D itself, represents set of

corresponding links and Taking into account disease DAG, the contribution of

disease D can be calculated using the following equation.

{

{ } (4)

∑ (5)



With the increase in distance between disease D and its ancestor diseases their

contribution to the disease D decreases. Contribution of disease D to itself is one as it is at the

0th

layer. Contribution from ancestor disease is multiplied by semantic contribution decay

factor ∆ [0, 1]. According to literature [15, 16] the best value for ∆ is 0.5. Two diseases

sharing common parts of DAG will have higher semantic similarity. Disease semantic

similarity is calculated as per the formula in equation 6.

( ) ∑ ( )

( ) (6)

2.4. Disease Functional Similarity

Disease functional similarity scores are downloaded from https://github.com/guofei-tju/MDA-

SKF as proposed in Jiang et al [18]. We denote it by DFS

We calculate the Final Disease Similarity as in equation 7

FDS = (DSS(i, j)+DFS(i, j))/2 (7)

3. MODIFIED WEIGHTED META -GRAPH ALGORITHM FOR MIRNA

DISEASE ASSOCIATION (MWMGMDA)

As a first step, we construct the heterogeneous information network inspired from Long et

al[2] as shown in Figure 2 using FMS and FDS (equations 3, 7) and adjacency matrix of

disease and miRNA associations.

Figure 2 Heterogeneous Network

Meta-graph search algorithm is designed and implemented on this HIN. We get the

probability scores for each pair of miRNAs and diseases. These probability scores are ranked

in order to predict the disease related miRNAs.

Meta-graphs have got wide application and have been used in the representational

learning and recommendation system [20]. Inspired by WMGHMDA [2], we have used Meta-

graph concept to predict the missing associations between miRNA and disease. Meta-graph is

subset of HIN [2]. Formally, Meta-graph could be defined as sub-graph Gs = (V, E), where V

= {di/i = 1, 2, . . . , nd}∪{mj/j = 1, 2, . . . , nm} represents the set of nodes including diseases

and miRNAs, and E = {(ei, ej)} implies the set of edges including inter-layer relationship

connections in the bipartite network and intra-layer similarity connections in both disease


Association


similarity network and miRNA similarity network[20]. In this work we take six meta-graphs

for consideration, which are shown below:

Similarity linking

Bipartite linking

Figure 3 Meta-graphs

These Meta-graphs depict possible semantic relation between a seed disease and target

miRNAs. Product of all weight values of the edges existing in the Meta-graph is calculated.

Each meta-graph contributes to the prediction probability of the miRNA disease association

pair. Contribution values of the Meta-graph when there is no observed relationship between

them is

( ) ∑ ∑

(8)

In this method we generalize common un-weighted meta-graph to weighted meta-graph

.In weighted meta-graph the weight values of the intra layer edges represent the similarities

between diseases or miRNAs and the weight values between the intra layer edges denote the

possibility of the existence of the association between disease and miRNAs. If miRNA is

experimentally verified to be associated with disease weight values of the corresponding

edges equals to 1 otherwise 0. The importance of the meta-graph decreases as the number of

the intermediate nodes and the number of edges increase, Hence, in this work we have taken

into consideration only six meta-graphs to keep the number of edges less than 5 and number

of nodes less than 3. Only the graph with single path and dual path are taken into account.

Contribution from each meta-graph is calculated according to the formulas given below

(equations 9 to14)

( ) ( ) (9)

( ) ∑ ( ) (10)



( ) ∑ ( ) (11)

( ) ∑ ∑

(12)

( ) ∑ ∑

(13)

( )

∑ ∑ ( )

(14)

Here all the meta-graphs have different structures and characteristics so their contributions

to the predictions of the miRNAs disease associations will be biased so we introduce the bias

rating to describe the differential contribution of the different meta-graphs. Differences

between meta-graphs mainly depend upon on the number of given nodes which is directly

connected to the bipartite edge could be seed disease or target miRNA Each of the graphs

have 2,1,1,0,1,1 nodes respectively as there is no edge between seed disease or target miRNA

graph(d) has 0 nodes. Based on the assumption meta-graph with more nodes have more

potential. Graph ‘a’ has more potential to contribute to the disease miRNAs association, ‘d’

is known to contribute least and ‘e’ &‘f ’can contribute considerably in disease association

pair. e & f are dual path weighted graph as in this graph seed node has more semantic paths

connected to a target miRNA node. We have assigned different bias ratings for different

weighted meta-graphs based on their contribution.

Here we apply weighted meta-graph algorithm of HIN to traverse all relevant meta-

graphs. The prediction score P is defined and calculated by summing up the contribution

values of the meta-graphs as follows

( ) ∑ ∑

(15)

is the contribution value of the rth meta-graph which belongs to l

th type of weighted

meta-graph to the pair ( . Here N=6, is a category number of weighted meta-graph as 6

types of meta-graphs are taken into the consideration for study. M is the number of meta-

graphs that are included in each category, λ [0, 1] is the bias rating applied to different

meta-graphs in order to distinguish their contributions to final predictions probability. We

iteratively implement the above search process which is based weighted meta-graph search

algorithm until Pt converges. Iterative formula with matrix is shown below

(16)

In the above equation Id and Im denote the unit matrices of size nd (no of diseases) and

nm(no of miRNAs). λ values are empirically set as , , , , , 0.2. is the decay factor which is similar to restart probability

in the random walk with restart. The initial values of the matrix Pt is set to the normalized

values of a (adjacent Matrix) A(di, mj). The element in the matrix Pt at ith

row and jth

column is

the probability value of the disease association pair. is the Hambard Product. To control

the contributions of the dural paths we set the value of the γ=0.1. According to Wang et al

[26] the probabilities computed in equation 11 will converge if the Average miRNA similarity

(FMS) and Average disease similarity (FDS) are properly normalized according to equations

17 and 18

( ) ( )

√∑ √∑ ( )

(17)

( ) ( )

√∑ √∑ ( )

(18)


Association


4. PERFORMANCE EVALUATION

We have used global LOOCV based on the known miRNA-disease associations in HMDD

database to evaluate the performance of WMGMDA. Further, WMGMDA are compared with

two previous classical computational methods: HGIMDA [22], RLSMDA [23]. In LOOCV

evaluation, each known association in the database is regarded as the test sample in turn,

while the other known associations are regarded as training samples. The miRNA-diseases

without known association evidences are considered as candidate samples. The scores of all

miRNA-disease pairs could be obtained after WMGMDA was implemented. In global

LOOCV, the score of the test sample was compared with the scores of all the candidate

samples.

Finally, a Receiver Operating Characteristics curve (ROC) compares WMGMDA with all

the previous methods. In this curve, the true positive rate (TPR, sensitivity) and false positive

rate (FPR, 1-specificity) are plotted [24]. Sensitivity represents the percentage of miRNA-

disease test samples whose ranks exceeded the given threshold while specificity represents the

percentage of negative miRNA-disease associations whose ranks were lower than the

threshold [25]. The area under the ROC (AUC) was calculated to evaluate the accuracy of

MWMGMDA. If AUC=1, MWMGMDA proves to be a prefect performance. Having AUC

0.5 means that the method merely has a random prediction performance. The AUCs of

MWMGMDA, NPCMDA [27], RLSMDA [28] are 0.9617833, 0.9148398, and 0.8549943

respectively in global LOOCV.

Figure 4 ROC curve

--------- our method

---x---x---- RLSMDA

o-o-o-o NPCMDA

It can be observed that, the proposed method performs better than RLSMDA and

NPCMDA.

Another performance evaluation is through leave one disease out strategy. We deleted all

the miRNA disease associations associated with a particular disease. The proposed method is

able to predict the miRNAs associated with the given diseases. This demonstrates that

proposed method is capable of predicting diseases which doesn’t have any associated

miRNAs.



We studied the effect of the parameter μ on the performance of proposed method. μ can

take values between 0 to 0.9. We have calculated the AUC of the proposed method at

different values of μ we were able to get the best performance at 0.2. The table below shows

the AUC obtained at different values of μ.

Table 1 AUC Performance

μ. AUC

0.1 0.953447

0.2 0.961972

0.3 0.951489

0.4 0.93862

0.5 0.9175464

0.6 0.895443

0.7 0.873503

0.8 0.85854

0.9 0.83

5. RESULTS

We have predicted the miRNAs for the 10 diseases which are given in Table 2.

Table 2 Results

Disease Number MiRNAs Confirmed in Top 50 Predictions

Kidney Cancer 50

Ovarian Cancer 48

Thyroid Cancer 45

Lung Cancer 44

Pancreatic Cancer 49

Brain Cancer 43

Leukemia 48

Cervical Cancer 47

Stomach Cancer 49

Breast Cancer 45

Top 50 predictions for 3 diseases are shown in Tables 3 to 5

MiRNAs predicted by the proposed method are confirmed by the three experimentally

verified databases namely dbDEMC[5], MirCANCER [28], HMDD 3.0[6]

Kidney Cancer

Table 3 Top 50 predictions for Kidney Cancer

MiRNA Confirmed by

hsa-mir-155 dbDEMC

hsa-mir-146a dbDEMC, HMDD 3.0,MiRCancer

hsa-mir-125b dbDEMC, MiRCancer

hsa-mir-210 dbDEMC, HMDD 3.0



hsa-mir-34a dbDEMC


hsa-mir-29a dbDEMC

hsa-mir-27a dbDEMC


Association


hsa-mir-221 dbDEMC, MiRCancer

hsa-mir-200b dbDEMC

hsa-mir-223 dbDEMC

hsa-mir-20a dbDEMC

hsa-mir-140 dbDEMC

hsa-mir-31 dbDEMC

hsa-mir-18a dbDEMC

hsa-mir-146b dbDEMC

hsa-mir-19a dbDEMC

hsa-mir-218 dbDEMC

hsa-mir-34c dbDEMC

hsa-mir-125a dbDEMC

hsa-let-7c dbDEMC

hsa-mir-106b dbDEMC, HMDD 3.0

hsa-mir-195 dbDEMC

hsa-mir-200a dbDEMC

hsa-mir-222 dbDEMC

hsa-let-7b dbDEMC

hsa-mir-423 dbDEMC

hsa-mir-107 dbDEMC

hsa-mir-30a dbDEMC, HMDD3.0, MiRCancer

hsa-mir-302b dbDEMC

hsa-mir-199a dbDEMC

hsa-mir-143 dbDEMC, MiRCancer

hsa-mir-139 dbDEMC

hsa-mir-205 dbDEMC

hsa-mir-122 dbDEMC

hsa-mir-196a dbDEMC

hsa-mir-34b dbDEMC

hsa-mir-150 dbDEMC

hsa-let-7g dbDEMC

hsa-mir-16 dbDEMC

hsa-mir-373 dbDEMC

hsa-mir-135b dbDEMC

hsa-let-7e dbDEMC

hsa-mir-20b dbDEMC

hsa-mir-99b dbDEMC

hsa-mir-130a dbDEMC

hsa-mir-99a dbDEMC

hsa-mir-429 dbDEMC

Ovarian Cancer

Table 4 Top 50 predictions for Ovarian Cancer

MiRNA Confirmed by


hsa-mir-98 dbDEMC,HMDD 3.0


hsa-mir-342 dbDEMC


hsa-mir-15a dbDEMC



hsa-mir-222 dbDEMC,HMDD 3.0,MirCancer



hsa-mir-196a dbDEMC,MirCancer

hsa-mir-10a dbDEMC


hsa-mir-122 dbDEMC

hsa-mir-423 dbDEMC

hsa-mir-206 HMDD 3.0

hsa-mir-23a dbDEMC,HMDD 3.0

hsa-mir-520d dbDEMC



hsa-mir-15b dbDEMC,HMDD 3.0


hsa-mir-29c dbDEMC,HMDD 3.0


hsa-mir-484 dbDEMC

hsa-mir-204 dbDEMC



hsa-mir-608 UNCONFIRMED



hsa-mir-203 HMDD 3.0


hsa-mir-483 dbDEMC





hsa-mir-345 dbDEMC



hsa-mir-301a dbDEMC

hsa-mir-144 dbDEMC

hsa-mir-196b HMDD 3.0

hsa-mir-181c dbDEMC


hsa-mir-424 dbDEMC

hsa-mir-326 dbDEMC


hsa-mir-218-2 UNCONFIRMED



Association


Pancreatic Cancer

Table 5 Top 50 predictions for Pancreatic Cancer

MiRNA Confirmed By

hsa-mir-29a dbDEMC,HMDD3.0,MirCANCER

hsa-mir-99b dbDEMC

hsa-mir-342 dbDEMC

hsa-mir-181d dbDEMC

hsa-mir-335 dbDEMC,HMDD3.0,MirCANCER

hsa-mir-125a dbDEMC

hsa-mir-19a dbDEMC

hsa-mir-20b dbDEMC

hsa-mir-106b dbDEMC


hsa-mir-30d dbDEMC

hsa-mir-302b dbDEMC

hsa-mir-30a dbDEMC

hsa-mir-22 dbDEMC

hsa-mir-23b dbDEMC




hsa-mir-520d dbDEMC


hsa-mir-148b dbDEMC,HMDD3.0,MirCANCER

hsa-mir-370 dbDEMC

hsa-mir-206 dbDEMC

hsa-mir-144 dbDEMC

hsa-mir-29c dbDEMC,HMDD3.0,MirCANCER

hsa-mir-27b dbDEMC

hsa-mir-19b dbDEMC



hsa-mir-520c dbDEMC


hsa-mir-30b dbDEMC

hsa-mir-181c dbDEMC,HMDD3.0,MirCANCER


hsa-mir-130b dbDEMC

hsa-mir-330 dbDEMC

hsa-mir-449b dbDEMC

hsa-mir-134 dbDEMC

hsa-mir-193b dbDEMC

hsa-mir-26b dbDEMC

hsa-mir-372 UNCONFIRMED

hsa-mir-423 dbDEMC

hsa-mir-326 dbDEMC

hsa-mir-149 dbDEMC

hsa-mir-503 dbDEMC

hsa-mir-339 dbDEMC

hsa-mir-874 dbDEMC

hsa-mir-490 dbDEMC

hsa-mir-130a dbDEMC




6. CONCLUSION

Weighted meta-graph for disease association is very effective in predicting miRNAs

associated with disease. We are able to predict the miRNAs for the diseases which are not

having any associations. In the proposed method we have integrated miRNA functional data,

disease based miRNA functional data, and Environmental factors data, Disease functional

Data, Disease semantic data. Rigorous experiments, performance comparison with other

methods and cross validation suggest the proposed method performs better. In future to

improve the prediction accuracy we are contemplating to integrate MiRNA target

information.

REFERENCES

[1] Chen X, Liu MX, Yan GY. RWRMDA: predicting novel human microRNA-disease

associations. Mol Biosyst. 2012; 8:2792–98. https://doi.org/10.1039/c2mb25180a

[2] Long, Y., & Luo, J. (2019). Open Access WMGHMDA : a novel weighted meta-graph-based

model for predicting human microbe-disease association on heterogeneous information

network, 1–18.

[3] Wang, D., Wang, J., Lu, M., Song, F., & Cui, Q. (2010). Inferring the human microRNA

functional similarity and functional network based on microRNA-associated diseases

Availability :, 26(13), 1644–1650. https://doi.org/10.1093/bioinformatics/btq241

[4] Xuan, P., Han, K., Guo, M., Guo, Y., Li, J., Ding, J., et alc(2013). Correction: prediction of

micrornas associated with human diseases based on weighted k most similar neighbors. PLoS

ONE 8:e70204. doi: 10.1371/annotation/a076115e-dd8c-4da7-989d-c1174a8cd31e

[5] Yang, Z., Ren, F., Liu, C., He, S., Sun, G., Gao, Q., et al. (2010). DBDEMC: a database of

differentially expressed miRNAs in human cancers. BMC Genomics 11(Suppl. 4):S5. doi:

10.1186%2F1471-2164-11-S4-S5

[6] Li, Y., Qiu, C., Tu, J., Geng, B., Yang, J., Jiang, T., et al. (2014). HMDD v2.0: a database for

experimentally supported human microRNA and disease associations. Nucleic Acids Res.

42(Database issue), D1070–D1074. doi: 10.1093/nar/gkt1023

[7] Xuan, P., Han, K., Guo, M., Guo, Y., Li, J., Ding, J., et al. (2013). Correction: prediction of

micrornas associated with human diseases based on weighted k most similar neighbors. PLoS

ONE 8:e70204. doi: 10.1371/annotation/a076115e-dd8c-4da7-989d-c1174a8cd31e

[8] Xuan, P., Han, K., Guo, Y., Li, J., Li, X., Zhong, Y., et al. (2015). Prediction of potential

disease-associated microRNAs based on random walk. Bioinformatics 31:1805. doi:

10.1093/bioinformatics/btv039

[9] HGIMDA: Heterogeneous graph inference for miRNA-disease association prediction Xing

Chen1, Cheng gang Clarence Yan, Xu Zhang, Zhu-Hong You, Yu-An Huang, Gui-Ying Yan

[10] You, Z. H., Huang, Z. A., Zhu, Z., Yan, G. Y., Li, Z. W., Wen, Z., et al. (2017). PBMDA: a

novel and effective path-based computational model for miRNA-disease association

prediction. PLoS Comput. Biol. 13:e1005455. doi: 10.1371/journal.pcbi.1005455

[11] Winter J., Diederichs S. (2011) MicroRNA Biogenesis and Cancer. In: Wu W. (eds)

MicroRNA and Cancer. Methods in Molecular Biology (Methods and Protocols), vol 676.

Humana Press, Totowa, NJ. https://doi.org/10.1007/978-1-60761-863-8_1

[12] Jihwan Haa, Hyunjin Kima, Youngmi Yoonb and Sanghyun Parka, * A method of extracting

disease-related microRNAs through the propagation algorithm using the environmental factor


Association


based global miRNA network Bio-Medical Materials and Engineering 26 (2015) S1763–

S1772, DOI 10.3233/BME-151477, IOS Press

[13] Q. Yang, C. Qiu, J. Yang, Q. Wu and Q. Cui, MIRenvironment database: Providing a bridge

for microRNAs, environmental factors and phenotypes, Bioinformatics 27 (2011), 3329–

3330)

[14] Ambros V (2004) The functions of animal microRNAs. Nature 431: 350±355.

https://doi.org/10.1038/nature02871 PMID: 15372042

[15] Chen X, You Z, Yan G, Gong D (2016) IRWRLDA: Improved Random Walk with Restart for

LncRNADisease Association prediction. Oncotarget 7: 57919±57931.

https://doi.org/10.18632/oncotarget.11141 PMID: 27517318

[16] Huang Y, Chen X, You Z, Huang D, Chan K (2016) ILNCSIM: improved lncRNA functional

similarity calculation model. Oncotarget 7: 25902±25914.

https://doi.org/10.18632/oncotarget.8296 PMID: 27028993

[17] van Laarhoven T, Nabuurs SB, Marchiori E (2011) Gaussian interaction profile kernels for

predicting drug-target interaction. Bioinformatics 27: 3036±3043.

https://doi.org/10.1093/bioinformatics/btr500 PMID: 21893517

[18] Dong Y, Chawla NV, Swami A. metapath2vec: Scalable representation learning for

heterogeneous networks. In: Proceedings of the 23rd ACM SIGKDD International Conference

on Knowledge Discovery and Data

[19] Mining; 2017. p. 135–44.

[20] Zhao H, Yao QM, Li JD, Song YQ, Lee DL. Meta-Graph based recommendation fusion over

heterogeneous information networks. In Proceedings of the 23rd ACM SIGKDD International

Conference on Knowledge Discovery and Data Mining; 2017. p. 635–44. https://doi.org/

10.1145/3097983.3098063.

[21] Fu TY, Lee WC, Lei Z. Hin2vec: Explore meta-paths in heterogeneous information networks

for representation learning. In: International Conference on Information and Knowledge

Management; 2017.p. 1797–806. https://doi.org/10.1145/3132847.3132953.

[22] Shi Y, Gui H, Zhu Q, Kaplan L, Han JW. Aspem: Embedding learning by aspects in

heterogeneous information networks. In: International Conference on Data Ming; 2018. arXiv

preprint arXiv:1803.01848

[23] Chen X, Yan C, Zhang X, You Z, Huang Y, Yan G. HGIMDA: Heterogeneous Graph

Inference for MiRNA-Disease Association prediction. Oncotarget. 2016; 7:65257- 65269. doi:

10.18632/oncotarget.11251.

[24] Chen X, Yan GY. Semi-supervised learning for potential human microRNA-disease

associations inference. Sci Rep. 2014; 4:5501.

[25] Jiang Q, Hao Y, Wang G, Juan L, Zhang T, Teng M, Liu Y, Wang Y. Prioritization of disease

microRNAs through a human phenome-microRNAome network. BMC Syst Biol. 2010; 4:S2.

[26] Oved K, Morag A, Pasmanikchor M, Oronkarni V, Shomron N, Rehavi M, Stingl JC, Gurwitz

D. Genome-wide miRNA expression profiling of human lymphoblastoid cell lines identifies

tentative SSRI antidepressant response biomarkers. Pharmacogenomics. 2012; 13:1129-1139.



[27] Wang WH, Yang S, Li J. Drug target predictions based on heterogeneous graph inference. In:

Proceedings of the Pacific Symposium; 2013. p. 53–64.

https://doi.org/10.1142/9789814447973_0006.

[28] Chen, X., Yan, G. Semi-supervised learning for potential human microRNA-disease

associations inference. Sci Rep 4, 5501 (2015). https://doi.org/10.1038/srep05501Chen, X.,

Yan, G. Semi-supervised learning for potential human microRNA-disease associations

inference. Sci Rep 4, 5501 (2015). https://doi.org/10.1038/srep05501

[29] Gu, C., Liao, B., Li, X. et al. Network Consistency Projection for Human miRNA-Disease

Associations Inference. Sci Rep 6, 36054 (2016). https://doi.org/10.1038/srep36054

[30] Boya Xie, Qin Ding, Hongjin Han, Di Wu, miRCancer: a microRNA–cancer association

database constructed by text mining on literature, Bioinformatics, Volume 29, Issue 5, 1

March 2013, Pages 638–644, https://doi.org/10.1093/bioinformatics/btt014

[31] Ambros V (2001) microRNAs: tiny regulators with great potential. Cell 107: 823±826. PMID:

11779458

modified weighted similarity in heterogeneous …...associations accurately. in this paper, weighted...

Documents