---
title: "miRSM: inferring miRNA sponge modules in heterogeneous data"
author: "\\
Junpeng Zhang (zhangjunpeng411@gmail.com)\\
School of Engineering, Dali University"
date: '`r Sys.Date()`'
output:
BiocStyle::html_document:
toc: yes
BiocStyle::pdf_document:
toc: yes
vignette: >
%\VignetteIndexEntry{miRSM: inferring miRNA sponge modules in heterogeneous data}
%\VignettePackage{miRSM}
% \VignetteEngine{knitr::rmarkdown}
% \usepackage[utf8]{inputenc}
% \VignetteEncoding{UTF-8}
---
```{r style, echo=FALSE, results="asis", message=FALSE}
BiocStyle::markdown()
knitr::opts_chunk$set(tidy = FALSE,
warning = FALSE,
message = FALSE)
```
```{r echo=FALSE, results='hide', message=FALSE}
library(miRSM)
```
# Introduction
MicroRNAs (miRNAs) play key roles in many biological processes including cancers [1-5]. Thus, uncovering miRNA functions and regulatory mechanisms is important for gene diagnosis and therapy.
Previous studies [6-9] have shown that a pool of coding and non-coding RNAs that shares common miRNA biding sites competes with each other, thus alter miRNA activity. The corresponding regulatory mechanism is named competing endogenous RNA (ceRNA) hypothesis [10]. These RNAs are called ceRNAs or miRNA sponges or miRNA decoys, and include long non-coding RNAs (lncRNAs), pseudogenes, circular RNAs (circRNAs) and messenger RNAs (mRNAs), etc. To study the module-level properties of miRNA sponges, it is necessary to identify miRNA sponge modules. The miRNA sponge modules will help to reveal the biological mechanism in cancer.
To speed up the research of miRNA sponge modules, we develop an R/Bioconductor package 'miRSM' to infer miRNA sponge modules. Unlike the existing R/Bioconductor packages ('miRspongeR' and 'SPONGE'), 'miRSM' focuses on identifying miRNA sponge modules by integrating expression data and miRNA-target binding information instead of miRNA sponge interaction networks.
# Identification of gene modules
Given matched ceRNA and mRNA expression data, we infer gene modules by using several methods from 21 packages, including 'WGCNA', 'GFA', 'igraph', 'ProNet', 'NMF', 'stats', 'flashClust', 'dbscan', 'subspace', 'mclust', 'SOMbrero', 'ppclust', 'biclust', 'runibic', 'iBBiG', 'fabia', 'BicARE', 'isa2', 's4vd', 'BiBitR' and 'rqubic'. We assemble these methods into 7 functions: module_WGCNA, module_GFA, module_igraph, module_ProNet, module_NMF, module_clust and module_biclust.
## Load BRCA sample data
The BRCA sample data includes matched miRNA, lncRNA, mRNA expression data, putative miRNA-target binding information and BRCA-related genes (lncRNAs and mRNAs).
```{r, eval=TRUE, include=TRUE}
data(BRCASampleData)
```
## module_WGCNA
By using WGCNA method [11], we identify co-expressed gene modules from matched ceRNA and mRNA expression data.
```{r, eval=TRUE, include=TRUE}
modulegenes_WGCNA <- module_WGCNA(ceRExp[, seq_len(80)],
mRExp[, seq_len(80)])
modulegenes_WGCNA
```
## module_GFA
The gene modules are identified by using GFA method [12, 13] from matched ceRNA and mRNA expression data.
```{r, eval=FALSE, include=TRUE}
modulegenes_GFA <- module_GFA(ceRExp[seq_len(20), seq_len(15)],
mRExp[seq_len(20), seq_len(15)],
iter.max = 2600)
modulegenes_GFA
```
## module_igraph
By using 'igraph' package [14], we infer gene modules from matched ceRNA and mRNA expression data. In the 'igraph' package, we can select "betweenness", "greedy", "infomap", "prop", "eigen", "louvain" and "walktrap" methods for gene module identification. The default method is "greedy".
```{r, eval=TRUE, include=TRUE}
modulegenes_igraph <- module_igraph(ceRExp[, seq_len(10)],
mRExp[, seq_len(10)])
modulegenes_igraph
```
## module_ProNet
In the 'ProNet' package, we can select FN [15], MCL [16], LINKCOMM [17] and MCODE [18] for gene module identification. The default method is MCL.
```{r, eval=TRUE, include=TRUE}
modulegenes_ProNet <- module_ProNet(ceRExp[, seq_len(10)],
mRExp[, seq_len(10)])
modulegenes_ProNet
```
## module_NMF
By using 'NMF' package [20], we infer gene modules from matched ceRNA and mRNA expression data. In the 'NMF' package, we can select "brunet", "Frobenius", "KL", "lee", "nsNMF", "offset", "siNMF", "snmf/l" and "snmf/r" methods for gene module identification. The default method is "brunet".
```{r, eval=TRUE, include=TRUE}
# Reimport NMF package to avoid conflicts with DelayedArray package
library(NMF)
modulegenes_NMF <- module_NMF(ceRExp[, seq_len(10)],
mRExp[, seq_len(10)])
modulegenes_NMF
```
## module_clust
We Identify gene modules from matched ceRNA and mRNA expression data using a series of clustering packages, including stats [21], flashClust [22], dbscan [23], subspace [24], mclust [25], SOMbrero [26] and ppclust [27]. The clustering methods include "kmeans", "hclust", "dbscan", "clique", "gmm", "som" and "fcm". The default method is "kmeans".
```{r, eval=TRUE, include=TRUE}
modulegenes_clust <- module_clust(ceRExp[, seq_len(30)],
mRExp[, seq_len(30)])
modulegenes_clust
```
## module_biclust
We Identify gene modules from matched ceRNA and mRNA expression data using a series of biclustering packages, including biclust [28], iBBiG [29], fabia [30], BicARE [31], isa2 [32], s4vd [33], BiBitR [34] and rqubic [35]. The biclustering methods include "BCBimax", "BCCC", "BCPlaid", "BCQuest", "BCSpectral", "BCXmotifs", "iBBiG", "fabia", "fabiap", "fabias", "mfsc", "nmfdiv", "nmfeu", "nmfsc", "FLOC", "isa", "BCs4vd", "BCssvd", "bibit" and "quBicluster". The default method is "fabia".
```{r, eval=TRUE, include=TRUE}
modulegenes_biclust <- module_biclust(ceRExp[, seq_len(30)],
mRExp[, seq_len(30)])
modulegenes_biclust
```
# Identification of miRNA sponge modules
The identified gene modules are regarded as candidate miRNA sponge modules. Based on the candidate miRNA sponge modules, we use the sensitivity canonical correlation (SCC), sensitivity distance correlation (SDC) and sensitivity RV coefficient (SRVC) methods to identify miRNA sponge modules. In addition, we also added the sponge module (SM) method proposed in [36] to predict miRNA sponge modules.
```{r, eval=TRUE, include=TRUE}
modulegenes_igraph <- module_igraph(ceRExp[, seq_len(10)],
mRExp[, seq_len(10)])
# Identify miRNA sponge modules using sensitivity RV coefficient (SRVC)
miRSM_igraph_SRVC <- miRSM(miRExp, ceRExp, mRExp, miRTarget,
modulegenes_igraph,
num_shared_miRNAs = 3, pvalue.cutoff = 0.05,
method = "SRVC", MC.cutoff = 0.8,
SMC.cutoff = 0.01, RV_method = "RV")
miRSM_igraph_SRVC
```
# Modular analysis of the identified miRNA sponge modules
## Functional analysis of miRNA sponge modules
We implement 'module_FA' function to conduct functional analysis of miRNA sponge modules. The functional analysis includes two types: functional enrichment analysis (FEA) and disease enrichment analysis (DEA). Functional enrichment analysis includes GO, KEGG and Reactome enrichment analysis. The ontology databases used contain GO: Gene Ontology database (), KEGG: Kyoto Encyclopedia of Genes and Genomes Pathway Database (), and Reactome: Reactome Pathway Database (). Disease enrichment analysis includes DO, DGN and NCG enrichment analysis. The disease databases used include DO: Disease Ontology database (), DGN: DisGeNET database () and NCG: Network of Cancer Genes database ().
```{r, eval=FALSE, include=TRUE}
modulegenes_WGCNA <- module_WGCNA(ceRExp[, seq_len(150)],
mRExp[, seq_len(150)])
# Identify miRNA sponge modules using sensitivity RV coefficient (SRVC)
miRSM_WGCNA_SRVC <- miRSM(miRExp, ceRExp, mRExp, miRTarget,
modulegenes_WGCNA, method = "SRVC",
SMC.cutoff = 0.01, RV_method = "RV")
miRSM_WGCNA_SRVC_genes <- miRSM_WGCNA_SRVC[[2]]
miRSM_WGCNA_SRVC_FEA <- module_FA(miRSM_WGCNA_SRVC_genes, Analysis.type = 'FEA')
miRSM_WGCNA_SRVC_DEA <- module_FA(miRSM_WGCNA_SRVC_genes, Analysis.type = 'DEA')
```
## Cancer enrichment analysis of miRNA sponge modules
To investigate whether the identified miRNA sponge modules are functionally associated with cancer of interest, we implement 'module_CEA' function to conduct cancer enrichment analysis by using a hypergeometric test.
```{r, eval=TRUE, include=TRUE}
modulegenes_WGCNA <- module_WGCNA(ceRExp[, seq_len(150)],
mRExp[, seq_len(150)])
# Identify miRNA sponge modules using sensitivity RV coefficient (SRVC)
miRSM_WGCNA_SRVC <- miRSM(miRExp, ceRExp, mRExp, miRTarget,
modulegenes_WGCNA, method = "SRVC",
SMC.cutoff = 0.01, RV_method = "RV")
miRSM_WGCNA_SRVC_genes <- miRSM_WGCNA_SRVC[[2]]
miRSM.CEA.pvalue <- module_CEA(ceRExp, mRExp, BRCA_genes, miRSM_WGCNA_SRVC_genes)
miRSM.CEA.pvalue
```
## Validation of miRNA sponge interactions in miRNA sponge modules
The function 'module_Validate' is implemented to validate the miRNA sponge interactions existed in each miRNA sponge module.
```{r, eval=FALSE, include=TRUE}
# Using the built-in groundtruth from the miRspongeR package
library(miRspongeR)
Groundtruthcsv <- system.file("extdata", "Groundtruth.csv", package="miRspongeR")
Groundtruth <- read.csv(Groundtruthcsv, header=TRUE, sep=",")
# Using the identified miRNA sponge modules based on WGCNA and sensitivity RV coefficient (SRVC)
miRSM.Validate <- module_Validate(miRSM_WGCNA_SRVC_genes, Groundtruth)
```
## Co-expression analysis of miRNA sponge modules
To evaluate whether the ceRNAs and mRNAs in the miRNA sponge modules are not randomly co-expressed, we implement 'module_Coexpress' function calculate average (mean and median) absolute Pearson correlation of all the ceRNA-mRNA pairs in each miRNA sponge module to see the overall co-expression level between the ceRNAs and mRNAs in the miRNA sponge module. For each miRNA sponge module, we perform a permutation test by generating random modules (the parameter 'resample' is the number of random modules to be generated) with the same number of ceRNAs and mRNAs for it to compute the statistical significance p-value of the co-expression level.
```{r, eval=TRUE, include=TRUE}
# Using the identified miRNA sponge modules based on WGCNA and sensitivity RV coefficient (SRVC)
miRSM_WGCNA_Coexpress <- module_Coexpress(ceRExp, mRExp, miRSM_WGCNA_SRVC_genes, resample = 10, method = "mean", test.method = "t.test")
miRSM_WGCNA_Coexpress
```
## miRNA distribution analysis of sharing miRNAs
To investigate the distribution of sharing miRNAs in the identified miRNA sponge modules, we implement 'module_miRdistribute' function. The miRNA distribution analysis can understand whether the sharing miRNAs act as crosslinks across different miRNA sponge modules.
```{r, eval=TRUE, include=TRUE}
# Using the identified miRNA sponge modules based on WGCNA and sensitivity RV coefficient (SRVC)
miRSM_WGCNA_share_miRs <- share_miRs(miRExp, ceRExp, mRExp, miRTarget, miRSM_WGCNA_SRVC_genes)
miRSM_WGCNA_miRdistribute <- module_miRdistribute(miRSM_WGCNA_share_miRs)
head(miRSM_WGCNA_miRdistribute)
```
## Predict miRNA-target interactions
Since the identified miRNA sponge modules and their sharing miRNAs can also be used to predict miRNA-target interactions (including miRNA-ceRNA and miRNA-mRNA interactions), we implement 'module_miRtarget' function to predict miRNA-target interactions underlying in each miRNA sponge module.
```{r, eval=FALSE, include=TRUE}
# Using the identified miRNA sponge modules based on WGCNA and sensitivity RV coefficient (SRVC)
miRSM_WGCNA_miRtarget <- module_miRtarget(miRSM_WGCNA_share_miRs, miRSM_WGCNA_SRVC_genes)
```
## Identify miRNA sponge interactions
To extract miRNA sponge interactions of each miRNA sponge module, we implement 'module_miRsponge' function to identify miRNA sponge interactions.
```{r, eval=FALSE, include=TRUE}
# Using the identified miRNA sponge modules based on WGCNA and sensitivity RV coefficient (SRVC)
miRSM_WGCNA_miRsponge <- module_miRsponge(ceRExp, mRExp, miRSM_WGCNA_SRVC_genes)
```
# Conclusions
miRSM provides several functions to study miRNA sponge modules, including popular methods for inferring gene modules (candidate miRNA sponge modules), and a function to identify miRNA sponge modules, as well as several functions to conduct modular analysis of miRNA sponge modules. It could provide a useful tool for the research of miRNA sponge modules.
# References
[1] Ambros V. microRNAs: tiny regulators with great potential. Cell, 2001, 107:823–6.
[2] Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell, 2004, 116:281–97.
[3] Du T, Zamore PD. Beginning to understand microRNA function. Cell Research, 2007, 17:661–3.
[4] Esquela-Kerscher A, Slack FJ. Oncomirs—microRNAs with a role in cancer.
Nature Reviews Cancer, 2006, 6:259–69.
[5] Lin S, Gregory RI. MicroRNA biogenesis pathways in cancer.
Nature Reviews Cancer, 2015, 15:321–33.
[6] Cesana M, Cacchiarelli D, Legnini I, et al. A long noncoding RNA
controls muscle differentiation by functioning as a competing endogenous
RNA. Cell, 2011, 147:358–69.
[7] Poliseno L, Salmena L, Zhang J, et al. A coding-independent function
of gene and pseudogene mRNAs regulates tumour biology. Nature, 2010,
465:1033–8.
[8] Hansen TB, Jensen TI, Clausen BH, et al. Natural RNA circles function
as efficient microRNA sponges. Nature, 2013, 495:384–8.
[9] Memczak S, Jens M, Elefsinioti A, et al. Circular RNAs are a large
class of animal RNAs with regulatory potency. Nature, 2013, 495:333–8.
[10] Salmena L, Poliseno L, Tay Y, et al. A ceRNA hypothesis: the Rosetta Stone
of a hidden RNA language? Cell, 2011, 146(3):353-8.
[11] Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics, 2008, 9:559.
[12] Bunte K, Lepp\"{a}aho E, Saarinen I, Kaski S. Sparse group factor analysis for biclustering of multiple data sources. Bioinformatics, 2016, 32(16):2457-63.
[13] Lepp\"{a}aho E, Ammad-ud-din M, Kaski S. GFA: exploratory analysis of multiple data sources with group factor analysis. J Mach Learn Res., 2017, 18(39):1-5.
[14] Csardi G, Nepusz T. The igraph software package for complex network research, InterJournal, Complex Systems, 2006:1695.
[15] Clauset A, Newman ME, Moore C. Finding community structure in very large networks. Phys Rev E Stat Nonlin Soft Matter Phys., 2004, 70(6 Pt 2):066111.
[16] Enright AJ, Van Dongen S, Ouzounis CA. An efficient algorithm for large-scale detection of protein families. Nucleic Acids Res., 2002, 30(7):1575-84.
[17] Kalinka AT, Tomancak P. linkcomm: an R package for the generation, visualization, and analysis of link communities in networks of arbitrary size and type. Bioinformatics, 2011, 27(14):2011-2.
[18] Bader GD, Hogue CW. An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics, 2003, 4:2.
[19] Zhang Y, Phillips CA, Rogers GL, Baker EJ, Chesler EJ, Langston MA. On finding bicliques in bipartite graphs: a novel algorithm and its application to the integration of diverse biological data types. BMC Bioinformatics, 2014, 15:110.
[20] Gaujoux R, Seoighe C. A flexible R package for nonnegative matrix factorization. BMC Bioinformatics, 2010, 11:367.
[21] R Core Team. R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria, 2018.
[22] Langfelder P, Horvath S. Fast R Functions for Robust Correlations and Hierarchical Clustering.
Journal of Statistical Software. 2012, 46(11):1-17.
[23] Hahsler M, Piekenbrock M. dbscan: Density Based Clustering of
Applications with Noise (DBSCAN) and Related Algorithms. R package version 1.1-2, 2018.
[24] Cebeci Z, Yildiz F, Kavlak AT, Cebeci C, Onder H. ppclust: Probabilistic and
Possibilistic Cluster Analysis. R package version 0.1.1, 2018.
[25] Scrucca L, Fop M, Murphy TB, Raftery AE. mclust 5: clustering, classification and density estimation using Gaussian finite mixture models The R Journal 8/1, 2016, pp. 205-233.
[26] Villa-Vialaneix N, Bendhaiba L, Olteanu M. SOMbrero: SOM Bound to Realize Euclidean and Relational Outputs. R package version 1.2-3, 2018.
[27] Cebeci Z, Yildiz F, Kavlak AT, Cebeci C, Onder H. ppclust: Probabilistic and Possibilistic Cluster Analysis. R package version 0.1.2, 2019.
[28] Kaiser S, Santamaria R, Khamiakova T, Sill M, Theron R, Quintales L, Leisch F, De TE. biclust: BiCluster Algorithms. R package version 1.2.0., 2015.
[29] Gusenleitner D, Howe EA, Bentink S, Quackenbush J, Culhane AC. iBBiG: iterative binary bi-clustering of gene sets. Bioinformatics, 2012, 28(19):2484-92.
[30] Hochreiter S, Bodenhofer U, Heusel M, Mayr A, Mitterecker A, Kasim A, Khamiakova T, Van Sanden S, Lin D, Talloen W, Bijnens L, G\"{o}hlmann HW, Shkedy Z, Clevert DA. FABIA: factor analysis for bicluster acquisition. Bioinformatics, 2010, 26(12):1520-7.
[31] Yang J, Wang H, Wang W, Yu, PS. An improved biclustering method for analyzing gene expression. Int J Artif Intell Tools, 2005, 14(5): 771-789.
[32] Bergmann S, Ihmels J, Barkai N. Iterative signature algorithm for the analysis of large-scale gene expression data. Phys Rev E Stat Nonlin Soft Matter Phys., 2003, 67(3 Pt 1):031902.
[33] Sill M, Kaiser S, Benner A, Kopp-Schneider A. Robust biclustering by sparse singular value decomposition incorporating stability selection. Bioinformatics, 2011, 27(15):2089-97.
[34] Rodriguez-Baena DS, Perez-Pulido AJ, Aguilar-Ruiz JS. A biclustering algorithm for extracting bit-patterns from binary datasets. Bioinformatics, 2011, 27(19):2738-45.
[35] Li G, Ma Q, Tang H, Paterson AH, Xu Y. QUBIC: a qualitative biclustering algorithm for analyses of gene expression data. Nucleic Acids Res., 2009, 37(15):e101.
[36] Zhang J, Le TD, Liu L, Li J. Identifying miRNA sponge modules using biclustering and regulatory scores.BMC Bioinformatics, 2017, 18(Suppl 3):44.
# Session information
```{r}
sessionInfo()
```