In jotsetung/mirtarbase: mirtarbase: a database of validated miRNA target gene interactions

Table of Contents

• 1. Introduction
• 2. Usage
  • 2.1. Get miRNA-target gene interactions for a specific gene
  • 2.2. Get miRNA-target gene interactions for a miRNA
  • 2.3. Get grouped miRNA-target gene interactions
  • 2.4. Alternative way to fetch data from the database
  • 2.5. Conversions between miRNA identifiers
  • 2.6. Using mirtarbase in the AnnotationDbi framework
• 3. Database layout
• 4. References

---

title: "mirtarbase: a database of validated miRNA target gene interactions" graphics: yes author: - name: Johannes Rainer output: BiocStyle::html_document: toc_depth: 2 vignette: > %\VignetteIndexEntry{mirtarbase: a database of validated miRNA target gene interactions} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} %\VignetteDepends{mirtarbase} %\VignettePackage{mirtarbase} %\VignetteKeywords{annotation,database,miRNA} bibliography: references.bib csl: biomed-central.csl references: - id: dummy title: no title author: - family: noname given: noname

---

BiocStyle::markdown() 

Introduction

The mirtarbase package provides the experimentally validated miRNA-target gene interactions (MTIs) defined in the miRTarbase database [@Hsu:2014co].

The database, which was build based on the Excel spread sheet that can be downloaded from the miRTarbase main site http://mirtarbase.mbc.nctu.edu.tw/, is automatically loaded upon library startup and bound to the environment variable MirtarbaseDb.v<release>, where <release> stands for the release of the miRTarBase. In addition, a shortcut mirtarbase is automatically generated and linked to the (most recent) release available in the package.

library(mirtarbase)

Usage

After the package is loaded the database can be accessed through the MirtarbaseDb object bound to the mirtarbase variable.

## print some information for the package
mirtarbase 

The miRTarbase defines a MTI for each mature miRNA - target gene pair and provides at least one publication in which this interaction was experimentally verified along with the support type (i.e. one of four evidence grades that was defined by the developers), the experiments supporting the interaction and the Pubmed ID. This information is stored internally in a SQLite database (see Section 3 for the layout of the database and the attributes/columns).

Before we start with some usage scenarios for the package it is important to understand the mechanism to fetch specific values from the database, i.e. how the results can be filtered. The package uses the same filtering system than the ensembldb Bioconductor package and extends it by some additional filters (filters are partially imported from the mirhostgenes package). The following filter classes are used by the package (in alphabetical order):

• EntrezFilter: filter the results based on the target genes' NCBI Entrezgene identifiers.
• ExperimentFilter: restrict the results to specific experiments (use listExperiments(mirtarbase) to get a list of all experiments).
• GenenameFilter: filter the results based on the target genes' names (symbols).
• MatMirnaFilter: search for MTIs for specific mature miRNA names (e.g. hsa-miR-15b-5p).
• MarMirnaIdFilter: search for MTIs for specific mature miRNAs identified by their mirbase accession number.
• MirfamFilter: filter the results based on miRNAs of specific miRNA families.
• MirfamIdFilter: filter the results based on miRNAs of specific miRNA families identified by their mirbase accession number.
• MirtarbaseIdFilter; filter the results based on the ID of the miRTarBase database.
• PreMirnaFilter: filter the results for MTIs of mature miRNAs derived from the specified pre-miRNA(s).
• PreMirnaidFilter: filter the results for MTIs of mature miRNAs derived from the specified pre-miRNA(s) identified by their mirbase accession number.
• PublicationFilter: filter the results based on the publication (identified by their Pubmed IDs) in which the MTI was described.
• SpeciesFilter: filter the results for target genes or miRNAs of the specified species (use listSpecies(mirtarbase, "gene") or listSpecies(mirtarbase, "mirna")).
• SupportTypeFilter: filter the results based on the evidence/support type of the MTI (use listSupportTypes(mirtarbase) to get a list of allowed values for support type).

These filters can be used individually or can be combined to generate more specific queries. Furthermore, the parameter condition allows some more flexibility to choose which entries should be fetched. condition can take the following values: =, !, contains, startsWith, endsWith with the latter 3 allowing partial matching. In addition, each filter can take a single or multiple values (see examples in the next section).

Get miRNA-target gene interactions for a specific gene

As an example we want to retrieve all MTIs for the gene BCL2, i.e. we want to get all (mature) miRNAs that have been shown to target this gene. To this end we define a GenenameFilter with the value BCL2 and submit this to the mtis call. As a result we get a MTIList, which is essentially a list of MTI objects that describe the interaction. Each interaction is defined by the mature miRNA name and the name of the target gene (accessible through the matmirna and gene methods, respectively) as well as the collection of evidences for the interaction. These evidences, or rather publications, are accessible through the reports method and specify by which experimental methods the interaction was validated and in which publication this interaction has been described. The curators of the miRTarbase manually assigned each evidence one of four support types which is accessible through the supportedBy method.

The code below simply fetches all MTIs for the gene BCL2 from the database.

## Query the database to fetch all MTIs for the target gene BCL2
BCL2 <- mtis(mirtarbase, filter=list(GenenameFilter("BCL2")))
BCL2

## To print some more information on a single MTI
BCL2[[1]]

## How many interactions did we get?
length(BCL2)

## These are however of all species as we did not specify a species filter
## and miRTarBase lists interactions for all species.
sort(table(mirnaSpecies(BCL2)), decreasing=TRUE) 

In order to restrict the MTIs to human genes and human miRNAs it is advisable to add one or more SpeciesFilter to the query.

## We can use the listSpecies method to get the names of all supported species
## from the database:
sort(listSpecies(mirtarbase))

## We want to get all human mature miRNAs that target human gene BCL2
BCL2 <- mtis(mirtarbase, filter=list(GenenameFilter("BCL2"),
                                     SpeciesFilter("Homo sapiens", feature="gene"),
                                     SpeciesFilter("Homo sapiens", feature="mirna")))

## Now we have only human miRNAs. We can now make a table of the miRNA,
## the support type and the number of publications for each MTI
BCL2.df <- data.frame(miRNA=matmirna(BCL2),
              reports=reportCount(BCL2),
              support_type=unlist(lapply(supportedBy(BCL2), function(z){
                          return(paste(unique(z), collapse=";"))
              })))

## Display the MTIs described by the most publications
head(BCL2.df[order(BCL2.df$reports, decreasing=TRUE), ]) 

So, there is evidence that e.g. miR-16-5p is targeting the gene BCL2, along with miR-15a-5p. We can also enrich this table with the information of the pre-miRNA(s) in which the mature miRNA is encoded. In addition, we can group the miRNAs also by the miRNA family. Note that each mature miRNA can be eventually encoded in more than one pre-miRNA, each mature miRNA (and each pre-miRNA) is supposed to be part of one miRNA family.

BCL2.df <- cbind(BCL2.df,
                 premirna=unlist(lapply(BCL2, function(z){
                     return(paste(premirna(z), collapse=";"))
                 })),
                 mirfam=mirfam(BCL2))

## Note: there are some mature miRNAs that can not be mapped to pre-miRNA
## or mirfam names.
sum(is.na(as.character(BCL2.df$mirfam)))

## the miRNA with most evidences (miR-16-5p) is actually encoded in two
## precursors:
premirna(BCL2$MIRT001800)

## The miRNA families from which most miRNAs target BCL2 are listed below:
sort(table(as.character(BCL2.df$mirfam)), decreasing=TRUE)

## The miRNAs from the mir-15 family targeting BCL2 are
MTI.mir15 <- BCL2[ which(unlist(lapply(BCL2, mirfam))=="mir-15") ]
## the mature miRNAs from this family:
MTI.mir15

## Extract the mature miRNA IDs
matmirna(MTI.mir15)
## And the pre-miRNAs:
premirna(MTI.mir15)

The missing mapping of mature miRNAs to pre-miRNA names or mirfam identifiers observed above is in many instances caused by different mirbase versions on which the mirbase.db package and the miRTarbase bases. In addition, not all mature miRNAs are annotated to miRNA families.

As we have seen above, we can use the methods matmirna, premirna and mirfam on MTI or MTIList objects to retrieve the mature miRNA involved in the miRNA-target gene interaction, the pre-miRNA in which the mature miRNA is encoded and the miRNA family to which the pre-miRNA(s) belong.

Get miRNA-target gene interactions for a miRNA

Next we retrieve MTIs between miRNAs of the mir-15 family and genes which names start with BCL2. For this we define a GenenameFilter with "like" as condition and a pattern for the gene name.

## Get all miRNA-target gene interactions betwee mature miRNAs from the
## mir-15 family and genes starting with BCL2
BCLs <- mtis(mirtarbase, filter=list(MirfamFilter("mir-15"),
                                     GenenameFilter("BCL2", condition="startsWith"),
                                     SpeciesFilter("Homo sapiens"))
            )
BCLs 

According to this information the miRNA miR-195-5p targets both, a pro- and an anti-apoptotic member of the BCL2 gene family (BCL2L11 and BCL2, respectively).

By default, the results are returned by the mtis method as MTIList object, but we could also specify "data.frame" as the return.type to retrieve the data as data.frame. This allows to retrieve only specific information from the database by specifying the columns that should be returned.

onlyGeneNames <- mtis(mirtarbase, filter=list(MirfamFilter("mir-15"),
                          GenenameFilter("BCL2", condition="startsWith"),
                          SpeciesFilter("Homo sapiens")),
              columns=c("mirna", "target_gene"), return.type="data.frame")
head(onlyGeneNames) 

Also members of the mir-17 family have been reported to target genes from the BCL2 gene family [@Ventura:2008gk], thus we retrieve next all MTIs between miRNAs of the miRNA families mir-15 or mir-17 and some of the genes from the BCL2 gene family, a gene family involved in, and regulating, the intrinsic apoptotic pathway.

To retrieve values for more than one gene, respectively miRNA family, we can submit a character vector of the respective ids to the filters.

## retrieving all MTIs between miRNAs from the mir-15 and mir-17 families
## and some genes from the BCL2 gene family
BCLs <- mtis(mirtarbase,
             filter=list(MirfamFilter(c("mir-15", "mir-17")),
                 GenenameFilter(c("BCL2", "BCL2L11", "PMAIP1", "MCL1")),
                 SpeciesFilter("Homo sapiens"))
            )
BCLs
## the miRNA - gene pairs:
data.frame(miRNA=matmirna(BCLs),
           gene=gene(BCLs),
           report_count=reportCount(BCLs)) 

Apparently, miRNAs from both the miR-15 and the miR-17 family target genes of the BCL2 gene family and are thus also involved in the regulation of the apoptotic pathway.

Next we evaluate the evidence grades of the interaction and remove all MTIs that are not of the Functional MTI support type (the type with the highest evidence grade).

funcMti <- unlist(lapply(BCLs, function(z){
    return(any(supportedBy(z)=="Functional MTI"))
}))
sum(funcMti)
length(funcMti)

## We could now use this logical vector to sub-set the list.
## Alternatively, we can also re-perform the query and fetch only interactions of that
## support type, which has the advantage that also only the publications of the
## corresponding support type are loaded.
BCLs <- mtis(mirtarbase,
               filter=list(MirfamFilter(c("mir-15", "mir-17")),
                   GenenameFilter(c("BCL2", "BCL2L11", "PMAIP1", "MCL1")),
                   SpeciesFilter("Homo sapiens"),
                   SupportTypeFilter("Functional MTI"))
            )
## the miRNA - gene pairs:
data.frame(miRNA=matmirna(BCLs),
           gene=gene(BCLs),
           report_count=reportCount(BCLs)
          ) 

This considerably reduced the list of interactions and also decreased the number of reports per MTI.

Get grouped miRNA-target gene interactions

Sometimes it might be useful to group the miRNA-target gene interactions by some factor, e.g. by genes or miRNAs. The method mtisBy allows to fetch MTIs grouped by any column from the database. It is possible to group the results by gene, (mature miRNA), entrezid, support type, Pubmed ID, pre-miRNA name, miRFam name or by species. The result will be a list with the names being the factor by which the interactions are grouped and each element being a MTIList of the MTIs.

In the example below we fetch all MTIs for the genes BCL2, BCL2L11, MCL1 and group them by miRNA family.

Filters <- list(SpeciesFilter(c("Homo sapiens")),
                GenenameFilter(c("BCL2", "BCL2L11", "MCL1")))

BCL2by <- mtisBy(mirtarbase, filter=Filters, by="mirfam")
head(BCL2by) 

In a similar way we can also fetch the data grouped by gene.

BCL2by <- mtisBy(mirtarbase, filter=Filters, by="gene")
BCL2by

Alternative way to fetch data from the database

By default, the mtis method returns a list of MTI objects (MTIList) which is sufficient for most use cases. Alternatively, however, the mtis method can also return the results as a data.frame. In addition to a significant performance improvement this also enables to select only specific columns from the database. Note however that by default the method returns all columns from the database which results in a data.frame with one MTI-publication per row, i.e. the same MTI represented by the miRNA-gene pair can be present in many rows of this data.frame depending in how many publications this interaction was identified.

## We perform the same call as above, but restrict the information to some selected
## columns and specify to return the results as a data.frame rather than a list
## of MTI objects.
BCLs.df <- mtis(mirtarbase,
                filter=list(MirfamFilter(c("mir-15", "mir-17")),
                    GenenameFilter(c("BCL2", "BCL2L11", "PMAIP1", "MCL1")),
                    SpeciesFilter("Homo sapiens"),
                    SupportTypeFilter("Functional MTI")),
                columns=c("mirna", "target_gene"),
                return.type="data.frame")

BCLs.df 

Conversions between miRNA identifiers

The mirtarbase package provides also methods and functions that allow to map mature miRNAs to their precursors or to miRNA families. These functions are essentially wrapper functions that use the information of the mirbase.db Bioconductor package for the conversion. However, since the mirtarbase and mirbase.db functions might provide information from different releases, some of the mappings might not be available. For a complete list of conversion function refer to the help page of the e.g. premirna2matmirna function.

## map from pre-miRNA name to mature miRNA name. The function returns by default
## a data.frame
premirna2matmirna(c("hsa-mir-16-1", "hsa-mir-16-2"))

## the same information but as a list:
premirna2matmirna(c("hsa-mir-16-1", "hsa-mir-16-2"), return.type="list") 

Using mirtarbase in the AnnotationDbi framework

The mirtarbase package implements also methods keys, keytypes, columns and select from the AnnotationDbi package that allow to query data from a MirtarbaseDb analogously to other AnnotationDbi objects. The supported columns by these methods are:

## List all supported columns that can be queried.
columns(mirtarbase)

## Note that these column names are different to those supported
## by the mtis method:
listColumns(mirtarbase) 

We can use the keys and keytypes methods to retrieve the supported keytypes for the select method.

## List all supported keytypes
keytypes(mirtarbase)

## List all keys for "Support Type"
keys(mirtarbase, keytype="SUPPORTTYPE")

## Use select to retrieve all MTIs for support type "Functional MTI"
mtis <- select(mirtarbase, keys="Functional MTI", keytype="SUPPORTTYPE")
head(mtis)
nrow(mtis) 

The select method for MirtarbaseDb allows in addition also to submit one or more filter objects with argument keys. This enables more flexible queries than possible with the standard usage. Below we retrieve all MTIs of support type Functional MTI for genes BCL2 and BCL2L11.

mtis <- select(mirtarbase, keys=list(SupportTypeFilter("Functional MTI"),
                                     GenenameFilter(c("BCL2", "BCL2L11"))))
head(mtis)
nrow(mtis) 

Database layout

The database consists of 3 tables, mirtarbase which contains all information stored in the xls file from the miRTarBase web site, pubmed_corpus, that contains the content of the MTI-PubMed_corpus.txt file from the miRTarBase web site and metadata with some internal informations. The column names and their properties are listed below. Each line in the table represents the MTI for a miRNA and one of its target genes as reported in a publication. Thus, an interaction between a miRNA and its target gene can be listed in more than one row, depending on the number of publications it was validated.

• mirtarbase:

  • mirtarbase_id: identifier for the miRNA target gene interaction (MTI). Note that this ID is not unique, i.e. MTIs reported in several publications have the same ID but are listed in several rows of the table.
  • mirna: mature miRNA name (a.k.a miRNA ID, e.g. hsa-miR-20a-5p).
  • species_mirna: the species of the miRNA (e.g. Homo sapiens).
  • target_gene: the official gene name (symbol) for the gene (e.g. DUSP6, or ush).
  • target_gene_entrez_gene_id: the NCBI Entrezgene ID for the target gene; either NA or the (numerical) Entrezgene ID. Contains only unique values, no multiple IDs collapsed by any separator.
  • species_target_gene: the species of the target gene.
  • experiments: the experiments providing the evidence for the interaction as reported in one publication.
  • support_type: the different types of support (from weak to strong).
  • references_pmid: the Pubmed ID of the publication reporting the MTI. Each line with a single Pubmed ID, no empty (NA) values.

• pubmed_corpus

  • pmid: the PubMed ID of the paper describing the MTI.
  • key: the key of the entry (can be Abstract, Title, Experiment_method, miRNA or Target_gene).
  • value: the value for the key.

• metadata

  • name: the names of the keys.
  • value: the value for the key.

References

jotsetung/mirtarbase documentation built on May 19, 2019, 9:42 p.m.