LErNet: characterization of lncRNAs via context-aware network expansion and enrichment analysis
the old verion can be found here
LErNet is distributed under the MIT license. This means that it is free for both academic and commercial use. Note however that some third party components in LErNet require that you reference certain works in scientific publications. You are free to link or use LErNet inside source code of your own program. If do so, please reference (cite) LErNet and this website. We appreciate bug fixes and would be happy to collaborate for improvements. License
If you have used the package LErNet in your project, please cite the following paper:
Bonnici V., Caligola S., Fiorini G., Giudice L., Giugno R.
LErNet: characterization of lncRNAs via context-aware network expansion and enrichment analysis.
In 2019 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB) (pp. 1-8). IEEE.
Before to install the LErNet package make sure that all the required packages are already installed on your computer.
# if you have not installed "devtools" package
install.packages("BiocManager")
BiocManager::install("GenomicRanges")
BiocManager::install("GenomicFeatures")
BiocManager::install("STRINGdb")
BiocManager::install("biomaRt")
BiocManager::install("ReactomePA")
install.packages("visNetwork")
install.packages("igraph")
install.packages("R.utils")
install.packages("rmarkdown")Then, you can install the LErNet package.
library(devtools)
install_github("InfOmics/LErNet")or
library(devtools)
install_github("InfOmics/LErNet", ref="lernet.1.0")We report a step-by-step example to execute LErNet on published data (Zhao et al., Scientific reports, 2018). The dataset is composed by a list of differentially expressed genes and long non-coding RNA (lncRNA). Original excel files are provided with the LErNet package in order to correctly execute the analysis. Further, a GTF file (from GENCODE database) is provided to retrieve genomic context of genes and lncRNAs.
It's necessary to install and load the following libraries to run the example:
library(R.utils)
library(biomaRt)
The first step of the analysis is to retrieve a set of genes and lncRNAs of interest and the information of the genomic context. In the following lines of code DE genes and DE lncRNAs are obtained directly from the excel files provided within the LErNet package and loaded after several preprocess operations.
lncrna_file <- system.file("extdata", "41598_2018_30359_MOESM2_ESM.csv", package = "LErNet")
pcrna_file <- system.file("extdata", "41598_2018_30359_MOESM3_ESM.csv", package = "LErNet")
gtf_file <- system.file("extdata", "gencode.vM20.chr_patch_hapl_scaff.annotation.gtf.gz", package = "LErNet")
pcgenes<-(read.csv(pcrna_file, stringsAsFactors=FALSE))$gene_id
length(pcgenes)
lncrnaInfo<-read.csv(lncrna_file, stringsAsFactors=FALSE)
lncrnaInfo<-lncrnaInfo[lncrnaInfo$significant != 'FALSE', ]
lncrnaAll<-as.character(lncrnaInfo$gene_id)
nrow(lncrnaInfo)
length(lncrnaAll)
LErNet provides the function load_gtf to load into a dataframe the necessary information from a GTF file.
complete_positions <- LErNet::load_gtf(gtf_file)
nrow(complete_positions)It is necessary that the dataframe contains the information for all genes and lncRNAs in input. In this example, data comes with information about novel lncRNAs. These information must be added to the dataframe complete_positions.
# Extract the novel lncRNAs from the dataframe "lncrnaInfo"
novel<-lncrnaInfo[lncrnaInfo$isoform_status == "lncRNA_Novel", ]
nrow(novel)
# Some elaboration to extract the necessary information about lncRNAs
chrs <- paste0("chr",sapply(strsplit(sapply(strsplit( novel$locus, "-"), `[`, 1), ":"), `[`, 1))
starts <- sapply(strsplit(sapply(strsplit(novel$locus, "-"), `[`, 1), ":"), `[`, 2)
ends <- sapply(strsplit(novel$locus, "-"), `[`, 2)
novel_gtf <- data.frame( "id" = novel$gene_id, "type" = rep('novel lncRNA', times = nrow(novel)),
"seqname" = chrs, "start" = starts, "end" = ends )
nrow(novel_gtf)
# Add information of novel lncRNAs into the dataframe containing information about known genes/lncRNAs
complete_positions <- rbind(complete_positions, novel_gtf)
rownames(complete_positions) <- seq(1:nrow(complete_positions))
nrow(complete_positions)LErNet exploits PPI network to expand a set of protein coding genes associated with lncRNAs. In this example the database STRING is exploited to build the PPI network, however LErNet can take as input a dataframe with 2 columns containing the edges of the network. Each element of the network must be identified with its ENSEMBL id. To build the network with STRING is necessary to specify a threshold of significance for protein interactions and the taxonomy id of the organism of interest.
mart <- useMart(biomart = "ensembl", dataset = "mmusculus_gene_ensembl")
#WARNING: on R <= 3.4 this may cause mutiple errors. Please, run it until no errors are arised.
# or, artenatively, use
#mart <- useEnsembl(biomart = "ensembl", dataset = "mmusculus_gene_ensembl", mirror = "useast")
stringdb_tax = 10090
stringdb_thr = 900
# alteratively, for human
# mart <- useMart(biomart = "ensembl", dataset = "hsapiens_gene_ensembl")
# #WARNING: on R <= 3.4 this may cause mutiple errors. Please, run it until no errors are arised.
# stringdb_tax = 9606
The function get_stringdb returns a list with a dataframe named ppi containing the PPI network. This dataframe can be provided to LErNEt without the use of STRING.
library(STRINGdb)
library(igraph)
ppi_network <- LErNet::get_stringdb( stringdb_tax = stringdb_tax, stringdb_thr = stringdb_thr)
annot<-getBM(attributes = c("ensembl_gene_id", "ensembl_gene_id_version", "ensembl_transcript_id", "ensembl_peptide_id",
"hgnc_symbol", 'transcript_biotype', 'transcript_length', 'entrezgene_id', "mgi_symbol"), mart = mart)
# write.csv(annot, system.file("extdata", "mouse_mart_export.csv", package = "LErNet"), row.names = FALSE)
# annot <- read.csv(system.file("extdata", "mouse_mart_export.csv", package = "LErNet"), sep=',', stringsAsFactors = FALSE)
ensp_to_ensg <- subset(annot, ensembl_peptide_id %in% unique(union(ppi_network$protein1, ppi_network$protein2)) )[, c('ensembl_peptide_id','ensembl_gene_id')]
print(nrow(ppi_network))
print(nrow(ensp_to_ensg))This step is used to generate the genomic context, i.e. to find the genomic seeds necessary to run the expansion phase through th PPI network. The function to perform accomplish this task is get_genomic_context. The function takes in input the information retrieved from the GTF file (complete_positions), the list of protein coding genes and lncRNAs and a window in which to search for genomic neighbors. The function returns a dataframe containing for each lncRNA one or more partner coding genes.
library(GenomicRanges)
genomic_context <- LErNet::get_genomic_context(
positions = complete_positions,
lncgenes = lncrnaAll,
pcgenes = pcgenes,
max_window = 100000)
nrow(genomic_context)It can be useful to show some basic statistics on the generated seeds:
# Number of genomic seeds
length(unique(genomic_context$partner_coding_gene))
# Mean number of genomic seeds for each lncRNA
mean(table(genomic_context$lnc_known))
The following lines of codes are necessary to match the seeds
length(unique(genomic_context$partner_coding_gene))
genomic_seeds <- unique((merge(genomic_context, ensp_to_ensg, by.x='partner_coding_gene', by.y='ensembl_gene_id'))$ensembl_peptide_id)
length(genomic_seeds)
length(pcgenes)
de_proteins <- merge(annot[annot$ensembl_peptide_id != '', ], data.frame('ensembl_gene_id' = unique(pcgenes)))$ensembl_peptide_id
length(de_proteins)The next step is the core phase of LErNEet, i.e. the expansion phase with the function expand_seeds. Expansion takes as input the genomic context, the PPI network and the list of starting proteins.
The function expand_seeds returns a list containing a dataframe with the network components.
components <- LErNet::expand_seeds(genomic_seeds, ppi_network, de_proteins=NULL)
length(unlist(components))
length(components[[1]])
components[[1]]This step is necessary to retrive the list of the protein mates which interact with lncRNAs of interest.
lncrna_context <- unique( (merge(genomic_context, annot[annot$ensembl_peptide_id != '',], by.x='partner_coding_gene', by.y='ensembl_gene_id'))[ , c('lnc_known','ensembl_peptide_id')])
colnames(lncrna_context) <- c('lncrna','mate')
To go through with the next step and then to facilitate the reading of the network, we need to compute the labels.
labels <- data.frame(
'id' = unique(genomic_context$lnc_known),
'label' = unique(genomic_context$lnc_known)
)
x <- (merge(data.frame('ensembl_peptide_id' = as.character(unique(unlist(components)))), annot))[, c('ensembl_peptide_id','mgi_symbol')]
colnames(x) <- c('id','label')
labels <- rbind(labels, x)LErNet allows to visualize the results of the analysis through the use of the package visNetwork. The function visualize takes in input the lncrna context, the list of strict starting proteins, the network seeds, the PPI network, one or more network components extracted by LErNet and the labels.
LErNet::visualize(
lncrna_context,
de_proteins,
genomic_seeds,
unlist(components),
ppi_network,
labels)
The last step is the functional enrichment of the results. Basically LErNet exploits the package ReactomePA to retrieve significant pathways through the function enrich:
library(ReactomePA)
#BiocManager::install("org.Mm.eg.db")
library(org.Mm.eg.db)
# for human
# BiocManager::install("org.Hs.eg.db")
# library(org.Hs.eg.db)
comp <- components[[1]]
entrez_ids <- unique((merge( data.frame('ensembl_peptide_id' = comp), annot ))$entrezgene_id)
enrichment <- LErNet::enrich(entrez_ids, 'mouse')
#print(enrichment)
# for human: organism = "human"
barplot(enrichment)
However, the user can use the preferred tool to make functional enrichment.