Version 1.4 of epivizr was released as part of the new Bioconductor 3.0 release. There are a good number of new features in epivizr in this release, including the ability of running the Epiviz UI within R, making R/Bioconductor capable of running it’s own genome browser!
epiviz as a standalonePrevious versions of epivizr used the {{ site.epiviz }} web application hosted at the University of Maryland as the front-end for UI. In this new version, we bundle the JavaScript source for the {{ site.epiviz }} UI in the epivizr package which allows R to serve as it’s own web host. Like Shiny, we use the httpuv package to serve the interactive application. In fact, lots of thanks to Joe Cheng and RStudio for helping to make this happen!
A genome browser without a genome is not very useful, so you need to tell it about the genome annotation you want to use. One place in Bioconductor where you can get genome annotation information is in their OrganismDb packages.
(Make sure you have version 1.4.2 installed to follow along! Just run biocLite("epivizr") to get it). The startStandalone function can take an object of this class and start the UI with that genome and its gene annotation loaded. Here’s how you can browse the mouse genome with Bioconductor:
library(epivizr)
library(Mus.musculus)
# this call makes the gene annotation from Mus.musculus,
# takes a couple of seconds (see more info below)
mgr <- startStandalone(Mus.musculus, "mm10", keepSeqlevels=paste0("chr",c(1:19,"X","Y")))
mgr$stopServer()
This release also allows to add genome and gene information on any Epiviz UI, both standalone and hosted (e.g., at the University of Maryland). For instance, to add the mouse genome to the Epiviz UI hosted at UMD you can use:
mgr <- startEpiviz(workspace="OJS2BPGrh7v")
# remove the pre-loaded hg19 annotation
mgr$rmSeqinfo(paste0("chr",c(1:22,"X","Y")))
# make a genome annotation object (takes a few seconds)
anno <- epivizr::makeGeneTrackAnnotation(Mus.musculus, keepSeqlevels=paste0("chr", c(1:19,"X","Y")))
# add chromosome names and lengths
mgr$addSeqinfo(seqinfo(anno))
# check on UI, only 19 autosomes now
# add the gene annotation track
annoDevice <- mgr$addDevice(anno, "mm10", type="geneInfo")
# now you have a mouse genome browser!
mgr$stopServer()
You can now add heatmaps to your Epiviz sessions. Here is an example using exon-level RNA-seq count data from the TCGA project (this data is included in the epivizr package).
library(epivizr)
data(tcga_colon_expression)
# make the names easier to understand
sampleNames <- paste0(seq(len=nrow(colData(colonSE))),":",colData(colonSE)$sample_type)
colnames(colonSE) <- sampleNames
# normalize counts using DE-Seq's size factors
ref_sample <- 2 ^ rowMeans(log2(assay(colonSE) + 1))
scaled <- (assay(colonSE) + 1) / ref_sample
scaleFactor <- Biobase::rowMedians(t(scaled))
assay_normalized <- sweep(assay(colonSE), 2, scaleFactor, "/")
assay(colonSE) <- log2(assay_normalized + 1)
# start the UI manager
mgr <- startEpiviz(workspace="qyOTB6vVnff")
# add the count data
msObj <- mgr$addMeasurements(colonSE, "tcga exon expression")
# make a heatmap of the first 15 samples
devObj <- mgr$heatmapChart(msObj$getMeasurements()[1:15])
mgr$stopServer()
And since Epiviz now supports clustering on heatmaps, you have a dynamic heatmap visualization of your expression data as you browse the genome.
For the next release of epivizr we are planning some new fun things:
BigWig and BAM files to support NGS workflows.We’ve started working on #1 above, you can see how it’s going by using the epivizr development version.