Introduction to epivizr: interactive visualization for genomic data

Héctor Corrada Bravo

2016-05-15

Epiviz is an interactive visualization tool for functional genomics data. It supports genome navigation like other genome browsers, but allows multiple visualizations of data within genomic regions using scatterplots, heatmaps and other user-supplied visualizations. It also includes data from the Gene Expression Barcode project for transcriptome visualization. It has a flexible plugin framework so users can add d3 visualizations. You can find more information about Epiviz at http://epiviz.cbcb.umd.edu/help and see a video tour here.

The epivizr package implements two-way communication between the R/Bioconductor computational genomics environment and Epiviz. Objects in an R session can be displayed as tracks or plots on Epiviz. Epivizr uses Websockets for communication between the browser Javascript client and the R environment, the same technology underlying the popular Shiny system for authoring interactive web-based reports in R.

Preliminaries: the data

In this vignette we will look at colon cancer methylation data from the TCGA project and expression data from the gene expression barcode project. The epivizrData package contains human chromosome 11 methylation data from the Illumina 450kHumanMethylation beadarray processed with the minfi package. We use expression data from the antiProfilesData bioconductor package.

library(epivizr)
library(antiProfilesData)
library(SummarizedExperiment)
data(tcga_colon_blocks)
data(tcga_colon_curves)
data(apColonData)

The tcga_colon_blocks object is a GRanges object containing chromosome 11 regions of hypo or hyper methylation in colon cancer identified using the blockFinder function in the minfi package.

show(tcga_colon_blocks)
## GRanges object with 129 ranges and 11 metadata columns:
##         seqnames                 ranges strand |               value
##            <Rle>              <IRanges>  <Rle> |           <numeric>
##     [1]    chr11 [  4407026,   6435089]      * |  -0.142954896408814
##     [2]    chr11 [131239366, 133716186]      * |  -0.135137323912239
##     [3]    chr11 [ 55041873,  57022542]      * |   -0.17334711677526
##     [4]    chr11 [114645223, 116602403]      * |  -0.140934365473709
##     [5]    chr11 [ 78357700,  80184550]      * |   -0.15603691753532
##     ...      ...                    ...    ... .                 ...
##   [125]    chr11   [29644815, 29650449]      * | -0.0940450729648221
##   [126]    chr11   [41943963, 41956273]      * | -0.0760193132782017
##   [127]    chr11   [16298618, 16314417]      * | -0.0748443876820716
##   [128]    chr11   [38740154, 38740154]      * |  -0.117248761155248
##   [129]    chr11   [81757203, 81757203]      * | -0.0710557720226853
##                       area   cluster indexStart  indexEnd         L
##                  <numeric> <numeric>  <integer> <integer> <numeric>
##     [1]   30.3064380386685       495     130755    131000       212
##     [2]   23.9193063324663       520     141959    142173       177
##     [3]   19.5882241956043       507     134251    134374       113
##     [4]   14.0934365473709       520     140000    140138       100
##     [5]   14.0433225781788       507     138132    138249        90
##     ...                ...       ...        ...       ...       ...
##   [125]  0.188090145929644       497     132733    132734         2
##   [126]  0.152038626556403       502     133379    133380         2
##   [127]  0.149688775364143       495     131986    131987         2
##   [128]  0.117248761155248       499     133331    133331         1
##   [129] 0.0710557720226853       508     138263    138263         1
##          clusterL            p.value      fwer       p.valueArea  fwerArea
##         <integer>          <numeric> <numeric>         <numeric> <numeric>
##     [1]      1629                  0         0                 0         0
##     [2]      1759                  0         0                 0         0
##     [3]      1816                  0         0                 0         0
##     [4]      1759                  0         0                 0         0
##     [5]      1816                  0         0                 0         0
##     ...       ...                ...       ...               ...       ...
##   [125]       367 0.0949029198604193         1 0.416952488890214         1
##   [126]        45  0.316487220018491         1 0.485758597035402         1
##   [127]      1629  0.342091920427093         1 0.494228876494974         1
##   [128]         7 0.0564138506964121         1 0.592024814339825         1
##   [129]        22  0.729905454979272         1 0.860076351814847         1
##   -------
##   seqinfo: 23 sequences from an unspecified genome; no seqlengths

The columns value and p.value can be used to determine which of these regions, or blocks, are interesting by looking at a volcano plot for instance.

plot(tcga_colon_blocks$value, -log10(tcga_colon_blocks$p.value), main="Volcano plot", xlab="Avg. methylation difference", ylab="-log10 p-value",xlim=c(-.5,.5))

The tcga_colon_curves object is another GRanges object which contains the basepair resolution methylation data used to define these regions.

show(tcga_colon_curves)
## GRanges object with 7135 ranges and 7 metadata columns:
##          seqnames                 ranges strand |        id    type
##             <Rle>              <IRanges>  <Rle> | <numeric> <array>
##      [1]    chr11       [131996, 132411]      * |    129466 OpenSea
##      [2]    chr11       [189654, 189654]      * |    129467 OpenSea
##      [3]    chr11       [190242, 190242]      * |    129468 OpenSea
##      [4]    chr11       [192096, 192141]      * |    129469 OpenSea
##      [5]    chr11       [192763, 193112]      * |    129470 OpenSea
##      ...      ...                    ...    ... .       ...     ...
##   [7131]    chr11 [134892703, 134892703]      * |    142360 OpenSea
##   [7132]    chr11 [134903175, 134903175]      * |    142361 OpenSea
##   [7133]    chr11 [134910774, 134910774]      * |    142362 OpenSea
##   [7134]    chr11 [134911302, 134911302]      * |    142363 OpenSea
##   [7135]    chr11 [134945848, 134945848]      * |    142364 OpenSea
##          blockgroup         diff      smooth normalMean cancerMean
##           <numeric>     <matrix>   <numeric>  <numeric>  <numeric>
##      [1]        495 -0.088871013 -0.12366708 0.75449261 0.66562159
##      [2]        495 -0.001244860 -0.08656241 0.95952884 0.95828398
##      [3]        495 -0.164759184 -0.08625123 0.68006830 0.51530911
##      [4]        495  0.003094723 -0.08526643 0.05134196 0.05443669
##      [5]        495 -0.102463563 -0.08484055 0.55235839 0.44989482
##      ...        ...          ...         ...        ...        ...
##   [7131]        520  -0.11394759  -0.1421404  0.4013819  0.2874343
##   [7132]        520  -0.01148901  -0.1530226  0.8482137  0.8367247
##   [7133]        520  -0.20066455  -0.1627672  0.6365234  0.4358588
##   [7134]        520  -0.14013791  -0.1635050  0.5831618  0.4430239
##   [7135]        520  -0.24670370  -0.2308602  0.6757470  0.4290433
##   -------
##   seqinfo: 24 sequences from an unspecified genome; no seqlengths

This basepair resolution data includes mean methylation levels for normal and cancer and a smoothed estimate of methylation difference. This smoothed difference estimate is used to define regions in the tcga_colon_blocks object.

Finally, the apColonData object is an ExpressionSet containing gene expression data for colon normal and tumor samples for genes within regions of methylation loss identified this paper. Our goal in this vignette is to visualize this data as we browse the genome.

The epivizr app

The connection to Epiviz is managed through a session manager object of class EpivizApp. We can create this object and open Epiviz using the startEpiviz function.

app <- startEpiviz(workspace="qyOTB6vVnff", gists="2caf8e891201130c7daa")

This opens a websocket connection between the interactive R session and the browser client. This will allow us to visualize data stored in the Epiviz server along with data in the interactive R session.


Windows users: In Windows platforms we need to use the service function to let the interactive R session connect to the epiviz web app and serve data requests. We then escape (using ctl-c or esc depending on your environment) to continue with the interactive R session. This is required anytime you want epivizr to serve data to the web app, for example, when interacting with the UI. (We are actively developing support for non-blocking sessions in Windows platforms).

app$server$service()

Listing available chart types

Once the browser is open and is connected to an active epivizr session, the epivizr session registers available chart types supported in the epiviz JS app. To list available chart types we use method list_chart_types. In this vignette, we only show three chart types available for use in the epiviz JS app. The production epiviz JS app has a larger number of chart types available. Chart types can also be added dynamically as described in the epiviz JS documentation: http://epiviz.cbcb.umd.edu/help. Chart types that are added dynamically are listed and usable within an epivizr session.

app$chart_mgr$list_chart_types()
##          type                          js_class num_settings
## 1 BlocksTrack epiviz.plugins.charts.BlocksTrack            6
## 2   LineTrack   epiviz.plugins.charts.LineTrack           15
## 3 ScatterPlot epiviz.plugins.charts.ScatterPlot           11
##                                                                              settings
## 1                title,marginTop,marginBottom,marginLeft,marginRight,minBlockDistance
## 2 title,marginTop,marginBottom,marginLeft,marginRight,measurementGroupsAggregator,...
## 3 title,marginTop,marginBottom,marginLeft,marginRight,measurementGroupsAggregator,...
##   num_colors
## 1         10
## 2          8
## 3         10

Chart type settings, e.g., line widths, colors, margins etc. can be modified dynamically. Settings available for each chart type are briefly listed in this call. See below for further detail on customizing charts through settings and colors.

Adding block region devices

Once the browser is open we can visualize the tcga_colon_blocks object containing blocks of methylation modifications in colon cancer. We use the plot method to do so.

blocks_chart <- app$plot(tcga_colon_blocks, datasource_name="450k colon_blocks")

Windows users: We need the service function to let the interactive R session serve data requests from the browser client as you interact with the UI. Escape (using ctl-c or esc depending on your environment) to continue with the interactive R session.

app$server$service()

You should now see that a new track is added to the Epiviz web app. You can think of this track as an interactive display device in R. As you navigate on the browser, data is requested from the R session through the websocket connection. Remember to escape to continue with your interactive R session if you are not running “non-blocking” mode. The blocks_chart object inherits from class EpivizChart, which we can use to get information about the data being displayed and the chart used to display it. Note that the “brushing” effect we implement in Epiviz works for epivizr tracks as well.

Now that we have interactive data connections to Epiviz we may want to iteratively compute and visualize our data. For example, we may want to only display methylation blocks inferred at a certain statistical significance level. In this case, we will filter by block size.

# subset to those with length > 250Kbp
keep <- width(tcga_colon_blocks) > 250000

# get the data object for chart
ms_obj <- app$get_ms_object(blocks_chart)
app$update_measurements(ms_obj, tcga_colon_blocks[keep,])

Now, only this subset of blocks will be displayed in the already existing track.

Modifying chart settings and colors

To modify default chart colors or settings, we need to know what settings can be applied to the chart. Again, as these are defined dynamically in the epiviz JS app we use a method, print_chart_type_info on the EpivizChartMgr class to list settings that can be applied to specific chart type. For example, to list chart settings available for a BlocksTrack chart we use the following:

app$chart_mgr$print_chart_type_info("BlocksTrack")
## Settings for chart type  BlocksTrack 
##                 id                  label default_value possible_values
## 1            title                  Title                              
## 2        marginTop             Top margin            25                
## 3     marginBottom          Bottom margin            23                
## 4       marginLeft            Left margin            20                
## 5      marginRight           Right margin            10                
## 6 minBlockDistance Minimum block distance             5                
##     type
## 1 string
## 2 number
## 3 number
## 4 number
## 5 number
## 6 number
## Colors: #1f77b4, #ff7f0e, #2ca02c, #d62728, #9467bd, #8c564b, #e377c2, #7f7f7f, #bcbd22, #17becf

There are two ways to set settings and colors to a chart: when the chart is initially created using the plot method, or after the chart is created using the set method in class EpivizChart. We will illustrate both ways of doing this:

If using the plot method, use the list_chart_settings method as above to list settings that can be applied to a chart type. For example, for a BlocksTrack chart, we can set minBlockDistance as a setting for the plot function which controls how close genomic regions can be before they are merged into a single rectangle in the chart. We can also change colors used in the chart this way.

settings <- list(minBlockDistance=50)
colors <- c("#d15014", "#5e97eb", "#e81ccd")
blocks_chart <- app$plot(tcga_colon_blocks, datasource_name="450k colon_blocks", settings=settings, colors=colors)

This will create a second blocks track using a different color map.

On the other hand, if the BlockChart is already added to the epiviz JS application session, use the set_chart_settings method to update settings and colors.

# create a list of settings and colors to update the plot
settings <- list(minBlockDistance=100)
colors <- c("#5e97eb", "#d15014", "#e81ccd")
app$chart_mgr$set_chart_settings(blocks_chart, settings=settings, colors=colors)

This changes the color map for the second blocks track added. Use the print_chart_info method to list settings and colors currently used in the chart.

# to list applied chart settings
app$chart_mgr$print_chart_info(blocks_chart)
## Chart settings for chart id  epivizChart_2 :
##                 id                  label default_value possible_values
## 1            title                  Title                              
## 2        marginTop             Top margin            25                
## 3     marginBottom          Bottom margin            23                
## 4       marginLeft            Left margin            20                
## 5      marginRight           Right margin            10                
## 6 minBlockDistance Minimum block distance           100                
##     type
## 1 string
## 2 number
## 3 number
## 4 number
## 5 number
## 6 number
## Colors: #5e97eb, #d15014, #e81ccd

Methods are also available directly on the EpivizChart objects. E.g., calls blocks_chart$set(settings=settings, colors=colors) and blocks_chart$print_info() yield the same result.

Printing charts

Charts can be printed as a pdf or png file through the epivizr session. To do so, use the print_chart method:

app$chart_mgr$print_chart(blocks_chart, file_name="blocks_chart", file_type="pdf")

This will create a file named blocks_chart.pdf which will be downloaded through your web browser. It will be found in the default file download location for your web browser.

Adding line plots along the genome

There are a number of different data types available to use through epivizr. You can see a list of R/BioC classes that are supported using ?register. In the previous section, we used the block data type to register a GenomicRanges object. To visualize methylation data at base-pair resolution from the tcga_colon_curves GenomicRanges object, we will use the bp type.

# add low-filter smoothed methylation estimates
means_track <- app$plot(tcga_colon_curves, datasource_name="450kMeth",type="bp",columns=c("cancerMean","normalMean"))

NOTE: You can adjust track settings to change how this new track looks like. For instance, to show all points in the window set the step parameter to 1, and to see a smooth interpolation of the data set the interpolation parameter to basis:

means_track$set(settings=list(step=1, interpolation="basis"))

Notice that we added two lines in this plot, one for mean methylation in cancer and another for mean methylation in normal. The columns argument specifies which columns in mcols(colon_curves) will be displayed.

We can also add a track containing the smooth methylation difference estimate used to define methylation blocks.

diff_chart <- app$plot(tcga_colon_curves, datasource_name="450kMethDiff",type="bp",columns=c("smooth"),ylim=matrix(c(-.5,.5),nc=1))

We pass limits for the y axis in this case. To see other arguments supported, you can use the help framework in R ?"EpivizApp". As before, we can specify settings for this new track.

The app connection again

We can use the app connection object to list charts we have added so far, or to remove charts.

app$chart_mgr$list_charts()
##              id                              type
## 1 epivizChart_1 epiviz.plugins.charts.BlocksTrack
## 2 epivizChart_2 epiviz.plugins.charts.BlocksTrack
## 3 epivizChart_3   epiviz.plugins.charts.LineTrack
## 4 epivizChart_4   epiviz.plugins.charts.LineTrack
##                                  measurements connected
## 1       450k colon_blocks_1:450k colon_blocks          
## 2       450k colon_blocks_2:450k colon_blocks          
## 3 450kMeth_3:cancerMean,450kMeth_3:normalMean          
## 4                       450kMethDiff_4:smooth
app$chart_mgr$rm_chart(means_track)
app$chart_mgr$list_charts()
##              id                              type
## 1 epivizChart_1 epiviz.plugins.charts.BlocksTrack
## 2 epivizChart_2 epiviz.plugins.charts.BlocksTrack
## 3 epivizChart_4   epiviz.plugins.charts.LineTrack
##                            measurements connected
## 1 450k colon_blocks_1:450k colon_blocks          
## 2 450k colon_blocks_2:450k colon_blocks          
## 3                 450kMethDiff_4:smooth

Adding a scatterplot

Now we want to visualize the colon expression data in apColonData object as an MA plot in Epiviz. First, we add an "MA" assay to the ExpressionSet:

keep <- pData(apColonData)$SubType!="adenoma"
apColonData <- apColonData[,keep]
status <- pData(apColonData)$Status
Indexes <- split(seq(along=status),status)

exprMat <- exprs(apColonData)
mns <- sapply(Indexes, function(ind) rowMeans(exprMat[,ind]))
mat <- cbind(colonM=mns[,"1"]-mns[,"0"], colonA=0.5*(mns[,"1"]+mns[,"0"]))
assayDataElement(apColonData, "MA") <- mat
show(apColonData)
## ExpressionSet (storageMode: lockedEnvironment)
## assayData: 5339 features, 2 samples 
##   element names: MA, exprs 
## protocolData: none
## phenoData
##   sampleNames: GSM95473 GSM95474 ... GSM215114 (53 total)
##   varLabels: filename DB_ID ... Status (7 total)
##   varMetadata: labelDescription
## featureData: none
## experimentData: use 'experimentData(object)'
## Annotation: hgu133plus2

epivizr will use the annotation(apColonData) annotation to determine genomic locations using the AnnotationDbi package so that only probesets inside the current browser window are displayed.

eset_chart <- app$plot(apColonData, datasource_name="MAPlot", columns=c("colonA","colonM"), assay="MA")
## Loading required package: hgu133plus2.db
## Loading required package: AnnotationDbi
## Loading required package: org.Hs.eg.db
## 
## 
## 'select()' returned 1:many mapping between keys and columns

In this case, we specify which data is displayed in each axis of the scatter plot using the columns argument. The assay arguments indicates where data is obtained.

The RangedSummarizedExperiment Object

Epiviz is also able to display plots of data in the form of a RangedSummarizedExperiment object. After loading the tcga_colon_expression dataset in the epivizrData package, we can see that this object contains information on 239322 exons in 40 samples.

data(tcga_colon_expression)
show(tcga_colon_expression)
## class: RangedSummarizedExperiment 
## dim: 12800 40 
## metadata(0):
## assays(1): ''
## rownames(12800): 143686 149186 ... 149184 149185
## rowData names(2): exon_id gene_id
## colnames(40): TCGA-A6-5659-01A-01R-1653-07
##   TCGA-A6-5659-11A-01R-1653-07 ... TCGA-D5-5540-01A-01R-1653-07
##   TCGA-D5-5541-01A-01R-1653-07
## colData names(110): bcr_patient_barcode bcr_sample_uuid ...
##   Basename fullID

The assay slot holds a matrix of raw sequencing counts, so before we can plot a scatterplot showing expression, we must first normalize the count data. We use the geometric mean of each row as a reference sample to divide each column (sample) by, then use the median of each column as a scaling factor to divide each row (exon) by.

ref_sample <- 2 ^ rowMeans(log2(assay(tcga_colon_expression) + 1))
scaled <- (assay(tcga_colon_expression) + 1) / ref_sample
scaleFactor <- Biobase::rowMedians(t(scaled))
assay_normalized <- sweep(assay(tcga_colon_expression), 2, scaleFactor, "/")
assay(tcga_colon_expression) <- assay_normalized

Now, using the expression data in the assay slot and the sample data in the colData slot, we can compute mean exon expression by sample type.

status <- colData(tcga_colon_expression)$sample_type
index <- split(seq(along = status), status)
logCounts <- log2(assay(tcga_colon_expression) + 1)
means <- sapply(index, function(ind) rowMeans(logCounts[, ind]))
mat <- cbind(cancer = means[, "Primary Tumor"], normal = means[, "Solid Tissue Normal"])

Now, create a new RangedSummarizedExperiment object with the two column matrix, and all the information about the features of interest, in this case exons, are stored in the rowRanges slot to be queried by Epiviz.

sumexp <- SummarizedExperiment(mat, rowRanges=rowRanges(tcga_colon_expression))
se_chart <- app$plot(sumexp, datasource_name="Mean by Sample Type", columns=c("normal", "cancer"))

Again, the columns argument specifies what data will be displayed along which axis.

Slideshow

We can navigate to a location on the genome using the navigate method of the app object:

app$navigate("chr11", 110000000, 120000000)

There is a convenience function to quickly navigate to a series of locations in succession. We can use that to browse the genome along a ranked list of regions. Let’s navigate to the 5 most up-regulated exons in the colon exon expression data.

foldChange <- mat[,"cancer"]-mat[,"normal"]
ind <- order(foldChange,decreasing=TRUE)

# bounding 1Mb around each exon
slideshowRegions <- trim(rowRanges(sumexp)[ind] + 1000000L)
app$slideshow(slideshowRegions, n=5)

Printing the epivizr workspace

To print the current epiviz workspace as a pdf or png, we use the print_workspace method:

app$print_workspace(file_name="workspace", file_type="pdf")

This will create a file named workspace.pdf and will be downloaded through your web browser. It will be saved to the default file download location for your web browser.

Closing the session

To close the connection to Epiviz and remove all tracks added during the interactive session, we use the stop_app function.

app$stop_app()

Standalone version and browsing arbitrary genomes

The epivizrStandalone all files required to run the web app UI locally. This feature can be used to browse any genome of interest. See that packages vignette for more information.

SessionInfo

sessionInfo()
## R version 3.3.0 (2016-05-03)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 16.04 LTS
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] stats4    parallel  stats     graphics  grDevices utils     datasets 
## [8] methods   base     
## 
## other attached packages:
##  [1] hgu133plus2.db_3.2.2       org.Hs.eg.db_3.3.0        
##  [3] AnnotationDbi_1.35.2       SummarizedExperiment_1.3.2
##  [5] GenomicRanges_1.25.0       GenomeInfoDb_1.9.2        
##  [7] IRanges_2.7.0              S4Vectors_0.11.0          
##  [9] antiProfilesData_1.9.0     Biobase_2.33.0            
## [11] BiocGenerics_0.19.0        epivizr_2.3.3             
## 
## loaded via a namespace (and not attached):
##  [1] Rcpp_0.12.5             BiocInstaller_1.23.3   
##  [3] formatR_1.4             XVector_0.13.0         
##  [5] GenomicFeatures_1.25.9  bitops_1.0-6           
##  [7] tools_3.3.0             zlibbioc_1.19.0        
##  [9] biomaRt_2.29.0          digest_0.6.9           
## [11] RSQLite_1.0.0           evaluate_0.9           
## [13] graph_1.51.0            DBI_0.4-1              
## [15] OrganismDbi_1.15.0      yaml_2.1.13            
## [17] epivizrData_1.1.3       rtracklayer_1.33.2     
## [19] stringr_1.0.0           knitr_1.13             
## [21] Biostrings_2.41.0       R6_2.1.2               
## [23] XML_3.98-1.4            RBGL_1.49.0            
## [25] BiocParallel_1.7.2      rmarkdown_0.9.6        
## [27] magrittr_1.5            epivizrServer_1.1.3    
## [29] Rsamtools_1.25.0        htmltools_0.3.5        
## [31] GenomicAlignments_1.9.0 mime_0.4               
## [33] httpuv_1.3.3            stringi_1.0-1          
## [35] RCurl_1.95-4.8          rjson_0.2.15