Compiled date: 2019-05-02
Last edited: 2018-03-08
License: MIT + file LICENSE
The default start configuration with one plot of each type may not always be the most appropriate.
iSEE allows the user to programmatically modify the initial settings (Rue-Albrecht et al. 2018), avoiding the need to click through the choices to obtain the desired panel setup.
Almost every aspect of the initial app state can be customized, down to whether or not the parameter boxes are open or closed!
To demonstrate this feature, let’s assume that we are only interested in feature assay plot panels.
The default set of panels can be changed via the initialPanels argument of the iSEE function call.
Given a SingleCellExperiment or SummarizedExperiment object named sce1 We’ll re-use the example from the previous workflow., the following code opens an app with two adjacent feature assay plots.
Note that each panel is set to occupy half the width of the application window, which is set to 12 units by the shiny package.
library(iSEE)
init <- DataFrame(
Name = c("Feature assay plot 1", "Feature assay plot 2"),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init)The genes to show on the Y-axis in the two plots can be specified via the featAssayArgs argument to iSEE. This is a DataFrame, specifying the initial parameters for each plot.
In this case, we want to modify the YAxis and YAxisFeatName defaults for the two plots:
fexArg <- featAssayPlotDefaults(sce, 2)
fexArg$YAxisFeatName <- c("0610009L18Rik", "0610009B22Rik")
app <- iSEE(sce, initialPanels = init, featAssayArgs = fexArg)This will open the app with two feature assay plots, showing the selected genes. Of course, from this starting point, all the interactive functionality is still available, and new panels can be added, modified and linked via point selection.
The data parameters control the source of the information represented on the X-axis and Y-axis of each plot. Those parameters are accessible at runtime in the eponymous collapsible box.
We refer users to the individual help page of each panel type listed below to learn more about the choices of X-axis variables for each type of panel.
From a running iSEE application, you can also display all the R code that is required to set up the current configuration by clicking on Display panel settings under the wrench icon in the top-right corner.
The nature of the Y-axis is defined by the type of panel.
For instance, column data plot panels require the name of a column in the colData(sce).
Users can preconfigure the Y-axis of individual column data plot panels as follows:
cdArgs <- colDataPlotDefaults(sce, 2)
cdArgs$DataBoxOpen <- TRUE
cdArgs$YAxis <- c("NREADS", "TOTAL_DUP")
init <- DataFrame(
Name = c("Column data plot 1", "Column data plot 2"),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, colDataArgs = cdArgs)The X-axis can be set to different types of variables, which generally requires setting at least two columns in the DataFrame of arguments.
The type of variable is generally set in the XAxis column, while the name of the variable is stored in a different column for each type of variable.
At runtime, this allows the app to remember the last selected variable of each type.
For instance, the XAxis column of feature assay plot arguments can have up to four different values:
"None": does not require any addition input (draws a single violin for all features)"Column data": requires XAxisColData to be set to a column name in the colData(sce)"Feature name": requires either
XAxisFeatName to be set to a feature name (or positional index) in rownames(sce)XAxisRowTable to be set to the name of a Row statistics table panel with an active selection set in its own Selected columnEach of these scenarios is demonstrated below:
fexArg <- featAssayPlotDefaults(sce, 4)
fexArg$DataBoxOpen <- TRUE
# Example 1
fexArg[1, "XAxis"] <- "None"
# Example 2
fexArg[2, "XAxis"] <- "Column data"
fexArg[2, "XAxisColData"] <- "Core.Type"
# Example 3a
fexArg[3, "XAxis"] <- "Feature name"
fexArg[3, "XAxisFeatName"] <- "Zyx"
# Example 3b (also requires a row statistic table)
fexArg[4, "XAxis"] <- "Feature name"
fexArg[4, "XAxisRowTable"] <- "Row statistics table 1"
rowData(sce)[, "gene_id"] <- rownames(sce)
rsArg <- rowStatTableDefaults(sce, 1)
rsArg$Selected <- "Ints2"
rsArg$SearchColumns <- list(gene_id="Ints")
# Initialisation
init <- DataFrame(
Name = c(
paste("Feature assay plot", 1:4),
"Row statistics table 1"),
Width = c(rep(6, 4), 12)
)
app <- iSEE(sce, initialPanels = init, featAssayArgs = fexArg, rowStatArgs = rsArg)Note how Example 3b requires an active row statistics table as a source of selection.
To facilitate visualisation, we added the features identifiers as the gene_id column in rowData(sce),
we preselected the feature "Ints2", and we prefiltered the table using the pattern "Ints" on the gene_id column to show this active selection.
In the case of reduced dimension plots, data parameters include the name of the reduced dimension slot from which to fetch coordinates.
This information is stored in the Type column:
rdArgs <- redDimPlotDefaults(sce, 1)
rdArgs$DataBoxOpen <- TRUE
rdArgs$Type <- "TSNE"
rdArgs$XAxis <- 2
rdArgs$YAxis <- 1
init <- DataFrame(
Name = c("Reduced dimension plot 1"),
Width = c(6)
)
app <- iSEE(sce, initialPanels = init, redDimArgs = rdArgs)We refer users to the ?redDimPlotDefaults help page to learn more about the choices of default parameters for reduced dimension plot panels.
For axes linked to an assay, such as the Y-axis of feature assay plot panels, the assay to display can be set through the Assay column of the arguments:
fexArg <- featAssayPlotDefaults(sce, 2)
fexArg$DataBoxOpen <- TRUE
fexArg$Assay <- c("rsem_counts", "tophat_counts")
init <- DataFrame(
Name = paste("Feature assay plot", 1:2),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, featAssayArgs = fexArg)The visual parameters control the appearance of data points. Those parameters include: color, shape, size, opacity, facet, as well as font size and legend position. Some visual parameters can be associated to variables and controlled through ggplot2 aesthetics, while others can be set to constant user-defined values. All those parameters are accessible at runtime in the eponymous collapsible box.
We refer users to the ?"iSEE point parameters" help page to learn more about the visual parameters variables configurable for each type of panel;
and to the ?"iSEE selection parameters" help page to learn more about the choices of parameters that control the appearance of point selections in receiver plot panels.
Certain visual parameters can be set to a constant user-defined value. Those include: color, transparency (i.e., alpha), downsampling resolution, as well as text font size and legend position.
For instance, the default color of data points in column data plot panels can be set to
a value different than the default "black" through the ColorByDefaultColor column,
while the default transparency value is controlled through the PointAlpha column.
Here, we alter several default visual parameters in the second panel:
cdArgs <- colDataPlotDefaults(sce, 2)
cdArgs$VisualBoxOpen <- TRUE
cdArgs$VisualChoices <- list(c("Color", "Points", "Other"), c("Color", "Points", "Other"))
cdArgs$ColorByDefaultColor[2] <- c("chocolate3")
cdArgs$PointAlpha[2] <- 0.2
cdArgs$PointSize[2] <- 2
cdArgs$Downsample[2] <- TRUE
cdArgs$SampleRes[2] <- 150
cdArgs$FontSize[2] <- 2
init <- DataFrame(
Name = c("Column data plot 1", "Column data plot 2"),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, colDataArgs = cdArgs)Note that for this demonstration, we facilitate visualization of the preconfigured arguments by setting VisualChoices to display both the "Color" and "Shape" UI panels.
The color and point of data points can be linked to variables in a manner similar to the X-axis parameters demonstrated above.
For instance, the color of data points in column data plot panels can be set to a variable in colData(sce) by setting the ColorBy value to "Column data", and the ColorByColData value to the desired coloumn name:
cdArgs <- colDataPlotDefaults(sce, 2)
cdArgs$VisualBoxOpen <- TRUE
cdArgs$VisualChoices <- list(c("Color", "Shape"), c("Color", "Shape"))
cdArgs$ColorBy <- "Column data"
cdArgs$ShapeBy <- "Column data"
cdArgs$ColorByColData <- c("Core.Type", "TOTAL_DUP")
cdArgs$ShapeByColData <- c("Core.Type", "driver_1_s")
init <- DataFrame(
Name = c("Column data plot 1", "Column data plot 2"),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, colDataArgs = cdArgs)Note that points may only be shaped by a categorical variable.
Categorical variables may also be used to facet plots by row and column.
For instance, column data plot panels can be facet by variables stored in the columns of colData(sce).
Users can enable faceting by setting FacetByRow and/or FacetByColumn columns to TRUE, and set RowFacetByColData and ColumnFacetByColData to the appropriate column name in colData(sce).
We demonstrate below how faceting may be enable by row, column, or both:
cdArgs <- colDataPlotDefaults(sce, 3)
cdArgs$VisualBoxOpen <- TRUE
cdArgs$VisualChoices <- list("Facets", "Facets", "Facets")
cdArgs$FacetByRow <- c(TRUE, FALSE, TRUE)
cdArgs$FacetByColumn <- c(FALSE, TRUE, TRUE)
cdArgs$RowFacetByColData <- "driver_1_s"
cdArgs$ColumnFacetByColData <- "Core.Type"
init <- DataFrame(
Name = paste("Column data plot", 1:3),
Width = c(4, 4, 4)
)
app <- iSEE(sce, initialPanels = init, colDataArgs = cdArgs)The initial state of iSEE applications can be configured all the way down to point selections and links between panels.
For instance, in the example below, we preconfigure the SelectByPlot column of a column data plot panel to receive a point selection from a reduced dimension plot panel.
This requires an active selection in the reduced dimension plot panel, which is achieved by preconfiguring the BrushData column of the arguments.
The simplest way to obtain a valid input for the BrushData column is to launch an iSEE application, make the desired selection using a Shiny brush, open the iSEE code tracker, and copy paste the relevant point selection data.
The result should look as below:
# Preconfigure the receiver panel
cdArgs <- colDataPlotDefaults(sce, 1)
cdArgs$SelectBoxOpen <- TRUE
cdArgs$SelectByPlot <- "Reduced dimension plot 1"
cdArgs$SelectEffect <- "Color"
cdArgs$SelectColor <- "darkgoldenrod1"
# Preconfigure the sender panel, including the point selection
rdArgs <- redDimPlotDefaults(sce, 1)
rdArgs$BrushData <- list(
list(xmin = 2.8, xmax = 10.4, ymin = 0.6, ymax = 13.2,
mapping = list(x = "X", y = "Y"), domain = list(left = -14.0, right = 10.9,
bottom = -12.0, top = 16.4), range = list(left = 38.7,
right = 541.5, bottom = 466.0, top = 23.7),
log = list(x = NULL, y = NULL), direction = "xy", brushId = "redDimPlot1_Brush",
outputId = "redDimPlot1")
)
init <- DataFrame(
Name = c("Reduced dimension plot 1", "Column data plot 1"),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, redDimArgs = rdArgs, colDataArgs = cdArgs)Note that in the example above, we chose to color selected data points in the receiver panel, by setting the SelectEffect argument to "Color",
and the color of selected data points to "darkgoldenrod1".
Other choices include "Restrict", to show only the selected data points;
and "Transparent" (the default), to increase the transparency of unselected data points.
An identical process can be followed to preconfigure a lasso point selection:
# Preconfigure the receiver panel
cdArgs <- colDataPlotDefaults(sce, 1)
cdArgs$SelectBoxOpen <- TRUE
cdArgs$SelectByPlot <- "Reduced dimension plot 1"
cdArgs$SelectEffect <- "Color"
cdArgs$SelectColor <- "darkgoldenrod1"
# Preconfigure the sender panel, including the point selection
rdArgs <- redDimPlotDefaults(sce, 1)
rdArgs$LassoData <- list(
list(lasso = NULL, closed = TRUE, panelvar1 = NULL, panelvar2 = NULL, mapping = list(
x = "X", y = "Y"), coord = structure(c(9.7, 4.0, 2.0,
8.2, 10.5, 9.7, 9.0, 12.8, 7.9, 0.9, 2.1, 9.0), .Dim = c(6L,
2L)))
)
init <- DataFrame(
Name = c("Reduced dimension plot 1", "Column data plot 1"),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, redDimArgs = rdArgs, colDataArgs = cdArgs)As we mentioned earlier, certain parameters such as assay types may be set to either character or integer values.
Those include parameters referring to assays *Assay, features *FeatName, and samples *Sample.
We generally recommend to set those parameters using integer values; those are used internally and will always work even in the presence of unnamed assays, for instance:
fexArg <- featAssayPlotDefaults(sce, 2)
fexArg$DataBoxOpen <- TRUE
fexArg$Assay <- c(6, 6)
fexArg$YAxisFeatName <- c(10, 11)
fexArg$XAxis <- "Feature name"
fexArg$XAxisFeatName <- c(12, 13)
init <- DataFrame(
Name = paste("Feature assay plot", 1:2),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, featAssayArgs = fexArg)Alternatively, users may refer to assays, features, and samples by their character name, if any. However, users are advised against mixing character and integer values, as this will cause integer values to be converted to the character type.
For instance, the example below will fail to find an assay called "6" for the first panel, and it will instead fall back to using the first assay ("tophat_counts):
fexArg <- featAssayPlotDefaults(sce, 2)
fexArg$DataBoxOpen <- TRUE
fexArg$Assay <- c(6, "counts")
init <- DataFrame(
Name = paste("Feature assay plot", 1:2),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, featAssayArgs = fexArg)We provide below an example where all parameters are defined as character values:
fexArg <- featAssayPlotDefaults(sce, 2)
fexArg$DataBoxOpen <- TRUE
fexArg$Assay <- c("logcounts", "counts")
fexArg$YAxisFeatName <- c("A3galt2", "A4galt")
fexArg$XAxis <- "Feature name"
fexArg$XAxisFeatName <- c("L3mbtl1", "L3mbtl2")
init <- DataFrame(
Name = paste("Feature assay plot", 1:2),
Width = c(6, 6)
)
app <- iSEE(sce, initialPanels = init, featAssayArgs = fexArg)We refer users to the individual help page of each panel type listed below to learn more about the parameters of each type of panel that accept integer and character values.
By providing a data frame to the tour argument of iSEE, you can create your own tour that will start when the app is launched2 In theory. On servers, sometimes the tour does not recognize the UI elements at start-up and needs to be restarted via the “Click me for quick tour” button to work properly..
The data frame should have two columns, element and intro:
introtour <- read.delim(
system.file("extdata/intro_firststeps.txt", package = "iSEE"),
sep = ";", header = TRUE)
head(introtour)
#> element
#> 1 #Welcome
#> 2 #allpanels
#> 3 #redDimPlot1
#> 4 #colDataPlot1
#> 5 #featAssayPlot1
#> 6 #rowStatTable1
#> intro
#> 1 Welcome to the interactive tour for iSEE - the Interactive SummarizedExperiment Explorer.<br/><br/>You will be shown around the different components of iSEE and learn the basic usage mechanisms by doing. Highlighted elements will respond to the user's actions, while the rest of the UI will be shaded. You will be prompted at particular spots to perform some actions, which will be marked with <strong>Action:</strong> text. Please take care to follow these instructions, since later parts of the tour may assume that all the actions from previous steps have been performed.<br/><br/><strong>Action:</strong> now, click on the 'Next' button or use the right arrow of your keyboard to proceed into your tour.
#> 2 iSEE provides a Shiny interface that allows you to generate a series of panels for exploring <code>SummarizedExperiment</code> objects. Here, we use single-cell RNA sequencing data from the <a href=https://doi.org/doi:10.18129/B9.bioc.scRNAseq><i>scRNAseq</i> package</a>, specifically a subset of gene expression profiles from cells in the mouse visual cortex <a href=https://doi.org/10.1038/nn.4216>(Tasic <i>et al.</i>, 2016)</a>. Each column of the <code>SummarizedExperiment</code> corresponds to a cell, while each row corresponds to a gene.<br/><br/>Using iSEE, you can generate a variety of different panel types to visualize the data. These are described in more detail in the following steps.
#> 3 For example, you can construct a <font color=#402ee8>Reduced dimension plot</font> (e.g., PCA, <i>t</i>-SNE) to efficiently explore high-dimensional datasets. Here, each point represents a cell.
#> 4 You can construct a <font color=#402ee8>Column data plot</font> involving column-level metadata, where each point represents a column (in this case, a cell) in the <code>SummarizedExperiment</code> object. Points can be displayed using violin plots stratified by an experimental factor on the x-axis, or as a scatter plot involving continuous variables on both the x- and y-axes.
#> 5 A <font color=#402ee8>Feature assay plot</font> displays the assay values (here, gene expression) across all cells. Like the <font color=#402ee8>Column data plot</font>, points represent cells and can be shown in both violin or scatter plots, depending on the variable selected on the x-axis.
#> 6 A <font color=#402ee8>Row statistic table</font> displays values in the <code>rowData</code> slot of your input object. They can also be used to efficiently select genes for the other plots.Each entry of the element column contains the name of a UI element in the application, prefixed by a hash sign (#).
The intro column contains the corresponding text (or basic HTML) that is to be shown at each step.
The simplest way to get started is to copy the intro_firststeps.txt file from the inst/extdata folder and edit it for your specific data set.
More customized tours require some knowledge of the names of the UI elements to put in the element column.
We recommend one of the following options:
Tools menu.
Select Web Developer, and in the submenu, select Inspector.
This will toggle a toolbar, and you will be able to read out the name of the element of interest when you hover and click on it.
If you want to select another element, you might need to re-click on the icon in the upper left corner of the toolbox, Pick an element from the page.Develop menu and select Show Web Inspector.
To toggle the selection of elements, you need to click on the crosshair icon in the top part of the toolbox, then again, explore the desired element by clicking or hovering on it.View menu, select first Developer and then Developer Tools.
Click then on the selecting arrow in the top left corner, and similarly to the other browsers, hover or click on the element of interest to obtain its name.Most elements can be identified using the above strategies.
Selectize widgets are trickier but can be handled with, e.g., #heatMapPlot1_ColData + .selectize-control.
Please see the intro_firststeps.txt file in the inst/extdata folder for more of these examples.
Sometimes it is useful to place one step of the tour in the center.
To do so, simply put a hash sign before a word which does not link directly to any CSS selector (e.g., as we do for #Welcome) in the corresponding element column.
Users should refer to the following help pages for the full list of values that can be specified in iSEE:
?redDimPlotDefaults?colDataPlotDefaults?featAssayPlotDefaults?rowDataPlotDefaults,?rowStatTableDefaults?sampAssayPlotDefaults?heatMapPlotDefaults?"iSEE point parameters"?"iSEE selection parameters"Some fairly complex configurations for a variety of data sets can be found at https://github.com/LTLA/iSEE2018. These may serve as useful examples for setting up your own configurations.
sessionInfo()
#> R version 3.6.0 (2019-04-26)
#> Platform: x86_64-pc-linux-gnu (64-bit)
#> Running under: Ubuntu 18.04.2 LTS
#>
#> Matrix products: default
#> BLAS: /home/biocbuild/bbs-3.9-bioc/R/lib/libRblas.so
#> LAPACK: /home/biocbuild/bbs-3.9-bioc/R/lib/libRlapack.so
#>
#> 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] parallel stats4 stats graphics grDevices utils datasets
#> [8] methods base
#>
#> other attached packages:
#> [1] scater_1.12.0 ggplot2_3.1.1
#> [3] scRNAseq_1.9.0 iSEE_1.4.0
#> [5] SingleCellExperiment_1.6.0 SummarizedExperiment_1.14.0
#> [7] DelayedArray_0.10.0 BiocParallel_1.18.0
#> [9] matrixStats_0.54.0 Biobase_2.44.0
#> [11] GenomicRanges_1.36.0 GenomeInfoDb_1.20.0
#> [13] IRanges_2.18.0 S4Vectors_0.22.0
#> [15] BiocGenerics_0.30.0 BiocStyle_2.12.0
#>
#> loaded via a namespace (and not attached):
#> [1] nlme_3.1-139 bitops_1.0-6
#> [3] bit64_0.9-7 tools_3.6.0
#> [5] irlba_2.3.3 R6_2.4.0
#> [7] DT_0.5 vipor_0.4.5
#> [9] DBI_1.0.0 lazyeval_0.2.2
#> [11] mgcv_1.8-28 colorspace_1.4-1
#> [13] withr_2.1.2 gridExtra_2.3
#> [15] tidyselect_0.2.5 bit_1.1-14
#> [17] compiler_3.6.0 BiocNeighbors_1.2.0
#> [19] shinyjs_1.0 colourpicker_1.0
#> [21] bookdown_0.9 scales_1.0.0
#> [23] stringr_1.4.0 digest_0.6.18
#> [25] rmarkdown_1.12 rentrez_1.2.2
#> [27] XVector_0.24.0 pkgconfig_2.0.2
#> [29] htmltools_0.3.6 htmlwidgets_1.3
#> [31] rlang_0.3.4 RSQLite_2.1.1
#> [33] DelayedMatrixStats_1.6.0 shiny_1.3.2
#> [35] jsonlite_1.6 dplyr_0.8.0.1
#> [37] RCurl_1.95-4.12 magrittr_1.5
#> [39] BiocSingular_1.0.0 GenomeInfoDbData_1.2.1
#> [41] Matrix_1.2-17 ggbeeswarm_0.6.0
#> [43] Rcpp_1.0.1 munsell_0.5.0
#> [45] viridis_0.5.1 stringi_1.4.3
#> [47] yaml_2.2.0 rintrojs_0.2.0
#> [49] zlibbioc_1.30.0 Rtsne_0.15
#> [51] plyr_1.8.4 grid_3.6.0
#> [53] blob_1.1.1 promises_1.0.1
#> [55] shinydashboard_0.7.1 crayon_1.3.4
#> [57] miniUI_0.1.1.1 lattice_0.20-38
#> [59] cowplot_0.9.4 splines_3.6.0
#> [61] knitr_1.22 pillar_1.3.1
#> [63] igraph_1.2.4.1 reshape2_1.4.3
#> [65] XML_3.98-1.19 glue_1.3.1
#> [67] evaluate_0.13 BiocManager_1.30.4
#> [69] httpuv_1.5.1 gtable_0.3.0
#> [71] purrr_0.3.2 assertthat_0.2.1
#> [73] xfun_0.6 rsvd_1.0.0
#> [75] mime_0.6 xtable_1.8-4
#> [77] later_0.8.0 viridisLite_0.3.0
#> [79] tibble_2.1.1 beeswarm_0.2.3
#> [81] AnnotationDbi_1.46.0 memoise_1.1.0
#> [83] shinyAce_0.3.3
# devtools::session_info()Rue-Albrecht, K., F. Marini, C. Soneson, and A. T. L. Lun. 2018. “ISEE: Interactive Summarizedexperiment Explorer.” F1000Research 7 (June):741.