This document outlines how to set up a flow cytometry dashboard using the flowDashboard Shiny modules.
flowDashboard Fits in an Analysis WorkflowflowDashboard enables you to visualize data at both the Sample and Population Level. The package assumes that you either have FCS data in a directory, or have generated a GatingSet for your gating scheme. This gating scheme can either be automated (from a pipeline such as openCyto) or manually (imported from flowJo).
There are multiple modules that will enable you to examine each step of the analysis.
Annotation display - displays annotation for all the FCS files in an experiment.QC plots - plots that let you compare the overall state of the data before and after preprocessing.Gating plots - plots that display how the data was gatedDot plots - to visualize populations based on a covariateViolin plots of markers in a populationWaterfall plots to examine how population percentage correlates with an outcome.Coexpression plots - a heatmap and a two-way expression explorer.Once your data is in a gatingSet (preferably in ncdf format), we will need to transform and prep it into the data.table format. This package uses data.table because it is a very fast format for accessing, merging, sorting, and displaying data.
The annotation structure is another data.table that we need for the dashboard. Based on your unit of analysis and what the FCS file represents (which may be patients, cell lines, or conditions within a patient), you will need an annotation table that provides covariate information at the level of analysis you are interested in. You can generate the skeleton of an annotation table with the `` command, or you can use the phenoData slot in your GatingSet.
Please note that the usefulness of flowDashboard is only as good as your annotation. Careful curation of variables in your annotation will allow you to sort, subset, and aggregate on these variables.
library(flowWorkspace)## Loading required package: flowCore## Loading required package: ncdfFlow## Loading required package: RcppArmadillo## Loading required package: BHlibrary(flowDashboard)## Loading required package: tidyr## Loading required package: dplyr##
## Attaching package: 'dplyr'## The following object is masked from 'package:ncdfFlow':
##
## filter## The following object is masked from 'package:flowCore':
##
## filter## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union## Loading required package: shiny## Loading required package: data.table##
## Attaching package: 'data.table'## The following objects are masked from 'package:dplyr':
##
## between, first, last## Loading required package: dtplyr## Loading required package: ggvis## Loading required package: lazyeval## Loading required package: gplots##
## Attaching package: 'gplots'## The following object is masked from 'package:stats':
##
## lowess## Loading required package: ggplot2##
## Attaching package: 'ggplot2'## The following object is masked from 'package:ggvis':
##
## resolution## Loading required package: R6## Loading required package: forcats## Loading required package: stringigsFile <- system.file("extdata/gvHDgs/", package = "flowDashboard")
gs <- load_gs(gsFile)## loading R object...## loading tree object...## Done##look at phenoData slot in gs
phenotypes <- pData(gs@data@phenoData)
phenotypes## Patient Visit Days Grade name
## s5a01 5 1 -6 3 s5a01
## s5a02 5 2 0 3 s5a02
## s5a03 5 3 6 3 s5a03
## s5a04 5 4 12 3 s5a04
## s5a05 5 5 19 3 s5a05
## s5a06 5 6 26 3 s5a06
## s5a07 5 7 33 3 s5a07
## s6a01 6 1 -8 3 s6a01
## s6a02 6 2 0 3 s6a02
## s6a03 6 3 5 3 s6a03
## s6a04 6 4 12 3 s6a04
## s6a05 6 5 19 3 s6a05
## s6a06 6 6 27 3 s6a06
## s6a07 6 7 32 3 s6a07
## s7a01 7 1 -4 3 s7a01
## s7a02 7 2 0 3 s7a02
## s7a03 7 3 7 3 s7a03
## s7a04 7 4 14 3 s7a04
## s7a05 7 5 21 3 s7a05
## s7a06 7 6 28 3 s7a06##look at marker information in gs
markerInfo <- pData(parameters(gs@data[[1]]))
markerInfo## name desc range minRange maxRange
## $P1 FSC-H FSC-Height 1024 0.0000000 1023.000000
## $P2 SSC-H SSC-Height 1024 0.0000000 1023.000000
## $P3 FL1-H CD15 FITC 1024 0.8813736 9.903488
## $P4 FL2-H CD45 PE 1024 0.8813736 9.903488
## $P5 FL3-H CD14 PerCP 1024 0.8813736 9.903488
## $P6 FL2-A <NA> 1024 0.0000000 7.623642
## $P7 FL4-H CD33 APC 1024 1.0000000 10000.000000
## $P8 Time Time (51.20 sec.) 1024 0.0000000 1023.000000#build a qcFlowObj
QCO <- QCOFromGatingSet(gs, samplePop = 1000)
#show annotation slot
QCO$annotation## Patient Visit Days Grade name
## 1: 5 1 -6 3 s5a01
## 2: 5 2 0 3 s5a02
## 3: 5 3 6 3 s5a03
## 4: 5 4 12 3 s5a04
## 5: 5 5 19 3 s5a05
## 6: 5 6 26 3 s5a06
## 7: 5 7 33 3 s5a07
## 8: 6 1 -8 3 s6a01
## 9: 6 2 0 3 s6a02
## 10: 6 3 5 3 s6a03
## 11: 6 4 12 3 s6a04
## 12: 6 5 19 3 s6a05
## 13: 6 6 27 3 s6a06
## 14: 6 7 32 3 s6a07
## 15: 7 1 -4 3 s7a01
## 16: 7 2 0 3 s7a02
## 17: 7 3 7 3 s7a03
## 18: 7 4 14 3 s7a04
## 19: 7 5 21 3 s7a05
## 20: 7 6 28 3 s7a06#show first few lines of qcData
QCO$qcData## idVar cellNum variable value
## 1: s5a01 1560 FSC.Height 812
## 2: s5a01 2999 FSC.Height 164
## 3: s5a01 1825 FSC.Height 256
## 4: s5a01 811 FSC.Height 79
## 5: s5a01 1275 FSC.Height 218
## ---
## 159996: s7a06 10589 Time..51.20.sec.. 79
## 159997: s7a06 3426 Time..51.20.sec.. 26
## 159998: s7a06 6128 Time..51.20.sec.. 46
## 159999: s7a06 8299 Time..51.20.sec.. 61
## 160000: s7a06 5875 Time..51.20.sec.. 43The QC Heatmap provides a quick overview of how markers vary across populations. This heatmap is based on the idea of examining a QC panel of markers to compare whether there is a large technical variation between samples.
The QC Violin plots allow you to compare fluoresence distibution of individual functions.
The underlying QC data is sampled using the buildSampledDataList() function. This function generates a data.table that samples from each FCS file for a selected control panel of markers.
The gating module consists of a heatmap visualization and plots for each cell in the heatmap, corresponding to the gating scheme. This module is useful in understanding the impact of automated gating pipelines, such as openCyto, and comparing manual gating schemes across samples.
The plots for each sample must be precomputed using the plotAllPopulations() function. This function takes a GatingSet as input.
The Violin plots