ggcyto : Visualize `Cytometry` data with `ggplot`

library(knitr)
opts_chunk$set(message = FALSE, warning = FALSE, fig.height= 3, fig.width= 5)
library(ggcyto)
dataDir <- system.file("extdata",package="flowWorkspaceData")

1: suppoort 3 types of plot constructor

low level: ggplot

The overloaded fority methods empower ggplot to work with all the major Cytometry data structures right away, which allows users to do all kinds of highly customized and versitled plots.

GatingSet

gs <- load_gs(list.files(dataDir, pattern = "gs_manual",full = TRUE))
attr(gs, "subset") <- "CD3+"
ggplot(gs, aes(x = `<B710-A>`, y = `<R780-A>`)) + geom_hex(bins = 128) + scale_fill_gradientn(colours = gray.colors(9))

flowSet/ncdfFlowSet/flowFrame

fs <- getData(gs, "CD3+")
ggplot(fs, aes(x = `<B710-A>`)) + geom_density(fill = "blue", alpha= 0.5)

gates

gates <- filterList(getGate(gs, "CD8"))
ggplot(gs, aes(x = `<B710-A>`, y = `<R780-A>`)) + geom_hex(bins = 128) + geom_polygon(data = gates, fill = "transparent", col = "purple")

medium level: ggcyto

ggcyto constructor along with overloaded + operator encapsulate lots of details that might be tedious and intimidating for many users.

ggcyto(gs, aes(x = CD4, y = CD8)) + geom_hex(bins = 128) + geom_gate("CD8")

It simplies the plotting by: add a default scale_fill_gradientn for you fuzzy-matching in aes by either detector or fluorochromes names determine the parent popoulation automatically exact and plot the gate object by simply referring to the child population name

top level: autoplot

Inheriting the spirit from ggplot's Quick plot, it further simply the plotting job by hiding more details from users and taking more assumptions for the plot.

#1d
autoplot(fs, "CD4")
#2d
autoplot(fs, "CD4", "CD8", bins = 64)

autoplot(gs, c("CD4", "CD8"), bins = 64)

2: in-line transformation

It is done by different scales layers speically designed for cytometry

data(GvHD)
fr <- GvHD[[1]]
p <- autoplot(fr, "FL1-H")
p #raw scale
p + scale_x_logicle() #flowCore logicle scale
p + scale_x_flowJo_fasinh() # flowJo fasinh
p + scale_x_flowJo_biexp() # flowJo biexponential

3: generic geom_gate layer

It hides the complex details pf plotting different geometric shapes

fr <- fs[[1]]
p <- autoplot(fr,"CD4", "CD8") + ggcyto_par_set(limits = "instrument")
#1d gate vertical
gate_1d_v <- openCyto::gate_mindensity(fr, "<B710-A>")
p + geom_gate(gate_1d_v)
#1d gate horizontal
gate_1d_h <- openCyto::gate_mindensity(fr, "<R780-A>")
p + geom_gate(gate_1d_h)
#2d rectangle gate
gate_rect <- rectangleGate("<B710-A>" = c(gate_1d_v@min, 4e3), "<R780-A>" = c(gate_1d_h@min, 4e3))
p + geom_gate(gate_rect)
#ellipsoid Gate
gate_ellip <- getGate(gs[[1]], "CD4")
class(gate_ellip)
p + geom_gate(gate_ellip)

4: geom_stats

p <- ggcyto(gs, aes(x = "CD4", y = "CD8"), subset = "CD3+") + geom_hex()
p + geom_gate("CD4") + geom_stats()
p + geom_gate("CD4") + geom_stats(type = "count") #display cell counts 

5: axis_inverse_trans

It can display the log scaled data in the original value

p # axis display the transformed values
p + axis_x_inverse_trans() # restore the x axis to the raw values

It currently only works with GatingSet.

6: auto limits

Optionally you can set limits by instrument or data range

p <- p + ggcyto_par_set(limits = "instrument")
p

7: labs_cyto

You can choose between marker and channel names (or both by default)

p + labs_cyto("markers")

8: ggcyto_par_set

It aggregates the different settings in one layer

#put all the customized settings in one layer
mySettings <- ggcyto_par_set(limits = "instrument"
                             , facet = facet_wrap("name")
                             , hex_fill = scale_fill_gradientn(colours = rev(RColorBrewer::brewer.pal(11, "Spectral")))
                            , lab = labs_cyto("marker")
                            )
# and use it repeatly in the plots later (similar to the `theme` concept)
p + mySettings

Currently we only support 4 settings, but will add more in future.

9: as.ggplot

It allows user to convert ggcyto objects to pure ggplot objects for further the manipulating jobs that can not be done within ggcyto framework.

class(p) # may not fully compatile with all the `ggplot` functions
p1 <- as.ggplot(p)
class(p1) # a pure ggplot object, thus can work with all the `ggplot` features

10: ggcyto_layout

Layout many gate plots on the same page

When plooting a GatingHierarchy, multiple cell populations with their asssociated gates can be plotted in different panels of the same plot.

gh <- gs[[1]]
nodes <- getNodes(gh, path = "auto")[c(3:9, 14)]
nodes
p <- autoplot(gh, nodes, bins = 64)
class(p)
p

As you see, this generates a special ggcyto_GatingLayout object which is a container storing multiple ggcyto objects. You can do more about the plot layout with the helper function ggcyto_arrange. For example, to arrange it as one-row gtable object

gt <- ggcyto_arrange(p, nrow = 1)
class(gt)
plot(gt)

or even combine it with other ggcyto_GatingLayout objects(or any gtable objects) and print it on the sampe page

p2 <- autoplot(getData(gh, "CD3+")[,5:8]) # some density plot
p2@arrange.main <- ""#clear the default title
gt2 <- ggcyto_arrange(p2, nrow = 1)

gt3 <- gridExtra::gtable_rbind(gt, gt2)
plot(gt3)


Try the ggcyto package in your browser

Any scripts or data that you put into this service are public.

ggcyto documentation built on Jan. 15, 2019, 6:01 p.m.