Introduction
A key challenge in the analysis of high-dimensional data is
visualization. In the context of population analysis, once a group of
cells has been identified either through manual gating or clustering how
do we efficiently check the main trend and homogeneity of this group.
Fan plots are an attractive method that combines the easy interpretation
of line charts, combined with an overview of the underlying distribution
similar to that provided through smooth scatter plots. We will
illustrate these 2 properties using an example flow cytometry
dataset.
Building a Fan plot
To build a fan plot, one first breaks the data distribution into a
predefined number of bins and records the bin limits. In the default
implementation (using 100 bins), the center bin corresponds to the
distribution median. Each bin is plotted as a small rectangle, filled
with a color shade that corresponds to the bin position. The center bin
gets the strongest shade of the base color, and the color is
increasingly decreased for bins that correspond to higher or smaller
values. When multiple channels are plotted side-by-side, the rectangle
can be joined leading to a visual metaphor similar to a line chart.
Installation
The package can be installed using the following command:
devtools::install_github('yannabraham/cytofan')
Once installed the package can be loaded using the standard
library command.
library(cytofan)
#> Loading required package: ggplot2
Visualizing mass cytometry data from the bodenmiller
package
To illustrate the use of cytofan in the context of cytometry, we will
use data published in Bodenmiller et al Nat
Biotech 2012. Samples corresponding to untreated cells, stimulated
with BCR/FcR-XL, PMA/Ionomycin or vanadate or unstimulated, were
downloaded from CytoBank as FCS files (NB the original data doesn’t seem
to be available anymore). Data was extracted and normalized using the
arcsinh function with a cofactor of 5.
We assemble a dataset containing all reference samples, and both
phenotypic and functional channels:
data('refPhenoMat')
data('refFuncMat')
data('refAnnots')
refFullFrame <- melt(cbind(refPhenoMat,
refFuncMat))
names(refFullFrame) <- c('cell_id','Channel','value')
refFullFrame$Channel <- factor(refFullFrame$Channel,
levels=c(colnames(refPhenoMat),
colnames(refFuncMat)))
refFullFrame$Cells <- rep(refAnnots$Cells,
ncol(refPhenoMat)+ncol(refFuncMat))
We first visualize the profile of CD4+ T cells using fan
plots:
refFullFrame %>%
filter(Cells=='cd4+') %>%
ggplot(aes(x=Channel,y=value))+
geom_fan()+
facet_grid(.~Cells)+
theme(axis.text.x = element_text(angle = 45, hjust = 1))
#> Warning: The `<scale>` argument of `guides()` cannot be `FALSE`. Use "none" instead as
#> of ggplot2 3.3.4.
#> ℹ The deprecated feature was likely used in the cytofan package.
#> Please report the issue at <https://github.com/yannabraham/cytofan/issues>.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
#> generated.
#> Warning: The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
Fan plots can be used to compare populations, eg CD4+ and
CD8+ T cells:
refFullFrame %>%
filter(Cells %in% c('cd4+','cd8+')) %>%
ggplot(aes(x=Channel,y=value))+
geom_fan()+
facet_grid(Cells~.)+
theme(axis.text.x = element_text(angle = 45, hjust = 1))
#> Warning: The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
Visualizing functional differences after stimulation.
We assemble a dataset containing all untreated samples, stimulated
with BCR/FcR-XL, PMA/Ionomycin or vanadate, and both phenotypic and
functional channels:
data('untreatedPhenoMat')
data('untreatedFuncMat')
data('untreatedAnnots')
untreatedFullFrame <- melt(cbind(untreatedPhenoMat,
untreatedFuncMat))
names(untreatedFullFrame) <- c('cell_id','Channel','value')
untreatedFullFrame$Channel <- factor(untreatedFullFrame$Channel,
levels=c(colnames(untreatedPhenoMat),
colnames(untreatedFuncMat)))
untreatedFullFrame$Cells <- rep(untreatedAnnots$Cells,
ncol(untreatedPhenoMat)+ncol(untreatedFuncMat))
untreatedFullFrame$Treatment <- rep(untreatedAnnots$Treatment,
ncol(untreatedPhenoMat)+ncol(untreatedFuncMat))
We then visualize the effects of stimulation on the profile of
CD4+ T cells:
untreatedFullFrame %>%
filter(Cells=='cd4+') %>%
ggplot(aes(x=Channel,y=value))+
geom_fan()+
facet_grid(Treatment~Cells)+
theme(axis.text.x = element_text(angle = 45, hjust = 1))
#> Warning: The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
We can also compare the effects of treatments across populations, eg
CD4+ and CD8+ T cells:
untreatedFullFrame %>%
filter(Cells %in% c('cd4+','cd8+')) %>%
ggplot(aes(x=Channel,y=value))+
geom_fan()+
facet_grid(Cells~Treatment)+
theme(axis.text.x = element_text(angle = 45, hjust = 1))
#> Warning: The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
#> The following aesthetics were dropped during statistical transformation: y.
#> ℹ This can happen when ggplot fails to infer the correct grouping structure in
#> the data.
#> ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
#> variable into a factor?
Conclusion
Fan plots provide a straightforward method to compare profiles across
conditions, while accounting for homogeneity of response or composition
of a given compartment. Fan plots enable quick exploration and
comparison of groups of cells such as manual gates and clusters.