Fan plots are an attractive method to visualize trends in complex data such as those generated in Flow and Mass Cytometry experiments. Using ggplot2, one can easily treillis the different conditions to observe trends in the data.

cytometry_fan.R

## build a fan plot from scratch using dplyr & ggplot2 for visualization of Flow data
## after an original idea by Guy J. Abel
## http://gjabel.wordpress.com/category/r/fanplot/

## libraries
library(flowCore)
library(reshape2)
library(dplyr)
library(ggplot2)
library(RColorBrewer)

## functions
do.fan <- function(x,step=0.01) {
  data.frame(ymin=quantile(x,probs=seq(0,1,step))[-length(seq(0,1,step))],
             ymax=quantile(x,probs=seq(0,1,step))[-1],
             id=seq(1,length(seq(step,1,step))),
             percent=abs(seq(step,1,step)-0.5))
}

StatFan <- ggproto("StatFan", Stat,
                   required_aes = "y",
                   default_aes = aes(fill=stat(percent),group=stat(id)),
                   compute_group = function(data,scales,step=0.01) do.fan(data$y,step)
)

stat_fan <- function(mapping = NULL, data = NULL, geom = NULL,
                     position = "identity", na.rm = FALSE, show.legend = NA,
                     inherit.aes = TRUE, ...) {
  list(
    layer(
      stat = StatFan, data = data, mapping = mapping, geom = geom,
      position = position, show.legend = show.legend, inherit.aes = inherit.aes,
      params = list(na.rm = na.rm, ...)
    )
  )
}

# GeomFan <- ggproto( "GeomFan", GeomRibbon )

geom_fan <- function(mapping = NULL, data = NULL,
                     position = "identity", na.rm = FALSE, show.legend = NA,
                     inherit.aes = TRUE, colorbase='Oranges', ...) {
  list(
    layer(
      stat = StatFan, geom = GeomRibbon, data = data, mapping = mapping,
      position = position, show.legend = show.legend, inherit.aes = inherit.aes,
      params = list(na.rm = na.rm, ...)
    ),
    scale_fill_gradientn(colours=rev(RColorBrewer::brewer.pal(9,colorbase)) ),
    guides(fill=FALSE)
  )
}

## parameters
theme_set( theme_bw(base_size=14) )

## load the data
data(GvHD)

## first lets transform the data using a logicle function
# estimate the parameters of the logicle function from the first sample
logicleTrans <- estimateLogicle(GvHD[[1]],
                                channels=c('FL1-H','FL2-H','FL3-H','FL4-H'))

tGvHD <- logicleTrans %on% GvHD

## extract the values
mat <- fsApply(tGvHD[,c('FL1-H','FL2-H','FL3-H','FL4-H')],exprs)

## get the annotation
df <- pData(tGvHD)[unlist(sapply(sampleNames(tGvHD),function(x) rep(x,nrow(tGvHD[[x]])))),]

## bind the 2 together...
df <- cbind(df,mat)

## ... and melt
melted <- melt(df,id.var=varLabels(tGvHD))

## fix the channel names
clean.names <- with(pData(parameters(tGvHD[[1]])),
                    desc[match(levels(melted$variable),name)])
clean.names <- sapply(clean.names,function(x) strsplit(x,' ')[[1]][1])
levels(melted$variable) <- clean.names

## Then you can compute and treillis your fan plot however you like
step <- 0.01

ggplot(melted,
       aes(x=variable,y=value))+
  geom_fan()+
  facet_grid(Visit~Patient)

ggplot(melted,
       aes(x=variable,y=value))+
  geom_fan(step=0.05)+
  facet_grid(Visit~Patient)

ggplot(melted,
       aes(x=variable,y=value))+
  geom_fan()+
  facet_grid(Grade~Visit,labeller=label_both)

ggplot(melted %>% filter(Visit=='1'),
       aes(x=variable,y=value))+
  geom_fan(colorbase='Greens')+
  facet_wrap(~Patient)

# it is still possible to adjust the color of the fan

ggplot(melted %>% filter(Visit=='1'),
       aes(x=variable,y=value))+
  geom_fan()+
  facet_wrap(~Patient)+
  scale_fill_gradientn(colours=rev(brewer.pal(9,'Greens')) )