Icons as x-axis labels in R - ggplot2

Question

(This is an extension to the question Icons as x-axis labels in R. It looks for a ggplot solution instead of a plot one. Since ggplot is based on grid and plot is based on graphics, the approach is very different)

I would like to plot something like this (from this paper) where icons, in this case small graphs, are used as tick labels.

The accepted answer from the original question is:

library(igraph)    
npoints <- 15
y <- rexp(npoints)
x <- seq(npoints)

# reserve some extra space on bottom margin (outer margin)
par(oma=c(3,0,0,0))
plot(y, xlab=NA, xaxt='n',  pch=15, cex=2, col="red")
lines(y, col='red', lwd=2)

# graph numbers 
x = 1:npoints   

# add offset to first graph for centering
x[1] = x[1] + 0.4
x1 = grconvertX(x=x-0.4, from = 'user', to = 'ndc')
x2 = grconvertX(x=x+0.4, from = 'user', to = 'ndc')

for(i in x){  

  print(paste(i, x1[i], x2[i], sep='; '))

  # remove plot margins (mar) around igraphs, so they appear bigger and 
  # `figure margins too large' error is avoided
  par(fig=c(x1[i],x2[i],0,0.2), new=TRUE, mar=c(0,0,0,0))
  plot(graph.ring(i), vertex.label=NA)  
}

How can we make a similar plot using ggplot?

This is the closer I get:

library(ggplot2)
library(grImport)
library(igraph)

npoints <- 5
y <- rexp(npoints)
x <- seq(npoints)

pics  <- vector(mode="list", length=npoints)
for(i in 1:npoints){
  fileps <- paste0("motif",i,".ps")
  filexml <- paste0("motif",i,".xml")

  # Postscript file
  postscript(file = fileps, fonts=c("serif", "Palatino"))
  plot(graph.ring(i), vertex.label.family="serif", edge.label.family="Palatino")
  dev.off()

  # Convert to xml accessible for symbolsGrob
  PostScriptTrace(fileps, filexml)
  pics[i] <- readPicture(filexml)
}
xpos <- -0.20+x/npoints
my_g <- do.call("grobTree", Map(symbolsGrob, pics, x=xpos, y=0))
qplot(x, y, geom = c("line", "point")) + annotation_custom(my_g, xmin=-Inf, xmax=Inf, ymax=0.4, ymin=0.3)

Answer 1

This builds of your attempt.

(I used set.seed(1) before the rexp function and also tweaked the graph to increase the edge thickness: plot(graph.ring(i), vertex.label=NA, edge.width=30))

Continuing from above:

# Initial plot
p <- qplot(x, y, geom = c("line", "point")) 

# Use the plot to get the x-positions of the labels
g <- ggplotGrob(p)    
xpos <- g$grobs[[grep("axis-b", g$layout$name)]]$children[[2]]$grobs[[2]]$x

# Create grob tree 
my_g <- do.call("grobTree", Map(symbolsGrob, pics, x=xpos, y=0.5))

# Make second plot
# Add extra space under the plot for the images 
# Remove x-axis details
# Note the annotation is below the lower y-axis limit
# The limits were selected by inspection
p2 <- p + annotation_custom(my_g, xmin=-Inf, xmax=Inf, ymax=-0.1, ymin=-0.2) + 
            theme(axis.text.x = element_blank(), 
                  plot.margin=unit(c(1,1,2,1), "cm"))

# remove clipping so the images render
g <- ggplotGrob(p2)
g$layout$clip[g$layout$name=="panel"] <- "off"

grid.newpage()
grid.draw(g)

There will be a way to do this properly / in line with the lovely previous solution, but anyways ...