missing factors in geom_barh with facet_plot(?)

[xvazquezc]

trafficstars

Hi,

I've been trying to plot a bar chart with community composition together with a hierarchical clustering dendrogram, and while I was able to plot it, there are always some of the factors that while they still appear in the legend, they are missing from the bar chart. I found out first trying to plot 10 factors but it happens even with only 5. Files for reproducible example: example_files.zip

library(ggtree)
library(tidyverse)
library(ggstance)

dist <- read.csv("dist.csv", header = TRUE, row.names = 1)
dist1 <- as.dist(dist)
hc <- hclust(dist1)
den <- as.dendrogram(hc)
p <- ggtree(den)

# load table with data for bar chart
mytable <- read.csv("mytable.csv", header = T, row.names = 1)

#put together
facet_plot(p, data = mytable, geom = geom_barh, panel = "Phylum", colour = "black",
           mapping = aes(fill = as.factor(Phylum),x=Abundance), stat = "identity") +
  geom_tiplab(size = 2) + xlim(0, 0.38)

This is the output, which in this simplified example, misses all Aquificae (should be >50% in some samples):

To be sure it wasn't a problem with mytable, I run it with ggplot as normal geom_bar and as geom_barh from ggstance (only showing the latter here):

ggplot(mytable, aes(y=LIMSID, x = Abundance)) +
  geom_barh(aes(fill=Phylum), position = "stack", stat = "identity")

and as you can see there are obvious differences:

I did further testing by removing one of the factors in mytable$Phylum:

mytable2 <- mytable[mytable$Phylum != "Aquificae", ]
facet_plot(p, data = mytable2, geom = geom_barh, panel = "Phylum", colour = "black",
           mapping = aes(fill = as.factor(Phylum),x=Abundance), stat = "identity") +
  geom_tiplab(size = 2) + xlim(0, 0.38)

And now it shows 4 categories without issue: or it seems...

If I remove a different one, the offending "Aquificae" only shows up in a few samples where is a very minor component:

mytable3 <- mytable[mytable$Phylum != "Proteobacteria", ]
facet_plot(p, data = mytable2, geom = geom_barh, panel = "Phylum", colour = "black",
           mapping = aes(fill = Phylum, x=Abundance), stat = "identity") +

I've run this on R3.6.3 and both ggtree versions 2.0.4 (from Bioconductor) and 2.3.0 (from Github). I also usually get the following message, which I wouldn't be surprised it would be related, but it doesn't matter if I remove zeroes from my table, the rows removed barely change:

Warning messages:
1: Removed 53 rows containing missing values (position_stack). 
2: Removed 82 rows containing missing values (geom_barh). 

I've also tried to ordered the data on my table based on the order from p$data$label or the order of the leaves from the hclust output, but all the same.

Can anyone reproduce this? Is there anything obvious I'm missing?

Thank you in advance, Xabi

[xiangpin]

Since you used xlim. The range of x is 0 ~ 0.38 in your example, the x of geom_barh will be removed. To solve the problem, you should not use xlim. And you can use geom_blank to show the tiplabels. PS: you can use geom_bar of ggplot2 (new version) instead of geom_barh of ggstance by adding orientation="y".

library(ggtree)
library(tidyverse)

dist <- read.csv("dist.csv", header = TRUE, row.names = 1)
dist1 <- as.dist(dist)
hc <- hclust(dist1)
den <- as.dendrogram(hc)
######## geom_blank##############
p <- ggtree(den) + geom_blank(aes(x=0.42))

# load table with data for bar chart
mytable <- read.csv("mytable.csv", header = T, row.names = 1)

#put together
facet_plot(p, data = mytable, geom = geom_bar, panel = "Phylum", colour = "black",
           mapping = aes(fill = as.factor(Phylum),x=Abundance), stat = "identity", orientation="y") +
  geom_tiplab(size = 2)