Adding a ggtree object to already existing ggplot with shared y-axis - r

Question

I have the following data and plot:
Data:
structure(list(type = c("mut", "mut", "mut", "mut", "mut", "mut",
"mut", "mut", "gene", "gene", "gene", "gene"), gene = c("gyrA",
"gyrA", "gyrB", "gyrB", "parC", "parC", "parE", "parE", "qnrA1",
"qnrA1", "sul3", "sul3"), type2 = c(1, 1, 1, 1, 1, 1, 1, 1, 2,
2, 2, 2), id = c("2014-01-7234-1-S", "2015-01-3004-1-S", "2014-01-2992-1-S",
"2016-17-299-1-S", "2015-01-2166-1-S", "2014-01-4651-1-S", "2016-02-514-2-S",
"2016-02-402-2-S", "2016-02-425-2-S", "2015-01-5140-1-S", "2016-02-522-2-S",
"2016-02-739-2-S"), result = c("1", "0", "0", "0", "0", "0",
"1", "1", "0", "0", "0", "1"), species = c("Broiler", "Pig",
"Broiler", "Red fox", "Pig", "Broiler", "Wild bird", "Wild bird",
"Wild bird", "Pig", "Wild bird", "Wild bird"), fillcol = c("Broiler_1",
"Pig_0", "Broiler_0", "Red fox_0", "Pig_0", "Broiler_0", "Wild bird_1",
"Wild bird_1", "Wild bird_0", "Pig_0", "Wild bird_0", "Wild bird_1"
)), row.names = c(NA, -12L), class = c("grouped_df", "tbl_df",
"tbl", "data.frame"), vars = "gene", drop = TRUE, indices = list(
0:1, 2:3, 4:5, 6:7, 8:9, 10:11), group_sizes = c(2L, 2L,
2L, 2L, 2L, 2L), biggest_group_size = 2L, labels = structure(list(
gene = c("gyrA", "gyrB", "parC", "parE", "qnrA1", "sul3")), row.names = c(NA,
-6L), class = "data.frame", vars = "gene", drop = TRUE, indices = list(
0:1, 2:3, 4:5, 6:7, 8:9, 10:11), group_sizes = c(2L, 2L,
2L, 2L, 2L, 2L), biggest_group_size = 2L, labels = structure(list(
gene = c("gyrA", "gyrB", "parC", "parE", "qnrA1", "sul3")), row.names = c(NA,
-6L), class = "data.frame", vars = "gene", drop = TRUE)))
Plot:
library(ggplot2)
p1 <- ggplot(test_df, aes(fct_reorder(gene, type2),
factor(id),
fill = fillcol,
alpha = result)) +
geom_tile(color = "white")+
theme_minimal()+
labs(fill = NULL)+
theme(axis.text.x = element_text(angle = 90,
hjust = 1,
vjust = 0.3,
size = 7),
axis.title = element_blank(),
panel.grid = element_blank(),
legend.position = "right")+
guides(alpha = FALSE)+
coord_fixed()
Additionally, I have the following tree object:
structure(list(edge = structure(c(23L, 23L, 22L, 22L, 21L, 21L,
20L, 20L, 19L, 19L, 18L, 18L, 17L, 17L, 16L, 16L, 15L, 15L, 14L,
14L, 13L, 13L, 1L, 3L, 2L, 9L, 22L, 23L, 4L, 5L, 20L, 21L, 11L,
12L, 18L, 19L, 10L, 17L, 8L, 16L, 6L, 7L, 14L, 15L), .Dim = c(22L,
2L)), edge.length = c(2, 2, 0, 0, 2.5, 0.5, 2, 2, 0.75, 0.25,
0.5, 0.5, 2.41666666666667, 0.166666666666667, 3.0625, 0.145833333333333,
3.38888888888889, 0.326388888888889, 3, 3, 0.5, 0.111111111111111
), tip.label = c("2016-02-425-2-S", "2016-02-522-2-S", "2015-01-2166-1-S",
"2016-02-402-2-S", "2016-02-514-2-S", "2016-17-299-1-S", "2016-02-739-2-S",
"2015-01-5140-1-S", "2014-01-2992-1-S", "2014-01-7234-1-S", "2014-01-4651-1-S",
"2015-01-3004-1-S"), Nnode = 11L), class = "phylo", order = "postorder")
Which is plotted like this:
library(ggtree)
p2 <- ggtree(tree)+
geom_treescale()+
geom_tiplab(align = TRUE, linesize = 0, size = 1)+
xlim(0, 4.2)
What I want to do is to combine the tree and the first plot, and order the first plot y-axis after the order in the tree, so that they match. I have tried to use some of the solutions here, but I can't seem to produce the same plot with the facet_plot function. Is there a way to identify maching values on the y-axis on both plots, and then combine them?
This is how I want it to look (approximately):

Answer 1

We need to arrange the tile plot in the same order as the tree plot and then we need to lay the two plots out so they correspond. The first task is relatively straightforward, but I'm not sure how to do the second without some manual tweaking of the layout.
library(tidyverse)
library(ggtree)
library(grid)
library(gridExtra)
p2 <- ggtree(tree)+
geom_treescale()+
geom_tiplab(align = TRUE, linesize = 0, size = 3)+
xlim(0, 4.2)
Now that we've created the tree plot, let's get the ordering of the y axis programmatically. We can do that using ggplot_build to get the plot structure.
p2b = ggplot_build(p2)
We can look at the data for the plot layout by running p2b$data in the console. This outputs a list with the various data frames that represent the plot structure. Looking these over, we can see that the fifth and six data frames have the node labels. We'll use the fifth one (p2b$data[[5]] and order them based on the y column to get a vector of node labels (p2b$data[[5]] %>% arrange(y) %>% pull(label))). Then we'll convert test_df$id to a factor variable with this node ordering.
test_df = test_df %>%
mutate(id = factor(id, levels=p2b$data[[5]] %>% arrange(y) %>% pull(label)))
(As another option, you can get the ordering of the nodes directly from p2 with p2$data %>% filter(isTip) %>% arrange(parent) %>% pull(label))
Now we can generate the tile plot p1 with a node order that corresponds to that of the tree plot.
p1 <- ggplot(test_df, aes(fct_reorder(gene, type2),
factor(id),
fill = fillcol,
alpha = result)) +
geom_tile(color = "white")+
theme_minimal()+
labs(fill = NULL)+
theme(axis.text.x = element_text(angle = 90,
hjust = 1,
vjust = 0.3,
size = 7),
axis.title = element_blank(),
panel.grid = element_blank(),
legend.position = "right")+
guides(alpha = FALSE)+
coord_fixed()
We can see in the plot below that the labels correspond.
grid.arrange(p2, p1, ncol=2)
Now we need to lay out the two plots with only one set of labels and with the node lines matching up vertically with the tiles. I've done this with some manual tweaking below by creating a nullGrob() (basically a blank space below p1) and adjusting the heights argument to get the alignment. The layout can probably be done programmatically, but that would take some additional grob (graphical object) manipulation.
grid.arrange(p2 + theme(plot.margin=margin(0,-20,0,0)),
arrangeGrob(p1 + theme(axis.text.y=element_blank()),
nullGrob(),
heights=c(0.98,0.02)),
ncol=2)