Center multiple pheatmap

I have two versions of the same heatmap, clustered and unclustered. The unclustered version doesn't have the dendrogram on top and when I combine them with patchwork the don't align. How can I center the second heatmap so it aligns with the first one?

library(ggplot2)
library(ggplotify)
library(pheatmap)
library(patchwork)

# dummy df
mat = data.frame(matrix(rnorm(100), nrow=20))
# pheatmap as ggplot
g1 <- as.ggplot(pheatmap(mat))
g2 <- as.ggplot(pheatmap(mat,cluster_cols = F))
# combine 
g1+g2

pheatmap patchwork

• 3,376 views • link

updated 1.8 years ago by yura.grabovska •

• written 5.7 years ago by 5utr

2 answers

I think you can use CompleHeatmap package

library(ComplexHeatmap)
mat = data.frame(matrix(rnorm(100), nrow=20))
a <- pheatmap(as.matrix(mat))
b <- pheatmap(as.matrix(mat), cluster_cols = F)
a + b

This aligns very well.

• 0 views • link

written 5.7 years ago by prabin.dm

Indeed. ComplexHeatmap renders all other heatmap packages useless by providing any features one might want. pheatmap especially is pretty obsolete, as CH was built to address its shortcomings - as the developer puts it:

The ComplexHeatmap package is inspired from the pheatmap package. You can find many arguments in ComplexHeatmap have the same names as in pheatmap. Also you can find this old package that I tried to develop by modifying pheatmap.

• 0 views • link

written 5.7 years ago by Ram

FWIW pheatmaps can be readily aligned with cowplot. But yes, in general it makes sense to move onto ComplexHeatmap. Historically it wasn't the fastest which made me stick to pheatmap for a while. But it has since overtaken it in most regards.

• 0 views • link

written 1.8 years ago by yura.grabovska •

Thank you - I was unaware of the speed difference. I loved the features and how intuitive it was. Has it gotten better than pheatmap in terms of speed as well?

• 0 views • link

written 1.8 years ago by Ram

I mean I'm talking like years ago back when I was doing my PhD. We use ComplexHeatmap a lot for plotting single-cell and CNA data so the speed is definitely much better now.

You have to factor in that before I started my degree I never touched R so there might have been an element of competence at play also.

• 0 views • link

written 1.8 years ago by yura.grabovska •

Another option: ggalign package

The usage of ggalign is simple if you're familiar with ggplot2 syntax, the typical workflow includes:

Initialize the layout using ggheatmap() or ggstack().
Customize the layout with:
- align_group(): Group layout axis into panel with a group variable.
- align_kmeans(): Group layout axis into panel by kmeans.
- align_reorder(): Reorder layout observations based on statistical weights or by manually specifying the observation index.
- align_dendro(): Reorder or Group layout based on hierarchical clustering.
Adding plots with ggalign() or ggpanel(), and then layer additional ggplot2 elements such as geoms, stats, or scales.

Here is the solution for this issue:

library(ggalign)
#> Loading required package: ggplot2

mat <- data.frame(matrix(rnorm(100), nrow = 20))
ggstack(mat, "h", sizes = unit(c(2, 1, 1), c("cm", "null", "null"))) +
    ggheatmap() +
    hmanno("t", free_spaces = "l") +
    align_dendro(aes(color = branch), k = 3L) +
    scale_color_brewer(palette = "Set2") +
    hmanno("l", size = unit(2, "cm")) +
    align_dendro(aes(color = branch), k = 4L) +
    scale_color_brewer(palette = "Set1") +
    ggheatmap() +
    hmanno("t") +
    align_dendro(aes(color = branch), k = 3L) +
    scale_color_brewer(palette = "Set2") +
    hmanno("r", size = unit(2, "cm")) +
    align_dendro(aes(color = branch), k = 4L) +
    scale_color_brewer(palette = "Set1")

enter image description here

^{Created on 2024-10-09 with [reprex v2.1.0](https://reprex.tidyverse.org)} ~

• 0 views • link

written 1.8 years ago by Yun •