Using heatmaply dendrogram sorting in pheatmap

Question

Is there a way of using the dendrogram clustering method used in the heatmaply package and to apply it to a heatmap produced by the pheatmap package? Basically the opposite to what was asked here: Clustering in pheatmap and heatmaply R packages

I usually see better clustering of my data when I use heatmaply over pheatmap. However, I do not always want to use an interactive plot. The heatmaply::ggheatmap function does not work properly for me since I have have col_side_colors and the annotation gets in the way of the dendrogram. It just looks very messy. So I use pheatmap instead.

Maybe someone can help with my question. Thanks!

Answer 1

library(pheatmap)
library(heatmaply)
library(seriation)
library(dendextend)

# A dataset    
x <- scale(mtcars)

# Interactive heatmap
p <- heatmaply(x)
print(p)

# The dendrogram for rows
dst <- dist(x)
hc_row <- hclust(dst)
row_dend <- as.dendrogram(hc_row)
row_dend <- seriate_dendrogram(row_dend, dst, method="OLO")

# The dendrogram for columns
dst <- dist(t(x))
hc_row <- hclust(dst)
col_dend <- as.dendrogram(hc_row)
col_dend <- seriate_dendrogram(col_dend, dst, method="OLO")
col_dend <- rotate(col_dend, order = rev(labels(dst)[get_order(as.hclust(col_dend))]))

# The pheatmap with the same clustering of heatmaply
pheatmap(x, cluster_rows=as.hclust(row_dend), cluster_cols=as.hclust(col_dend))

The output of heatmaply

and the output of pheatmap