Question

Error: node stack overflow when trying to plot large dataset using Complexheatmap

0

Entering edit mode

AB ▴ 110

@ab-8975

Last seen 2.2 years ago

United States

Hi,

I have a massive dataset that I want to use to make a heatmap using the Complexheatmap package. However, I'm getting the following error

Error: node stack overflow
Error during wrapup: node stack overflow

I am not at liberty to share my original dataset and hence here is a sample data.

new_df=data.frame("Spock2"=rep(0,10),"Col4a3"=c(rep(0,3),rep(2,2),rep(0,5)),"Syt1"=c(rep(0,8),rep(4,2)),"Pacrg"=c(rep(0,4),4,rep(0,2),rep(3,3)))

new_df=new_df[rep(seq_len(nrow(new_df)), each=300),]
Heatmap(as.matrix(new_df), row_labels = rep("",3000))

How do I fix this issue ? Any ideas ?

Thanks, Apoorva

Complexheatmap • 5.2k views

ADD COMMENT • link updated 5.2 years ago by Zuguang Gu ▴ 290 • written 5.2 years ago by AB ▴ 110

0

Entering edit mode

It is because you have too many rows which are identical. The internal function ComplexHeatmap:::grid.dendrogram() that draws the dendrogram can handle a matrix with a few identical rows, but it gives errors with too many of that. I will look at this and try to fix it soon.

ADD REPLY • link 5.2 years ago Zuguang Gu ▴ 290

0

Entering edit mode

I find a similar case when making heatmaps with too many identical rows with gplots package. https://stackoverflow.com/questions/16559250/error-in-heatmap-2-gplots

ADD REPLY • link 5.2 years ago Zuguang Gu ▴ 290

score 0 · Answer 1 · 2019-09-27

I cannot solve this problem, but you can add small random values to the original matrix to make no duplicated rows:

m = as.matrix(new_df)
m = m + runif(length(m), min = 0, max = 1e-10)
Heatmap(m, show_row_names = FALSE)

To control the legend, set proper values in heatmap_legend_param argument.

https://jokergoo.github.io/ComplexHeatmap-reference/book/legends.html#heatmap-and-annotation-legends

score 0 · Answer 2 · 2019-09-28

0

Entering edit mode

Zuguang Gu ▴ 290

@zuguang-gu-7797

Last seen 25 days ago

Germany / Heidelberg / DKFZ

Now I add a new argument jitter in Heatmap() which internally adds random shift to the original matrix. The matrix is only modified when making the dendrograms. The heatmap bodies and the legends are not affected:

m = as.matrix(new_df)
Heatmap(m, show_row_names = FALSE, jitter = TRUE)

or

m = as.matrix(new_df)
Heatmap(m, show_row_names = FALSE, jitter = 1e-10)

The random values are generated from the uniform distribution within (0, jitter) or (0, 1e-10).

Please update the package from GitHub:

library(devtools)
install_github("jokergoo/ComplexHeatmap")

ADD COMMENT • link 5.2 years ago Zuguang Gu ▴ 290

0

Entering edit mode

Hi. I had the same problem, I used the jitter argument now, but the error persists.

ADD REPLY • link 5.0 years ago yermatudela • 0

0

Entering edit mode

Indeed, I dont have any repeated value i.e.:

head(sort(table(HTdt$TotPC_FC), decreasing=TRUE))

-2.7531024129286 -2.60239600129202 -2.59882399747114 -2.50057728006338 -2.43532405332128 1 1 1 1 1 -2.35119425790377 1

head(sort(table(HTdt$SynPC_FC), decreasing=TRUE))

-12.8575440204269 -11.1288535950429 -8.36119814307584 -6.06653571829438 -5.52828129172012 1 1 1 1 1 -4.07528353549578 1

head(sort(table(HTdt$DegcrFC), decreasing=TRUE))

-8.74608567657745 -8.45771966702432 -7.94509403518839 -7.85642773252236 -7.42725845984559 1 1 1 1 1 -7.2987419615091 1

What could be causing the error then?

ADD REPLY • link 5.0 years ago yermatudela • 0