Question

Making a 3d tSNE Plot in Seurat for scRNA-seq

2

Entering edit mode

6.7 years ago

asyndeton17 ▴ 40

Hi,

Is there an easy way to make a 3D tSNE plot of the single cell clusters in Seurat?

Thanks

RNA-Seq scRNA-seq Seurat tSNE • 10k views

ADD COMMENT • link updated 4.9 years ago by DrAcula ▴ 140 • written 6.7 years ago by asyndeton17 ▴ 40

0

Entering edit mode

Is there a way to do so with another package?

ADD REPLY • link 6.6 years ago by asyndeton17 ▴ 40

WouterDeCoster · Answer 1 · 2017-10-27

I tried solution from halo22, it didn't work. However, I somehow figured it out myself.

#3d tsne projection
yourseuratobject <- RunTSNE(object = yourseuratobject, dims.use = 1:n, do.fast = TRUE, dim.embed = 3)

tsne_1 <- yourseuratobject@dr$tsne@cell.embeddings[,1]

tsne_2 <- yourseuratobject@dr$tsne@cell.embeddings[,2]

tsne_3 <- yourseuratobject@dr$tsne@cell.embeddings[,3]

library(scatterplot3d)

scatterplot3d(x = tsne_1, y = tsne_2, z = tsne_3, color = as.numeric(1:n)[yourseuratobject@ident])

library(rgl) #interactive 3d plotting

plot3d(x = tsne_1, y = tsne_2, z = tsne_3, col = as.numeric(1:n)[yourseuratobject@ident], type="s",radius=0.3)

rgl::rglwidget() #save as html

PS: tsne coordinates are stored in yourseuratobject@dr$tsne@cell.embeddings cluster identity by SNN is store in yourseuratobject@ident

ref. https://github.com/jkrijthe/Rtsne/issues/12

Hope this would help!

Gary

score 1 · Answer 2 · 2017-09-21

1

Entering edit mode

6.6 years ago

halo22 ▴ 300

You can use the R-package tsne to do this https://cran.r-project.org/web/packages/tsne/tsne.pdf

For your analysis you will have to figure out what input to provide to tSNE, ideally PC's obtained from seurat are the suggested input but you can also use your variable genes with scaled values.

ADD COMMENT • link 6.6 years ago by halo22 ▴ 300

2

Entering edit mode

A bit late by now, but I think the R package Rtsne is preferable over tsne since it is faster being a wrapper over C++ code.

ADD REPLY • link 6.5 years ago by dariober 14k

score 0 · Answer 3 · 2019-05-27

I know this is old now. But Ive written some code which will help you create a 3D expression plot using plotly out of a seurat v3.0.0 object. Dont forget to star and fork :)

Find the complete solution here: https://github.com/Dragonmasterx87/3D-Plotting-in-Seurat-3.0.0

Update: The code now supports UMAP 3D clustering as well :)