Question

Change point size and remove outline in scater's plotPCA

0

Entering edit mode

CodeAway ▴ 70

@codeaway-12991

Last seen 3.5 years ago

Hello All,

Just need a quick help, with regards to the plotPCA function of the "really excellent" scater package, that I am using for single-cell RNAseq analysis of 10X data. I wish to change the size of the points and also remove the gray outlines that currently appear around the points.

I tried using "scale_size_manual" and "scale_shape_manual", but these did not have any effect.

I would really appreciate any help.

Many thanks.

scater rnaseq single-cell • 4.8k views

ADD COMMENT • link updated 6.0 years ago by angela.ubn • 0 • written 7.0 years ago by CodeAway ▴ 70

score 0 · Answer 1 · 2017-05-09

0

Entering edit mode

James W. MacDonald 65k

@james-w-macdonald-5106

Last seen 4 hours ago

United States

Rather than trying to hack things by calling ggplot2 functions directly, you might just read the help page for plotPCA, which says

size_by: character string defining the column of 'pData(object)' to be
          used as a factor by which to define the size of points in the
          plot. Alternatively, a data frame with one column containing
          values to map to sizes.

You don't show your code, but presumably there is something in the help page that affects whether or not a gray outline should be drawn around your plotting symbols.

ADD COMMENT • link 7.0 years ago James W. MacDonald 65k

0

Entering edit mode

Thanks, James. I am actually not using "size_by", so that is left to default.

I was able to change the colors of the points as below:

ppcah <- plotPCA(sce, pca_data_input="pdata", colour_by="sample_type", shape_by=NULL) + 
  scale_fill_manual(values=c("WT"="blue", "R172K"="red")) +
  labs(fill="Type", title="PCA plot of QC metrics")

using scale_fill_manual. But if I use scale_shape_manual or scale_size_manual, nothing changes.

Thanks in advance, for any responses.

ADD REPLY • link 7.0 years ago CodeAway ▴ 70

0

Entering edit mode

For what it's worth, I usually just pull out the coordinates from the object returned by plotPCA and replot those in a separate call, rather than trying to wrestle with ggplot2 and scater at the same time. In my opinion, scater is intended for quick production of decent-looking plots for data exploration, usually for use in relatively informal analysis reports that get shown to colleagues or collaborators. If you want customized publication-quality figures, you'll have to do a bit more work.

ADD REPLY • link 7.0 years ago Aaron Lun ★ 28k

score 0 · Answer 2 · 2018-04-10

If you need to change the size and shape, you probably need to convert the associated attributes as a factor.

e.g.

New_sceset$sc3_7_clusters <- as.factor(New_sceset$sc3_7_clusters)

New_sceset$cell_type <- as.factor(New_sceset$cell_type)

p <- plotPCA(New_sceset, colour_by = 'sc3_7_clusters', shape_by = 'cell_type')

p + scale_color_manual(values = rainbow(7)) + scale_shape_manual(values = c(16, 17))

It works for me. :D