Scanpy: Not able to color groups on pca plot

[sonyasofia7]

I have an adata object and multiple CD4+ and CD8+ cells as observations and gene expression as variables. I want to draw a PCA with each dot corresponding to one cell and colored by cell type (e.g. all CD4 cells in blue, CD8 in orange). Each cell have a unique identifier, and I have a list cd4, which contains identifiers of only CD4 cells, and the same for CD8.

I tried the following:
scn.pl.pca(adata_needed, groups={"cd4":cd4, "cd8":cd8}, palette = ["red", "blue"])
But the result is the PCA plot in grey

I also tried to visualize only one sample:
scn.pl.pca(adata_needed, groups=["SRR1551000"])

But got exactly the same result.

I decided to use "color" parameter in pca plot:
scn.pl.pca(adata_needed, color="SRR1551000")
But got the following mistake:

ValueError: key 'SRR1551000' is invalid! pass valid observation annotation, one of [] or a gene name Index(['ENSG....', 'ENSG...', )

The observation names are there (adata_needed.obs_names outputs the observation I try to give to the function)

Here's the plot I need, but drawn with pandas.

How can I draw a PCA plot like this?

[LuckyMD]

Hi,

To make a pca plot you need the functions scn.tl.pca(adata_needed) and scn.pl.pca(adata_needed). If you would like to visualize your CD4+ and CD8+ cells on this plot, you need to store a variable in adata_needed.obs which contains the information which cell is CD4+ and which cell is CD8+. For example:

adata_needed.obs['cell_type'] = my_list_with_cd_labels
scn.tl.pca(adata_needed)
scn.pl.pac(adata_needed, color='cell_type')

Here the variable my_list_with_cd_labels should look someting like this:
my_list_with_cd_labels = ['CD8+', 'CD4+', 'CD4+', 'CD4+', 'CD4+', 'CD8+', 'CD8+', ..., 'CD8+', 'CD4+']
Where the length of the list is equal to the number of cells you have in adata_needed and the order is the same as well.

It may be helpful to consult this tutorial as well: https://nbviewer.jupyter.org/github/theislab/scanpy_usage/blob/master/170505_seurat/seurat.ipynb

I hope that helps!

[sonyasofia7]

Thank you!
It was just not clear from the tutorial in the beginning that the same thing is used to define colors in plots :(