scv.pl.scatter function

[huizizhang949]

Hello,

I want to plot the (filtered and normalized) phase portrait for some genes in particular using the raw plt.plot function. For example,

plt.plot(adata.layers['spliced'].todense()[:,adata.var_names=='Igfbpl1'],
adata.layers['unspliced'].todense()[:,adata.var_names=='Igfbpl1'],'o')

But the phase portrait is different from the one produced by

scv.pl.scatter(adata, basis='Igfbpl1', frameon=True)

The one shown by scv.pl.scatter seems to be more close to the desired 'almond' shape and x limits and y limits are also different. Is there any step in scv.pl.scatter that transforms or normalizes the values in adata.layers['spliced']?

Thank you!

[WeilerP]

@Huizi949-z, scvelo.plotting.scatter plots the smoothed impression by default, i.e., Mu and Ms. Otherwise, the data is too noisy to robustly infer rates.

[huizizhang949]

Thanks for the reply!!

Could you please tell me how Mu and Ms are computed? Where can I extract these values from adata? I look into the source code for the function here

scvelo/scvelo/plotting/scatter.py

Line 20 in 1805ab4

def scatter(

but I could not find where and how Mu and Ms get involved.

Thank you!

[WeilerP]

@Huizi949-z, please have a look at the paper, tutorials, and docs. Everything should be explained there.
First order moments (Mu and Ms) are calculated by smoothing a cell's gene expression over its neighbors. This is done when calling scvelo.preprocessing.moments(). Consequently, you'll find the corresponding entries in adata.layers.