Plot the expression of the common gene modules

[nadinegheziel]

Hello Vivian,

Thank you very much for this package.

I have run scINSIGHT on my own single-cell data of two conditions : control and PS
Everything looked fine all along the process except maybe that run_scINSIGHT took me a long time (but nevermind, the computer was clearly not adapted to fit)

I am now trying to plot the expression of the common gene modules exactly like you did in Fig 4 b and c of your paper but without any success, I really don't understand how to do it

I understood that I am supposed to use sim.scobj@V containing the modules expression and scale the color of the plot by it, but I am going nowhere. Would it be possible to have some help or some tutorial to reproduce ?
Thank you very much

Kindly
Nadine

print(sim.scobj@parameters)`

$K_j
[1] 2

$lda
[1] 0.01

$stability
      K_5       K_7       K_9      K_11      K_13      K_15 
0.8675092 0.8779438 0.9448068 0.9598742 0.9554114 0.9703640 

$K
[1] 13

$specificity
lda_0.001  lda_0.01   lda_0.1     lda_1    lda_10 
 1.236756  1.258037  1.207933  1.222736  1.235701 

[Vivianstats]

Hello Nadine,

obj$V gives the coefficients of the genes in the common gene modules, and obj$norm.W_2 gives the normalized expression levels of these gene modules. We made the figure using the normalized expression levels.

Hope this helps,
Vivian

[JahanRahman]

Hello,

How would you recommend that we visualize the condition-specific modules? There's of course the normalized matrix for the shared modules (obj$norm.W_2) but not the same for W_1. Should we normalize W_1 somehow first (ie using L2 norm or something else), or can we visualize using that matrix directly?

[Vivianstats]

Hello, the normalization is performed on W_2 to improve the clustering analysis across multiple samples, due to the presence of potential batch effects. For condition-specific modules, they are not comparable across conditions and need to be visualized for each condition separately. Therefore, normalization should not be necessary.

[JahanRahman]

that makes sense. Thank you for clarifying!

[JahanRahman]

Sorry just one follow up question. If the condition specific modules are meant to be visualized separately, how can we tell which module corresponds to which condition?

[Vivianstats]

I believe the modules are returned as a named list. Each element in the list corresponds to modules in one condition.

[JahanRahman]

ahh okay. If I'm understanding correctly, when we specify the K_j parameter, it's specifying how many condition specific modules per condition?