What is the effect of running pyscenic on the results of cell sampling of the sample？

[jklupup]

I tried to delete only one cell, but the results of the first step of GRN generation adjacencie.csv varied dramatically.
The final coincidence rate of TF-Target was only 56.21% between the complete cell expression profile and the cell expression profile with only one cell deleted.
I don't think it's a bit credible, but I've tested a lot of data sets with this result. The test process uses the same code and reference comment files.