Pyscenic vs scenic GENIE3[results]

[JensFGG]

With R I had many TF and targets with importance levels above 200 when running the Genie3 step

However, I when running almost the same dataset with pyscenic using the code below, my biggest importance level only reaches 0.08. Any idea what could trigger this difference? And what does the value in importance mean? is it relative? if so, can it be that I have too much noise in the data because in python I did not run a filter on the cells in contrast to R.

arboreto_with_multiprocessing.py
expr_mat.loom
allTFs_hg38.txt
--method genie3
--output adj.tsv
--num_workers 32
--seed 777

[SeppeDeWinter]

Hi Jens

We indeed mostly use these values relatively (e.g. by taking the top n target genes per TF based on the importance score), their absolute value alone is not that informative.

The noise in the data could affect the absolute values of the importance scores.

Does the ranking (e.g. top 500 genes for a particular TF) differ a lot between R and python?

Best,

Seppe