Differences between R and python versions

[jk86754]

Differences between R and python versions

[bramvds]

Dear Joe,

Thank you for your questions.

Regarding your first question, pySCENIC adheres to the same maxim "one TF - one regulon". Maybe you are referring to the results you get after running the command line version of pyscenic (or using the function pyscenic.prune2df)? This table gives you multiple enriched motifs (and their associated target genes) for a transcription factor. You need to apply the function transform.df2regulons to combine them into regulons. You can also use the command line version setting --output_type to json. Hope this helps.

This code snippet might also help to transform the CSV output to regulon objects in python.

import os
from pyscenic.utils import load_motifs
from pyscenic.transform import df2regulons

RESOURCES_FOLDER = "/resources/zeisel_et_al/"
MOTIFS_FNAME = os.path.join(RESOURCES_FOLDER, "zeisel_et_al.motifs.csv")
motifs = load_motifs(MOTIFS_FNAME)
regulons = df2regulons(motifs)

For the second question, I kindly request you to best this question on the arboreto package (https://github.com/tmoerman/arboreto). pySCENIC relies on this package for GRNBoost2. GENIE3 and GRNBoost2 both implement a tree-based regression per target gene approach. GENIE3 relies on Random Forests while GRNBoost2 uses Gradient Boosted Trees. For more information on the 'variable importance measure' I kindly refer you to the eponymous section in: Huynh-Thu, V. A., Irrthum, A., Wehenkel, L. & Geurts, P. Inferring regulatory networks from expression data using tree-based methods. PLoS ONE 5, (2010). The manuscript that covers GRNBoost2 has recently been published in Bioinformatics: Moerman, T. et al. GRNBoost2 and Arboreto: efficient and scalable inference of gene regulatory networks. Bioinformatics (2018). doi:10.1093/bioinformatics/bty916

Kindest regards,
Bram