I am using ARACNE-AP via Docker to generate networks from my gene matrix. As an input, I am using a text file, with genes on rows and samples on columns and a gene list( the ones that I am mostly interested in whether to learn if they create hubs or not).
ARACNE-AP returns a network.txt file where the first column is regulators the second is target the third is mutual information score and the fourth is the p-value. According to what I read & understood, with MI we create a rank, and then according to this threshold, we decide whether it is a direct interaction or not. However, there are some things that I still can't wrap my head around.
- How does that threshold is decided? (sample number affect this but I can't understand how)
- According to that threshold how does ARACNE-AP generates p-values?
For Viper, the network generated from ARACNE is used as an input and the format is in .adj. However, ARACNE-AP doesn't support this anymore. If anyone used ARACNE and Viper recently I would like to ask:
- Would changing the extension manually from .txt to .adj cause problems?
Viper also accepts 3 column .txt file but what surprised me here is that my gene expression matrix has much more genes (rows) than my network file generated from ARACNE-AP which is probably because not all of them have pairwise direct interactions. For this part I would like to ask:
- Would subsetting the gene matrix according to the network file is reasonable?
I guess this a bit more general question:
- Which network file from which ARACNE could be given as an input for Viper?