Different versions of RegulonDB regulatory network - which to use (and how)?

Question

[v.baca]

Hi there, I do a little project where I analyze regulatory networks motifs (such as feedforward/feedback loops etc.). Specifically, I explore attractor landscape of Boolean networks representing these motifs. What I would like is take some real biological appearances of these motifs in real regulatory networks, with some context (e.g. all 1-edge neighbours of motif nodes) and see how the motif behaves in these contexts, compared to isolated behavior.

I looked at RegulonDB. There are several different versions of the transcription regulatory network: TF-Gene, TF-Operon, TF-TF. My 1st question: What's difference between these? Which one would you suggest is most relevant to use here? The explanation on RegulonDB didn't help me much.

As I looked into the network files, the regulations were usually edges from TFs to genes, while the genes/TFs usually appear in both roles - the same gene/TF in different roles is differentiated by small/big first letter. That leads to bipartite graph in Cytoscape. This certainly makes sense, but in my use-case (that is, representation as a Boolean network), I don't want to distinguish between gene and its protein functioning as TF - I just want the network to represent the fact that gene A positively/negatively regulates gene B. I therefore solved this using "toLower()" on all the genes/TFs identifiers, merging genes and TFs together. 2nd question: Is this merging "correct" (does the resulting data make sense)? Do you know of any better data source for my use-case?

Any answers, ideas, help will be appreciated :-)

Answer 1

[Marcel Ribeiro-Dantas]

Regarding your first question, this is likely what relationships were statistically tested. TF-gene (or TF-centric) regulatory networks mean that only the statistical association between TFs and genes has been tested, not between genes and genes. Maybe gene foo is a Transcription Factor, you are not aware of it, and the relationship between it and gene bar, even if it exists, won't lead to an edge because it wasn't tested. By taking the assumption of a subset of genes being TF, you decrease the likelihood of false positives, but you can incur false negatives.

I'm not sure about the second question. If you want to know how A regulates B, and you removed the distinction between TFs and targets, you won't know if it was A or B that regulated positively/negatively the other. It's not clear to me why you had to remove this distinction.