Investigations of Attractor Behavior over the Decay of Modular RBNs

TL;DR

This study evaluates when simplified modular RBNs can reliably approximate more interconnected RBNs, finding static and dynamic modularity measures perform similarly in small networks as interconnectivity increases.

Contribution

It demonstrates that static modularity measures can be as effective as dynamic measures in predicting RBN behavior in small networks, highlighting limitations of static measures as networks become more interconnected.

Findings

Static and dynamic modularity measures perform similarly in small networks.

Both measures' effectiveness decline as network interconnectivity increases.

The Newman 2004 static modularity measure remains reliable in certain conditions.

Abstract

When is it safe to approximate a complicated random Boolean network (RBN) as a simplified, easier to model RBN? When can static measures of network structure be reliably used to infer the network's dynamics? This simple experiment tests the ability of disjoint modular RBNs to approximate the dynamics of progressively more interconnected RBNs, while characterizing the performance of both static and dynamic measures of modularity as both break down. We find that, at least in the small networks investigated, the Newman 2004 [1] measure of static modularity performs as well as a more complex dynamic measure of modularity, and that the progressively increasing failure of one tracks that of the other. The dynamic measure is based on the Hamming distance of attractor schemata in rewired networks from those in perfectly modular networks. This result holds for a range of p-values.