Impact of individual nodes in Boolean network dynamics

TL;DR

This paper investigates how the influence of individual nodes in Boolean networks can be predicted using eigenvector centrality, with implications for understanding biological regulation and signaling.

Contribution

It demonstrates that eigenvector centrality effectively predicts a node's impact on network dynamics, especially when based on weighted activity matrices.

Findings

Eigenvector centrality predicts damage spreading in Boolean networks.

Weighted matrices improve the accuracy of impact prediction.

Results are supported by simulations and analytical linear approximations.

Abstract

Boolean networks serve as discrete models of regulation and signaling in biological cells. Identifying the key controllers of such processes is important for understanding the dynamical systems and planning further analysis. Here we quantify the dynamical impact of a node as the probability of damage spreading after switching the node's state. We find that the leading eigenvector of the adjacency matrix is a good predictor of dynamical impact in the case of long-term spreading. This so-called eigenvector centrality is also a good proxy measure of the influence a node's initial state has on the attractor the system eventually arrives at. Quality of prediction is further improved when eigenvector centrality is based on the weighted matrix of activities rather than the unweighted adjacency matrix. Simulations are performed with ensembles of random Boolean networks and a Boolean model of signaling in fibroblasts. The findings are supported by analytic arguments from a linear approximation of damage spreading.