mpbn: a simple tool for efficient edition and analysis of elementary properties of Boolean networks

TL;DR

The paper addresses the need for scalable, interactive analysis of Boolean networks (BNs) defined by $f: \\{0,1\\}^n \\to \\{0,1\\}^n$, enabling efficient exploration of elementary dynamics under the Most Permissive update mode. It introduces mpbn, a Python-based tool leveraging Answer-Set Programming to enumerate fixed points and both minimal and maximal trap spaces and to perform MP reachability checks. Key contributions include broad format support (BooleanNet, biolqm), flexible model edition, DNF/BDD representations, MP dynamics, and large-scale benchmarks showing mpbn outperforms existing tools on big networks. The work advances BN analysis for systems biology by offering an open, teaching-friendly toolbox capable of guiding control strategies and mechanistic discovery.

Abstract

The tool mpbn offers a Python programming interface for an easy interactive editing of Boolean networks and the efficient computation of elementary properties of their dynamics, including fixed points, trap spaces, and reachability properties under the Most Permissive update mode. Relying on Answer-Set Programming logical framework, we show that mpbn is scalable to models with several thousands of nodes and is one of the best-performing tool for computing minimal and maximal trap spaces of Boolean networks, a key feature for understanding and controling their stable behaviors. The tool is available at https://github.com/bnediction/mpbn.

Figures (7)

• Figure 1: Example of usage within a Jupyter notebook
• Figure 2: Cumulative minimal trap space experiments completed (y-axis) until a specific time point (x-axis, logarithmic). Concerns the 235 real-world models.
• Figure 3: Cumulative minimal trap space experiments completed (y-axis) until a specific time point (x-axis, logarithmic). Concerns the 128 randomly generated models.
• Figure 4: Cumulative maximal trap space experiments completed (y-axis) until a specific time point (x-axis, logarithmic). Concerns the 235 real-world models.
• Figure 5: Cumulative maximal trap space experiments completed (y-axis) until a specific time point (x-axis, logarithmic). Concerns the 128 randomly generated models.