New research introduces AstroLogics for analysing Boolean model ensembles

A new study published in Bioinformatics presents AstroLogics, a framework designed to analyse Boolean network model ensembles and identify differences in their dynamical behaviour and logical regulation. AstroLogics combines simulation-based analysis, clustering, and the comparison of logical rules to explore the diversity within ensembles of Boolean models.

Background

Boolean networks are widely used to model cellular regulatory mechanisms. They represent genes, proteins, and other biological components as binary variables, active or inactive, and describe their regulatory relationships using logical rules.

Despite their simplicity, Boolean networks can capture important features of biological systems, including attractors and robustness to noise. Their discrete nature makes them particularly useful when quantitative data are scarce, while their interpretability can facilitate biological insights.

Methods for inferring Boolean networks from experimental data and existing biological knowledge can generate multiple candidate models, creating what are known as model ensembles. These ensembles can represent cell populations and their heterogeneity. However, according to the authors, proper and efficient tools for comparing individual models within these solution spaces are still lacking.

AstroLogics was developed to address this gap by analysing Boolean networks through two complementary approaches: their dynamical properties and their logical rules.

Key points

• AstroLogics compares Boolean networks within model ensembles, identifying differences in both their dynamical behaviour and logical regulation.

• The framework calculates dynamical distances between models and clusters models that share similar behaviours.

• It identifies key logical properties associated with each cluster, highlighting regulatory structures that differentiate model behaviours.

• AstroLogics uses MaBoSS for stochastic simulations, allowing Boolean network dynamics to be explored without exhaustive analysis of the complete state transition graph.

• The framework considers both long-term and transient dynamics. Transient dynamics can distinguish models that have the same attractors but differ in the timing and order of component activation.

• Three Boolean network ensembles were used to demonstrate the framework: a central nervous system differentiation model, a haematopoiesis model, and a cancer cell invasion model.

• The framework is modular and supports different strategies for calculating model dynamics.

• AstroLogics is integrated with the CoLoMoTo environment, facilitating interoperability with existing Boolean modelling tools.

• The AstroLogics package, datasets, models, scripts, and tutorials are openly available, enabling the experiments and figures presented in the study to be reproduced.

Conclusions

AstroLogics provides a new framework for exploring Boolean model ensembles by comparing their dynamics and logical rules. By identifying groups of models with similar behaviours and the regulatory features that distinguish them, the framework can help generate new hypotheses and guide further model refinement.

This work also contributes to the computational modelling expertise underpinning CERTAINTY. The project is developing a multi-scale, patient-specific virtual twin for CAR T cell therapy in multiple myeloma, bringing together computational models across different biological scales to support personalised treatment strategies and biomedical research.

Access the Publication

Read the open-access article in Bioinformatics: https://doi.org/10.1093/bioinformatics/btag555