[Paper Review] Synthesis of Boolean Networks from Biological Dynamical Constraints using Answer-Set Programming

Stéphanie Chevalier, Froidevaux, Christine+2 more|arXiv (Cornell University)|Sep 10, 2019

Gene Regulatory Network Analysis15 references4 citations

TL;DR

This paper presents an Answer-Set Programming (ASP)-based method to synthesize Boolean networks from biological dynamical constraints, including reachability and attractor properties, enabling complete, non-redundant enumeration of valid models. It achieves scalability up to 1,000-node networks and reduces a search space of over 226 million candidate models to just 3,735 consistent solutions when combining positive and negative reachability with trap space constraints in a CNS development case study.

ABSTRACT

Boolean networks model finite discrete dynamical systems with complex behaviours. The state of each component is determined by a Boolean function of the state of (a subset of) the components of the network. This paper addresses the synthesis of these Boolean functions from constraints on their domain and emerging dynamical properties of the resulting network. The dynamical properties relate to the existence and absence of trajectories between partially observed configurations, and to the stable behaviours (fixpoints and cyclic attractors). The synthesis is expressed as a Boolean satisfiability problem relying on Answer-Set Programming with a parametrized complexity, and leads to a complete non-redundant characterization of the set of solutions. Considered constraints are particularly suited to address the synthesis of models of cellular differentiation processes, as illustrated on a case study. The scalability of the approach is demonstrated on random networks with scale-free structures up to 100 to 1,000 nodes depending on the type of constraints.

Motivation & Objective

• To address the challenge of model synthesis in systems biology where prior knowledge is incomplete and dynamic behaviors must be captured.
• To provide a complete, non-redundant characterization of all Boolean networks satisfying specified dynamical constraints derived from partial observations.
• To enable the synthesis of models for cellular differentiation processes using reachability and attractor constraints.
• To overcome limitations of heuristic or incomplete search methods in existing approaches.
• To support biological hypothesis testing by quantifying data informativeness and detecting inconsistencies.

Proposed method

• The method encodes the Boolean network synthesis problem as a Boolean satisfiability problem using Answer-Set Programming (ASP).
• It models the network using a prior knowledge influence graph that restricts the domain of each node's Boolean function.
• Dynamical constraints—positive reachability, negative reachability, trap spaces, and fixpoints—are formally encoded as ASP rules.
• The approach uses the most permissive semantics of Boolean networks to improve scalability and model biological plausibility.
• Constraints are checked via stable model generation, ensuring exhaustive and non-redundant solution enumeration.
• The encoding is parametrized to support varying network sizes and constraint types, with scalability demonstrated on scale-free networks.

Experimental results

Research questions

• RQ1 Can Answer-Set Programming be used to synthesize Boolean networks that satisfy complex dynamical constraints derived from partial biological observations?
• RQ2 How can reachability and attractor properties be encoded in ASP to ensure complete and non-redundant solution enumeration?
• RQ3 What is the scalability of the method for large-scale biological networks, particularly those with complex dynamics like cellular differentiation?
• RQ4 How do combinations of constraints (e.g., reachability and trap spaces) reduce the solution space while preserving biological relevance?
• RQ5 To what extent can this method detect inconsistencies in biological hypotheses or quantify the informativeness of experimental data?

Key findings

• The method successfully reduces the solution space from over 226 million candidate Boolean networks to just 3,735 consistent models when combining positive reachability, negative reachability, and trap space constraints in a CNS development model.
• The approach achieves scalability on random scale-free networks with up to 1,000 nodes, depending on the constraint type.
• The inclusion of negative reachability constraints significantly reduces the solution space, from 224 million (NR only) to 16 million (PR + NR).
• Combining positive reachability with trap space constraints reduces the solution set to 8,964 models, demonstrating strong filtering power.
• The method enables exhaustive analysis of model variability, supporting the identification of robust regulatory motifs across consistent models.
• The approach detects inconsistencies in hypotheses: for example, the combination of PR + NR + FP yields only 1,120 solutions, indicating high data informativeness and model precision.