ExaStoLog

ExaStoLog computes exact stationary probability distributions and performs parameter sensitivity analysis for stochastic asynchronous continuous time Boolean models to quantify how attractor probabilities depend on timescale parameters (transition rates).

Key Features:

• Exact calculation method: Computes stationary probability values for attractors without Monte Carlo simulation using deterministic matrix-based techniques.
• Graph-theoretical techniques: Applies graph theoretical methods previously used in chemical kinetics to analyze state transition structures.
• Matrix calculation techniques: Uses kinetic matrices and linear algebra to derive stationary solutions.
• Continuous time Markov chain formulation: Represents logical model states as a continuous time Markov chain.
• Topological sorting of state transition graphs: Employs topological sorting of the state transition graph to structure computations.
• Nullspace–kinetic matrix dependency analysis: Exploits dependencies between nullspaces and kinetic matrices to obtain stationary distributions.
• Master equation derivation: Derives stationary solutions directly from the master equation.
• Parameter sensitivity analysis: Quantifies the effect of changes in timescale parameters (transition rates) on stationary solutions.
• Identification of robust and sensitive processes: Detects which processes and parameters predominantly influence attractor probabilities.
• Avoids Monte Carlo uncertainty: Provides exact results to address uncertainties inherent in Monte Carlo estimates and convergence to asymptotic solutions.
• Scalability for intermediate models: Applies efficiently to intermediate-size models (e.g., up to ~23 nodes) using a matrix-based approach.

Scientific Applications:

• Precise attractor probability estimation: Determines exact stationary probabilities for attractors in stochastic asynchronous continuous time Boolean models.
• Timescale parameter analysis: Explores how transition rates affect stationary solutions and model behavior.
• Sensitivity-driven hypothesis generation: Identifies parameters whose variation produces robust or sensitive changes in attractor probabilities for targeted experimental or modeling follow-up.
• Methodological validation: Assesses the validity and convergence of Monte Carlo-based estimates by providing exact benchmarks.
• Biological interpretation of dynamics: Facilitates interpretation of model dynamics by linking kinetic parameters to long-term probabilistic outcomes.

Methodology:

Defines states as a continuous time Markov chain and uses graph-theoretical and matrix calculation techniques from chemical kinetics—including topological sorting of the state transition graph, analysis of nullspaces and kinetic matrices, and direct derivation from the master equation—to compute stationary solutions.