Stochastic parameter search for events.

Area of Science:

• Computational Biology
• Biochemical Systems Modeling
• Systems Biology

Background:

• High-performance computing (HPC) enables complex stochastic models for biochemical systems.
• Predicting parameter configurations for significant events is crucial but computationally intensive.
• Brute force search is infeasible for large models due to computational demands and parameter space dimensionality.

Purpose of the Study:

• To develop an automated and efficient parameter estimation algorithm for stochastic models.
• To enable prediction of parameter configurations that lead to specific events of interest.
• To overcome the limitations of brute force and existing algorithms for rare event analysis.

Main Methods:

• Developed Stochastic Parameter Search for Events (SParSE), a novel, automated, and parallelizable parameter estimation algorithm.
• SParSE updates all reaction rate parameters concurrently, with complexity independent of the number of parameters.
• Applied SParSE to birth-death, reversible isomerization, and SIRS disease transmission models.

Main Results:

• SParSE significantly accelerates the computation of the parametric solution hyperplane compared to uniform random search.
• The algorithm successfully computes biasing parameters for rare events not amenable to current importance sampling methods.
• Demonstrated SParSE's effectiveness across systems of increasing complexity.

Conclusions:

• SParSE offers a novel, efficient, and event-oriented parameter estimation method for stochastic systems obeying the Chemical Master Equation (CME).
• The algorithm's usability and efficiency are maintained for large systems, independent of the number of unknown parameters.
• SParSE is readily applicable to diverse stochastic systems for predicting significant events.