State and parameter estimation from exact partial state observation in stochastic reaction networks.

Area of Science:

• Computational Biology
• Chemical Kinetics
• Stochastic Processes

Background:

• Chemical reaction networks are often modeled using continuous-time Markov processes.
• Accurate estimation of species' copy numbers is crucial for understanding cellular dynamics.
• Existing methods face challenges in handling partially observed systems.

Purpose of the Study:

• To develop a novel particle filter for state and parameter estimation in discrete-state, continuous-time Markov process models of chemical reaction networks.
• To address the challenge of estimating unobserved species when only partial observations are available.
• To adapt the method for Bayesian parameter estimation and retrospective state estimation.

Main Methods:

• A novel particle filter method is proposed for state and parameter estimation.
• The conditional probability distribution of unobserved states is shown to satisfy differential equations with jumps.
• A weighted Monte Carlo simulation is used to approximate the distribution of unobserved species.

Main Results:

• The developed particle filter accurately estimates the state and parameters of chemical reaction networks.
• The method effectively computes the conditional probability distribution of unobserved species.
• The algorithm is adaptable for Bayesian parameter estimation and past state value estimation.

Conclusions:

• The proposed particle filter offers a robust approach for analyzing partially observed chemical reaction networks.
• This method advances the field of stochastic modeling in systems biology.
• The adaptability of the algorithm provides a versatile tool for various estimation tasks.