Moment estimation for chemically reacting systems by extended Kalman filtering

Area of Science:

• Computational chemistry
• Biochemical systems analysis
• Stochastic modeling

Background:

• Chemical reaction dynamics often rely on moment closure, which can be imprecise for systems with low molecule counts.
• Stochastic simulation offers exactness but is computationally expensive, especially for rare events or stiff systems.

Purpose of the Study:

• To develop a novel method for estimating moments in chemically reacting systems.
• To improve the accuracy and computational efficiency of analyzing stochastic chemical kinetics.
• To enable estimation of unmeasured species moments from experimental data.

Main Methods:

• Closing moment dynamics by replacing higher-order moments with estimates from limited stochastic simulations.
• Integrating the closed moment dynamics into an extended Kalman filter.
• Using simulation outputs as system states within the Kalman filter framework.

Main Results:

• The proposed method provides accurate moment estimates for chemically reacting systems.
• It offers a more computationally efficient alternative to traditional moment closure and pure stochastic simulation.
• Demonstrated utility in estimating unmeasurable species moments from observable ones in experimental settings.

Conclusions:

• The novel hybrid approach effectively addresses limitations of existing methods for stochastic chemical kinetics.
• This technique enhances the analysis of complex reacting systems, particularly those with low molecule counts or rare events.
• The method has practical implications for both theoretical modeling and experimental data interpretation.