State- versus Reaction-Based Information Processing in Biochemical Networks.

Area of Science:

• Biophysics
• Biochemical Systems Analysis
• Information Theory

Background:

• Trajectory mutual information quantifies information transfer in biochemical systems.
• The linear-noise approximation (LNA) with Gaussian channel theory provides tractable solutions for mutual information, typically accurate for large systems.
• Discrepancies exist between LNA and exact Markov jump process formalisms, even in large systems.

Purpose of the Study:

• To explain the qualitative differences in mutual information calculations between LNA and Markov jump processes.
• To introduce and differentiate between reaction-based and state-based trajectory descriptions.
• To propose a reaction-based Gaussian mutual information to prevent information loss.

Main Methods:

• Analysis of trajectory mutual information using linear-noise approximation (LNA) and Gaussian channel theory.
• Comparison with exact Markov jump process formalisms.
• Development and application of reaction-based versus state-based trajectory definitions within the Gaussian framework.

Main Results:

• Differences arise from reaction-based versus state-based trajectory descriptions.
• State-based trajectories, used in standard LNA, lead to systematic information loss.
• A reaction-based Gaussian mutual information approach avoids this information loss.

Conclusions:

• The choice of trajectory description (reaction-based vs. state-based) significantly impacts mutual information calculations in biochemical systems.
• Standard LNA-based mutual information inherently loses information due to its state-based nature.
• A reaction-based Gaussian mutual information offers a more accurate quantification of information transfer.