Information Thermodynamics for Time Series of Signal-Response Models.

Area of Science:

• Information Thermodynamics
• Stochastic Dynamical Systems
• Time Series Analysis

Background:

• Entropy production in stochastic systems connects to Bayesian network causal structures.
• Previous work formalized this for bipartite systems using fluctuation theorems.

Purpose of the Study:

• Introduce information thermodynamics for general (non-bipartite) time series.
• Investigate the link between irreversibility and information in incomplete causal representations.
• Quantify signaling efficiency in biological systems using time series irreversibility.

Main Methods:

• Developed information thermodynamics for non-bipartite time series.
• Introduced a backward transfer entropy lower bound for conditional time series irreversibility.
• Analyzed linear signal-response models and nonlinear receptor-ligand systems.

Main Results:

• Established that irreversibility arises from incomplete causal representations.
• Demonstrated a lower bound for time series irreversibility based on feedback absence.
• Showed time series irreversibility quantifies signaling efficiency in a biological model.

Conclusions:

• Incomplete causal representations are key to the irreversibility-information link in time series.
• Backward transfer entropy provides a measure of irreversibility due to missing feedback.
• Time series irreversibility offers a novel metric for biological signaling efficiency.