Validation of Markov state models using Shannon's entropy.

Area of Science:

• Computational chemistry
• Statistical mechanics
• Molecular dynamics

Background:

• Markov state models (MSMs) are crucial for analyzing molecular simulation dynamics.
• Validating the Markov assumption is essential for the accuracy of MSMs.
• Current validation methods like the eigenvalue method have limitations.

Purpose of the Study:

• To present a novel procedure for validating the Markov assumption in MSMs.
• To utilize information theory and Shannon entropy for this validation.
• To offer a method for identifying and improving states that violate the Markovian property.

Main Methods:

• Development of an entropy-based method for Markov assumption validation.
• Application of the entropy method to a model system.
• Comparison of the entropy method with the traditional eigenvalue method.

Main Results:

• The entropy method provides a robust way to assess the Markovian nature of states.
• The method successfully identified states deviating from Markovian behavior.
• The entropy method offers a direct comparison with the eigenvalue method, demonstrating its utility.

Conclusions:

• The proposed entropy-based method is a valuable tool for validating MSMs.
• This approach enhances the reliability of kinetic models derived from molecular simulations.
• Identifying and refining non-Markovian states leads to more accurate molecular dynamics analyses.