A comparison of weighted ensemble and Markov state model methodologies

Area of Science:

• Computational Chemistry
• Molecular Dynamics Simulations
• Biophysics

Background:

• Molecular dynamics (MD) simulations are crucial for understanding reaction mechanisms and rates.
• Brute force MD is computationally expensive for long time-scale events.
• Markov State Models (MSMs) and Weighted Ensemble (WE) methods accelerate MD simulations.

Purpose of the Study:

• To compare the accuracy and efficiency of MSMs and WE methods for computing reaction rates.
• To analyze the sensitivity of these methods to the definition of macro-states.
• To identify the most reliable approach for rate estimation in high-dimensional conformational spaces.

Main Methods:

• Utilized Markov State Models (MSMs) and Weighted Ensemble (WE) methods.
• Employed clustering of microscopic configurations into macro-states for both approaches.
• Introduced Accelerated Weighted Ensemble (AWE) using 'colors' for flux computation.
• Performed numerical experiments on alanine dipeptide and penta-alanine systems.

Main Results:

• MSMs introduce significant, boundary-dependent biases in reaction rate computations.
• WE's rate predictions are less sensitive to macro-state definition compared to MSMs.
• AWE demonstrates reliable flux estimation across varying macro-state definitions.
• MSMs excel at visualizing metastable sets and overall dynamics.

Conclusions:

• AWE provides more reliable reaction rate estimations with less effort in macro-state definition.
• MSMs are valuable for understanding metastable states and visualizing dynamics.
• AWE is recommended for accurate rate predictions in complex molecular systems.