tICA-Metadynamics for Identifying Slow Dynamics in Membrane Permeation

Area of Science:

• Computational chemistry and biophysics
• Machine learning applications in molecular dynamics
• Membrane transport mechanisms

Background:

• Accurate prediction of drug permeation through cell membranes is crucial for pharmaceutical development.
• Identifying rate-limiting molecular dynamics (reaction coordinates) is a major challenge in free energy calculations.
• Machine learning offers novel approaches to uncover slow dynamics in complex molecular systems.

Approach:

• Applied time-lagged independent component analysis (tICA), an unsupervised dimensionality reduction technique.
• Integrated tICA with well-tempered metadynamics simulations for molecular dynamics.
• Investigated the permeation of trimethoprim through a multicomponent membrane model.

Key Points:

• tICA-metadynamics identified effective translational and orientational collective variables (CVs), improving simulation convergence by approximately 1.5 times.
• Periodic boundary crossing artifacts in translational CVs were corrected by using absolute molecular features.
• Convergence of tICA CVs was achieved within five membrane crossings, with data reweighting essential to prevent translational CV deviations.

Conclusions:

• The tICA-metadynamics approach provides an efficient method for determining essential collective variables for membrane permeation.
• This strategy enhances the accuracy and speed of molecular simulations for drug discovery and development.
• Addressing simulation artifacts and employing appropriate data processing are key for reliable free energy calculations.