Comparing Multislice Projections of MD Simulations with CryoEM Exposes Structural Prediction Errors.

Area of Science:

• Structural Biology
• Computational Biophysics
• Biophysics

Background:

• Cryo-electron microscopy (cryoEM) determines atomic-resolution structures, while molecular dynamics (MD) simulations predict molecular motion.
• Direct experimental validation of MD predictions remains challenging.
• MD simulations are crucial for interpreting cryoEM data and understanding biomolecular dynamics.

Purpose of the Study:

• To develop and validate a method for directly comparing MD simulation outputs with experimental cryoEM data.
• To assess the accuracy of MD simulations, particularly coarse-grained (CG) models, in reproducing experimental membrane structures.
• To identify sources of error in MD simulations for lipid membrane dynamics.

Main Methods:

• Utilized a physics-based multislice wave propagation algorithm to project MD trajectories (all-atom and CG) into simulated cryoEM 2D images and 3D reconstructions.
• Compared simulated cryoEM images with experimental images of lipid membranes with varying curvature and composition.
• Analyzed bilayer dimensions and membrane thickness from simulated and experimental data.

Main Results:

• MD simulations qualitatively reproduced the fluidity and contrast of biological membranes observed in cryoEM.
• All-atom (AA) MD simulations accurately predicted bilayer dimensions for simple lipid bilayers.
• Martini3 CG-MD simulations failed to predict membrane thickness changes in complex lipid mixtures and high-curvature membranes, with errors attributed to polyunsaturated lipid tails and cholesterol.
• Discrepancies between cryoEM and small-angle X-ray scattering (SAXS) data were explained.

Conclusions:

• The developed multislice image simulation method enables direct comparison between MD simulations and cryoEM experiments.
• Current CG-MD force fields, like Martini3, require refinement, especially for simulating complex lipid membranes containing cholesterol and polyunsaturated lipids.
• This approach provides a pathway for improving MD force fields through direct experimental validation, leading to more accurate predictions of biomolecular dynamics.