Cryo-electron microscopy ensemble optimization using individual particles and physical constraints.

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Biomolecules are dynamic, and understanding their conformational states is key to their function.
• Cryo-electron microscopy (cryo-EM) determines biomolecule structures at atomic resolution.
• Current methods struggle to simultaneously infer structural heterogeneity and population weights.

Purpose of the Study:

• To develop a novel method for cryo-EM ensemble optimization.
• To directly infer optimal structures and population weights from cryo-EM images.
• To enable simultaneous inference of structural heterogeneity.

Main Methods:

• Utilized Bayesian optimization techniques for cryo-EM ensemble optimization.
• Iteratively optimized structures and weights using cryo-EM particle images.
• Employed projected gradient descent inspired approach for physical prior projection.

Main Results:

• Successfully recovered structures and population weights across various systems, from toy models to large proteins.
• Demonstrated robustness even when the number of inferred states did not match actual metastable states.
• Validated the method on real cryo-EM data.

Conclusions:

• The developed cryo-EM ensemble optimization method accurately infers structural and population heterogeneity.
• This approach is suitable for flexible biomolecules with complex conformational landscapes.
• Paves the way for advanced analysis of biomolecular dynamics using cryo-EM.