Best for the Eye, Not for the Algorithm: Anisotropy in Fitting Atomic Models in Cryo-EM

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Current cryo-EM atomic model refinement often treats all voxels in density maps equally.
• Fourier coefficient uncertainty is anisotropic due to signal-to-noise ratio variations and particle image distribution.
• Fitting models directly to particle images is theoretically superior but computationally intensive.

Purpose of the Study:

• To develop a computationally feasible method for atomic model refinement in cryo-EM that accounts for anisotropic uncertainty.
• To demonstrate the equivalence of fitting to weighted volumes and fitting directly to particle images under specific conditions.
• To propose an implementation strategy for existing cryo-EM software.

Main Methods:

• Developed a method to fit atomic models to weighted cryo-EM density maps, approximating direct fitting to particle images.
• Utilized proxies to capture uncertainty and distortions in Fourier coefficients.
• Implemented the method in a modified Servalcat program using data from standard RELION runs.

Main Results:

• Showed that fitting to weighted volumes can be equivalent to fitting directly to particle images.
• Demonstrated the feasibility of incorporating uncertainty information into atomic model refinement.
• Successfully applied the proof-of-concept method using standard cryo-EM data processing outputs.

Conclusions:

• Atomic model refinement in cryo-EM can be significantly improved by accounting for anisotropic uncertainty in Fourier coefficients.
• Fitting to weighted volumes offers a practical approach to leverage particle image information.
• The proposed method is adaptable to existing cryo-EM pipelines, enhancing structural model accuracy.