Protein model building for intermediate-resolution cryo-EM maps by integrating evolutionary and experimental...

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Accurate protein model building in intermediate-resolution cryo-EM maps is crucial but challenging.
• Deep learning models like AlphaFold2 provide accurate structures but often conflict with experimental EM data.
• Integrating predicted structures with experimental maps requires advanced computational methods.

Purpose of the Study:

• To develop an automated method, CryoEvoBuild, for improving protein model building in intermediate-resolution cryo-EM maps.
• To effectively integrate evolutionary and experimental information with AlphaFold2 predictions.
• To address the challenges posed by discrepancies between predicted and experimental structures.

Main Methods:

• CryoEvoBuild employs a novel domain-wise fitting, refinement, assembly, and rebuilding pipeline.
• A recycling framework guided by AlphaFold2 is implemented for iterative improvement.
• The method integrates evolutionary information with experimental cryo-EM map data.

Main Results:

• CryoEvoBuild significantly improves the accuracy of AlphaFold2-predicted protein structures.
• Benchmarking on 117 diverse maps (4.0-10.0 Å) demonstrates superior performance.
• The method outperforms existing state-of-the-art approaches like EMBuild and phenix.dock_and_rebuild.

Conclusions:

• CryoEvoBuild offers an effective solution for accurate protein model building from intermediate-resolution cryo-EM data.
• The integration of evolutionary and experimental information enhances model quality.
• This automated approach advances structural determination in cryo-EM studies.