CASP15 cryoEM protein and RNA targets: refinement and analysis using experimental maps

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Traditional CASP assessments rely on experimental structures, which can contain errors, especially cryo-EM structures.
• Evaluating protein models against cryo-EM density maps offers a more direct assessment of prediction accuracy.

Approach:

• Developed a novel method to rank protein chain prediction quality by rigid fitting to experimental cryo-EM density maps.
• Created an automated hierarchical refinement pipeline specifically for cryo-EM density maps.
• Applied refinement to top CASP15 protein and RNA predictions to improve atomic model building.

Key Points:

• Direct evaluation using cryo-EM density maps correlated well with CASP scores, indicating high overall model accuracy.
• Local assessment revealed inaccuracies in predicted side chains and persistent challenges in modeling loops.
• Refinement in cryo-EM maps successfully improved structures, demonstrating the utility of CASP predictions as starting points for atomic modeling.

Conclusions:

• Cryo-EM density map evaluation provides a robust assessment of protein model quality, complementing traditional methods.
• Automated refinement pipelines in cryo-EM maps are effective for improving CASP predictions for both protein and RNA.
• Challenges in loop modeling persist, highlighting areas for future development in protein structure prediction and refinement.