CASP15 cryo-EM protein and RNA targets: Refinement and analysis using experimental maps

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Protein structure prediction is crucial for understanding biological function.
• Current assessments like CASP rely on experimental structures, which are themselves models with inherent subjectivity.
• Evaluating predictions directly against experimental density maps offers a more objective approach.

Purpose of the Study:

• To develop and validate a method for assessing protein structure predictions using experimental density maps.
• To compare the accuracy of predictions against reference structures and density maps.
• To explore the utility of predictions as starting points for atomic model refinement in cryo-electron microscopy (cryo-EM).

Main Methods:

• A novel method was employed to rank protein chain predictions based on their fit into experimental density maps.
• Fit-based rankings were correlated with Critical Assessment of protein Structure Prediction (CASP) assessment scores.
• Selected high-quality predictions for protein and RNA targets underwent automated refinement.

Main Results:

• The fit-based ranking method showed good correlation with CASP scores.
• Evaluations against density maps indicated high accuracy of predictions, occasionally exceeding reference structures.
• Automated refinement of CASP15 predictions yielded structures closely resembling reference targets, even with significant conformational changes.
• Side-chain positioning in predictions was found to be occasionally poor at high resolution.

Conclusions:

• Direct evaluation using density maps provides a robust assessment of protein structure prediction accuracy.
• CASP-assessed predictions serve as valuable starting points for building atomic models in cryo-EM maps for both proteins and RNA.
• Loop modeling remains a challenge, and integrating model-fit to density may help identify flexible regions.