The X-ray crystallography phase problem solved thanks to AlphaFold and RoseTTAFold models: a case-study report

Area of Science:

• Structural Biology
• Computational Biology
• Molecular Biology

Background:

• Deep learning models like AlphaFold and RoseTTAFold are revolutionizing protein structure prediction.
• Accurate protein models are critical for solving the phase problem in X-ray crystallography via molecular replacement.
• Traditional methods failed to solve the crystal structure of a protein in the nonsense-mediated mRNA decay pathway.

Purpose of the Study:

• To evaluate the utility of deep learning protein structure models in solving challenging crystal structures.
• To demonstrate the application of AlphaFold and RoseTTAFold in overcoming the phase problem.
• To compare the performance of AlphaFold and RoseTTAFold models in molecular replacement.

Main Methods:

• Collected 2.45 Å resolution diffraction data for the target protein.
• Attempted structure solution using isomorphous replacement, anomalous diffraction, and molecular replacement.
• Utilized AlphaFold and RoseTTAFold predicted models for molecular replacement.

Main Results:

• All traditional structure solution methods failed for the target protein.
• The crystal structure was successfully solved using molecular replacement with AlphaFold and RoseTTAFold models.
• The AlphaFold model provided a superior search model compared to RoseTTAFold for this specific case.

Conclusions:

• Deep learning models significantly facilitate solving the X-ray crystallography phase problem.
• AlphaFold and RoseTTAFold are powerful tools for accelerating structural biology research.
• Careful generation of search models is crucial for successful molecular replacement outcomes.