Verification: model-free phasing with enhanced predicted models in ARCIMBOLDO_SHREDDER

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• Protein structure prediction methods are increasingly accurate, achieving accuracy comparable to experimental structures of homologous proteins.
• Predicted models, even those with discrepancies, often contain highly accurate regions suitable for crystallographic phasing.
• Traditional molecular replacement phasing relies on experimental structures, limiting its application with predicted models.

Purpose of the Study:

• To introduce and evaluate the "predicted_model" mode in ARCIMBOLDO for utilizing predicted protein structures in crystallographic phasing.
• To develop a robust pipeline for fragment-based phasing using predicted models, addressing their unique characteristics.
• To mitigate potential issues like model bias by critically assessing the experimental data's contribution.

Main Methods:

• ARCIMBOLDO's "predicted_model" mode processes predicted structures by converting B-values, removing unstructured regions, and decomposing structural units.
• The software systematically probes predicted models against experimental data to optimize their use in phasing.
• A phased approach is employed: first, checking if the predicted model itself is a solution, then extracting fragments if necessary, followed by model-free verification and expansion using SHELXE and ALIXE.

Main Results:

• Predicted models, even with domain movements, contain accurate regions enabling fragment-based phasing with ARCIMBOLDO.
• The "predicted_model" mode effectively handles the peculiarities of predicted structures, making preliminary treatments unnecessary.
• The developed procedure bypasses the need for a molecular replacement search model, relying instead on inferences from the predicted model.

Conclusions:

• The "predicted_model" mode in ARCIMBOLDO provides a powerful and direct method for utilizing predicted protein structures in crystallographic phasing.
• This approach enhances the utility of protein structure predictions, particularly in scenarios where experimental structures are unavailable.
• The method offers a robust alternative to traditional molecular replacement, especially when dealing with imperfect predicted models.