Automated refinement of macromolecular structures at low resolution using prior information

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Low-resolution structural data presents challenges due to a low signal-to-noise ratio.
• Accurate atomic model refinement requires complementary information and optimized protocols.
• Existing methods for selecting homologous structures and prior information can be suboptimal.

Purpose of the Study:

• To develop and validate an automated pipeline for refining low-resolution macromolecular structures.
• To identify optimal refinement protocols and strategies for selecting homologous structures.
• To improve the quality of atomic models derived from low-resolution data.

Main Methods:

• Implementation of the Low-Resolution Structure Refinement (LORESTR) pipeline.
• Automated selection of homologous structures via BLAST search or user input.
• Generation of restraints using ProSMART and REFMAC5.
• Auto-detection of twinning and selection of optimal scaling and solvent parameters.
• Execution of multiple refinement instances with varying parameters to identify the best protocol.

Main Results:

• The LORESTR pipeline successfully automates the selection of homologous structures and refinement strategies.
• The pipeline demonstrated improved R factors, geometry, and Ramachandran statistics.
• 94% of tested low-resolution Protein Data Bank (PDB) cases showed improvement after refinement with LORESTR.
• The pipeline effectively handles issues like twinning and optimizes scaling and solvent parameters.

Conclusions:

• The automated LORESTR pipeline provides an effective solution for refining low-resolution macromolecular structures.
• LORESTR enhances model accuracy and reliability by optimizing refinement protocols and homologous structure selection.
• This pipeline significantly improves the quality of structural models derived from challenging low-resolution datasets.