Refinement of severely incomplete structures with maximum likelihood in BUSTER-TNT

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Macromolecular refinement is crucial for understanding biological structures.
• Accurate structural models are essential for drug discovery and biological mechanism elucidation.
• Existing refinement packages face challenges with incomplete atomic models.

Purpose of the Study:

• To introduce BUSTER-TNT, a maximum-likelihood macromolecular refinement package.
• To enhance the process of structural model building and completion.
• To address limitations in refining structures with significant missing atomic information.

Main Methods:

• BUSTER assembles structural models, scales structure factors, and computes model likelihood.
• TNT manages stereochemistry, restraints, and coordinate/parameter shifts.
• BUSTER models missing structural regions as low-resolution probability distributions in real space.
• A 2D Gaussian model describes the structure-factor distribution in reciprocal space, incorporating errors from atomic, bulk-solvent, and missing-structure components.

Main Results:

• BUSTER-TNT improves the accuracy of overall scale factors.
• Bias in atomic model refinement is reduced by accounting for scattering from missing structures.
• The statistical model provides a more robust error estimation than traditional methods.
• Selective density modification can be performed in regions of unbuilt structure.

Conclusions:

• BUSTER-TNT effectively models and refines macromolecular structures, even with incomplete atomic models.
• The package enhances structural model completion and accuracy through advanced statistical modeling.
• This advancement aids in solving challenging crystallographic datasets and advancing structural biology research.