Refmac

Refmac performs maximum-likelihood refinement of macromolecular structures against X-ray diffraction data to improve the accuracy and reliability of crystallographic models.

Key Features:

• Likelihood Functions: Employs likelihood functions for amplitudes or intensities and adapts to twinning and SAD/SIRAS experimental data.
• Restraints and Parameterization: Provides multiple restraint classes and parameterization options, including secondary-structure restraints, homologous-structure restraints, automatic global and local NCS (Non-Crystallographic Symmetry) restraints, "jelly-body" restraints, and long-range restraints on atomic displacement parameters (ADPs) based on the Kullback-Leibler divergence, effective at resolutions as low as 4 Å.
• TLS Parameterization: REFMAC5 implements TLS (Translation-Libration-Screw) parameterization for rapid refinement of anisotropic ADPs when high-resolution data are available.
• Automated Twinning Refinement: Performs automated refinement in the presence of twinning.
• Mathematical Foundation: Extends the likelihood function to incorporate prior phase information and experimental standard uncertainties, allowing for varying error models across different parts of a structure.
• Sigma(A) Estimation: Estimates sigma(A) using "free" reflections for error analysis.

Scientific Applications:

• Macromolecular crystallography: Refines macromolecular structures across the full resolution spectrum using maximum-likelihood approaches.
• Twinned and experimental phasing data: Applicable to refinement with twinned data and SAD/SIRAS experimental datasets.
• Bacterial alpha-amylase: Has been applied to refine the structure of bacterial alpha-amylase.
• Cytochrome c': Has been applied to refine the structure of cytochrome c'.
• Cross-linked insulin: Has been applied to refine the structure of cross-linked insulin.
• Oligopeptide binding protein: Has been applied to refine the structure of oligopeptide binding protein.
• Comparison to least-squares refinement: Maximum-likelihood residuals obtained with Refmac consistently outperform those from traditional least-squares refinement.

Methodology:

Computational methods explicitly include maximum-likelihood refinement using likelihood functions for amplitudes or intensities, handling of twinning and SAD/SIRAS data, restraint schemes (secondary-structure, homologous-structure, global/local NCS, "jelly-body"), long-range ADP restraints based on the Kullback-Leibler divergence, TLS parameterization, extension of the likelihood to include prior phase information and experimental standard uncertainties, and sigma(A) estimation from "free" reflections.