Towards automated crystallographic structure refinement with phenix.refine

Abstract

phenix.refine is a program within the PHENIX package that supports crystallographic structure refinement against experimental data with a wide range of upper resolution limits using a large repertoire of model parameterizations. It has several automation features and is also highly flexible. Several hundred parameters enable extensive customizations for complex use cases. Multiple user-defined refinement strategies can be applied to specific parts of the model in a single refinement run. An intuitive graphical user interface is available to guide novice users and to assist advanced users in managing refinement projects. X-ray or neutron diffraction data can be used separately or jointly in refinement. phenix.refine is tightly integrated into the PHENIX suite, where it serves as a critical component in automated model building, final structure refinement, structure validation and deposition to the wwPDB. This paper presents an overview of the major phenix.refine features, with extensive literature references for readers interested in more detailed discussions of the methods.

Figure 1

Flowchart of structure refinement as implemented in phenix.refine. The execution of some steps is subject to user-defined options. The main refinement body (shown with the gray arrow) is called a macro-cycle and is repeated several times. See text for details.

Figure 2

Illustration of typical scenarios for occupancy refinement that phenix.refine handles automatically. (a) Residue having several alternative conformations marked with altLoc identifiers (two in this example, A and B). It is essential that all conformers have identical chain identifiers and residue numbers, while residue names can be different as shown in example (e). All atoms within each conformer must have identical occupancies. The sum of occupancies over all conformers is constrained to 1. (b) Single atoms with occupancy not equal to 0 or 1. (c) Exchangeable H/D sites (used in refinement against neutron data collected from partially deuterated sample). (d) Single-residue molecule with identical occupancies for all atoms (but not equal to 1 or 0). A user can overwrite this behavior or/and define constraints for any number of selected atoms or groups of atoms.

Figure 3

The graphical user interface (GUI) for phenix.refine. (a) Configuration tab showing the refinement strategy and commonly used restraint and optimization settings. (b) Display of results, including summary of output files, tables and graphs of statistics and links to molecular-graphics software.

Figure 4

Polygon images (Urzhumtseva et al., 2009 ▶) before (left) and after (right) re-refinement in phenix.refine for structures 1eic, 1g2y, 2elg and 2ppn. In all cases the polygon computed for structures before re-refinement in phenix.refine indicates one or more problems, for example high R _free and R _work and too small bond r.m.s.d. for 1eic or very high R factors and geometry deviations for 2elg (vertices are on the furthermost end of the histogram bar). Re-refinement in phenix.refine resulted in polygon vertices moved towards the center (squeezing the polygon) in most cases, indicating improvement of the corresponding model characteristics.

Figure 5

Selected examples of (2mF _obs − DF _model, ϕ_model) nuclear density map improvement after re-refinement of structure 1c57 (neutron data). Left, original structure; right, after re-refinement in phenix.refine. Maps are contoured at the 1.5σ level. Note the improved orientation of exchangeable H/D atoms at Ser and Tyr O atoms. The systematic lack of density around H atoms is a consequence of the negative scattering length of H atoms and related density-cancellation effects (Afonine, Mustyakimov et al., 2010 ▶).

Figure 6

Structures after refinement of a severely distorted model, shown in (a), using different refinement protocols: (b) dual-space refinement, (c) refinement using minimization only and (d) combined refinement using minimization and simulated annealing. The best available refined model is shown in gray in all panels.

Figure 7

(a) Ensemble of structures illustrating the outcome of 100 identical simulated-annealing refinement runs apart from the random seed. (b) The distribution of R _work and R _free corresponding to each structure of the ensemble.