Macromolecular refinement using shift field optimization and regularization

Description

For half a century the refinement of atomic model parameters to best explain the observed diffraction pattern has been fundamental to the process of crystallographic structure solution. This process has traditionally been carried out by the optimization of individual atomic parameters, with the use of stereochemical restraints to maintain plausible model geometry, particularly when data resolution is poor. However the data are often too poor to reliably indicate how individual atoms should be moved, and as a result the refinement calculation becomes a protracted battle between the noisy data pulling atoms in different directions and the restraints which are trying to maintain model geometry. This limits both the speed and radius of convergence of the calculation.

Shift field refinement is an approach in which smoothly varying shifts to the model are determined from large regions of the unit cell, where the region size is determined by the resolution of the data and the size of the features being refined. This enables refinement to capture large domain shifts at low resolution, and to be applied at any resolution with rapid convergence. The method can also be used to refine a map against a set of diffraction observations, even in the absence of an atomic model.

We also demonstrate how the incorporation of a separate regularization step can be used to improve the refinement results by allowing more cycles of refinement to be run without the risk of model degradation due to accumulated model distortions. This in turn leads to further improvement in the refinement results.

Period	6 Jan 2022
Event title	CCP4 Study Weekend 2022: Current trends in macromolecular model refinement and validation
Event type	Conference
Degree of Recognition	International