NAME: Randy Read

AFFILIATION: University of Cambridge, UK

CONTACT: rjr27@cam.ac.uk

 

TITLE: "Coping with translational NCS in Phaser"

ABSTRACT: Experience with likelihood-based molecular replacement in Phaser has shown that the likelihood targets provide increased sensitivity to the correct solution than more traditional molecular replacement targets. The increased sensitivity comes from accounting for the effects of different sources of error in the probability distributions underlying the likelihood targets.

 

However, increased sensitivity comes at a price. The likelihood targets are also more sensitive to the influences of effects that have not been modeled in the probability distributions. In their simplest form, the probability distributions are derived by assuming that the atoms are all equally well ordered, isotropic, and randomly distributed relative to the Bragg planes in the unit cell. We have already found that the presence of significant overall anisotropy can make it much more difficult to solve structures by molecular replacement. This can be overcome by correcting the data to remove anisotropy, which significantly improves Phaser's success rate in such cases.

 

Translational non-crystallographic symmetry violates the assumption that the atoms are randomly distributed, and cases showing translational NCS account for a large proportion of the structures that Phaser is not able to solve, even with a reasonable molecular replacement model. In the presence of translational NCS, the diffraction data show large oscillations in intensity over reciprocal space, which are not modeled in the current probability distributions.

 

We have recently derived modified likelihood targets that account for the effects of translational NCS, and have implemented them in Phaser. Preliminary results are promising: the signal of correct solutions is improved, and it is now possible to solve some of the structures that eluded previous versions of Phaser.