view article
Figure 5
Phase determination for test case 5hk7 (Arrigoni et al., 2016BB903) (solvent fraction 0.833, 2.95 Å resolution). (a) and (b) The trajectories of two successful runs. (c) The trajectory of an unsuccessful run. These runs were initiated with the same molecular envelope [Fig. 4[link](b)] but with different random phase sets. Across the first 7200 iterations, the discontinuities apparent every 240 iterations are associated with the steady reduction in data apodization, while the discontinuities apparent every 60 iterations are associated with switching of the DM algorithm parameter β between two values. In parts (a)–(c), plotted from top to bottom, as a function of iteration, are the convergence indicator of the DM algorithm (δDM); the variance in the solvent region; the Wasserstein distance between reconstructed and reference histograms in the protein region; the correlation between reconstructed and measured Fourier amplitudes; the weighted mean absolute difference between reconstructed and model phases; and the correlation between the reconstructed and model envelope. The metrics that could be followed during determination of an unknown structure are shown in black, while the metrics that assess agreement with the known solution are shown in blue. (d) Consensus electron-density map determined from 10 independent runs, with the molecular structure of the bacterial sodium channel (5hk7) shown in ribbon representation.

IUCrJ
Volume 9| Part 5| September 2022| Pages 648-665
ISSN: 2052-2525
https://doi.org/10.1107/S2052252522006996