Research Article: Making a difference in multi-data-set crystallography: simple and deterministic data-scaling/selection methods

Date Published: July 01, 2020

Publisher: International Union of Crystallography

Author(s): Greta M. Assmann, Meitian Wang, Kay Diederichs.

http://doi.org/10.1107/S2059798320006348

Abstract

Fast and deterministic methods, based on multi-dimensional scaling and weighted ΔCC1/2, to reject non-isomorphous data sets in multi-data-set crystallography are described, and their successful application to several difficult projects where phasing is based on weak anomalous signal is reported.

Partial Text

Obtaining large crystals and solving the phase problem remain the major bottlenecks in macromolecular crystallography. To overcome the problem of a lack of sufficiently large crystals for collecting a complete data set with little radiation damage, multi-crystal data-collection strategies were established early on and have recently experienced a renaissance (Kendrew et al., 1960 ▸; Dickerson et al., 1961 ▸; Ji et al., 2010 ▸; Liu et al., 2012 ▸; Akey et al., 2014 ▸; Huang et al., 2018 ▸). Serial synchrotron crystallography (SSX; Rossmann, 2014 ▸) typically collects a few degrees of rotation data from each of the small crystals available to the experimenter.

The paradigm of multi-data-set scaling and merging is that averaging reduces random errors in the merged intensities, according to the laws of error propagation. However, this assumes that the intensity differences of different data sets with respect to the unknown ‘true’ intensities are unrelated, which does not hold in the case of non-isomorphism. If the data sets have systematic differences, merging introduces systematic errors that are not necessarily reduced by averaging. Without non-isomorphism, the accuracy of the merged data is identical to their precision, for which a number of crystallographic indicators exist. However, in the presence of systematic differences (the crystallographic term for which is ‘non-isomorphism’), the accuracy of the merged data is worse than their precision by an amount that is difficult to quantify, but which can be large enough to prevent structure solution.

 

Source:

http://doi.org/10.1107/S2059798320006348

 

Leave a Reply

Your email address will not be published.